This is only a preview of the December 2026 issue of Silicon Chip. You can view 37 of the 104 pages in the full issue, including the advertisments. For full access, purchase the issue for $10.00 or subscribe for access to the latest issues. Articles in this series:
Items relevant to "Advanced Cable Analyser":
Items relevant to "Inrush Current Limiter":
Articles in this series:
Items relevant to "NeoPixel RGB LED Analog Clock, Part 2":
Items relevant to "The Magic Candle":
Purchase a printed copy of this issue for $15.00. |
The VERY BEST DIY Projects!
Inrush
Current Limiter
THE HISTORY OF
AMD
Pt 2
MAGIC CANDLE
Advanced
Cable Analyser
DECEMBER 2026
ISSN 1030-2662
12
9 771030 266001
$1500* NZ $1590
INC GST
INC GST
Contents
Vol.39, No.12
December 2026
14 The History of AMD, Part 2
We resume the story in 1993, with AMD having begun to design more of
their own chips such as the Am486. Following on from that is the creation
of the Athlon CPU line and AMD’s entry into the discrete graphics market.
By Dr David Maddison, VK3DSM
Electronics feature
The History of
AMD
Part 2
Page 14
50 Motor Control, Part 3
This series looks at many of the different motor types in depth, explaining
how they work and can be controlled. This month, we cover stepper motors,
which are useful in applications that require precise position control.
By Andrew Levido
Electronics theory
66 Working with AVR & PIC Fuses
In this short series, we investigate the configuration ‘fuses’, hardware and
software in 8-bit PIC and AVR parts. We start by looking at the PIC16F84
microcontroller and what changing its configuration fuses achieves.
Part 1 by Tim Blythman
Firmware development
94 STC Bantam model A141
Named after a small chicken, the Bantam model A141 is a four-valve
mantel radio manufactured by STC (Standard Telephones and Cables).
While smaller, it is very similar electrically to the AWA 500M.
Br Ian Batty
Vintage Radio
30 Advanced Cable Analyser
This Cable Analyser works on cables with up to eight wires over a wide
variety of connectors that you can choose. It can also be run in continuous
mode to check for intermittent faults, or against a reference cable.
By Richard Palmer
Test equipment project
45 Inrush Current Limiter
Using this project lets you limit the inrush current for most appliance mains
power supplies, including many power amplifiers, saving you from tripping
circuit breakers. It will work up to a maximum load current of 8A.
By Phil Prosser
Power supply project
56 NeoPixel RGB LED Analog Clock
Our ‘Analog’ Clock has built-in Wi-Fi and NTP support, can be configured
remotely and uses addressable RGB LEDs. This month we provide the
assembly instructions, explaining how to use it and what the firmware does.
Part 2 by Nicholas Vinen
Clock project
78 The Magic Candle
Impress your friends and family by lighting it like a candle with a flame from
a match or lighter and then blowing it out with your breath! This design was
inspired by the similar hobby projects that appeared in the 1950s to 1970s.
By Dr Hugo Holden
Toy/gimmick project
Part 2: page 56
NeoPixel
RGB LED
‘Analog’ Clock
2
Editorial Viewpoint
4
Mailbag
73
Circuit Notebook
76
Online Shop
88
Serviceman’s Log
99
DCC Kits
100
Subscriptions
101
Ask Silicon Chip
103
Market Centre
104
Advertising Index
104
Notes & Errata
1. Hybrid switchmode / charge pump
power supply
2. Accurate 1Hz oscillator
3. Simulating crystals in LTspice
SILICON
SILIC
CHIP
www.siliconchip.com.au
Publisher/Editor
Nicholas Vinen
Technical Editor
John Clarke – B.E.(Elec.)
Technical Staff
Bao Smith – B.Sc.
Tim Blythman – B.E., B.Sc.
Advertising Enquiries
(02) 9939 3295
adverts<at>siliconchip.com.au
Regular Contributors
Allan Linton-Smith
Dave Thompson
David Maddison – B.App.Sc. (Hons 1),
PhD, Grad.Dip.Entr.Innov.
Geoff Graham
Associate Professor Graham Parslow
Dr Hugo Holden – B.H.B, MB.ChB.,
FRANZCO
Ian Batty – M.Ed.
Phil Prosser – B.Sc., B.E.(Elec.)
Cartoonist
Louis Decrevel
loueee.com
Founding Editor (retired)
Leo Simpson – B.Bus., FAICD
Silicon Chip is published 12 times
a year by Silicon Chip Publications
Pty Ltd. ACN 626 922 870. ABN 20
880 526 923. All material is copyright ©. No part of this publication
may be reproduced without the written
consent of the publisher.
Subscription rates (Australia only)
6 issues (6 months): $77.50
12 issues (1 year):
$145
24 issues (2 years):
$270
Online subscription (Worldwide)
6 issues (6 months): $55
12 issues (1 year):
$105
24 issues (2 years):
$200
For overseas rates, see our website or
email silicon<at>siliconchip.com.au
* recommended & maximum price only
Postal address: PO Box 194,
Matraville, NSW 2036.
Phone: (02) 9939 3295.
ISSN: 1030-2662
Printing and Distribution:
Editorial Viewpoint
I seriously underestimated AI coding
Like it or not, AI is here to stay. I am neither in the
“AI is horrible” nor the “AI is amazing” camp. At least,
that was true until recently.
When I first became aware that AI was becoming more
than just science fiction, I started trying out ChatGPT
to see what the fuss was about. That was in early 2023.
I was very impressed with its capabilities from the
start, but it was clear that its uses at the time were
limited. It tended to “hallucinate” quite a bit and confidently made mistakes.
One of its biggest flaws was that it often wouldn’t admit when it didn’t know
something – it would just make up a plausible-sounding answer (still a problem
today). It also had quite out-of-date information.
However, things clearly improved over time. The developers added the ability
for AI to perform web searches and access websites to get fresh information.
They introduced new models that could spend time reasoning through a problem
before giving an answer, rather than simply blurting out their first response.
So, I guess you could call me lukewarm on AI. I could see its uses, but I
didn’t see it taking over whole industries or replacing jobs wholesale. That is,
until I started using it seriously for programming tasks.
I decided to give OpenAI’s Codex a go, since it was already included in the
roughly $30 monthly subscription we had to evaluate ChatGPT. I already had
VS Code installed, so I installed the Codex extension (with ChatGPT’s help –
it took about one minute) and logged into my account.
I was not prepared for what happened next.
I typed up a specification for a fairly sophisticated piece of software I wanted
to build. It involved graphics, sound, route planning, complex rule sets and a
whole lot of other features. By my estimation, it would have taken a small but
capable programming team at least a week to produce a prototype. I handed
over the document and asked the Codex agent to start implementing it.
When I came back a couple of hours later, it had done it. I had a fully working
prototype that did pretty much everything I had asked. Over the next few days,
by asking it to add more features, improve the performance and refine the data,
I managed to get the project into a usable state. Doing all that myself would
have taken me at least several months. I couldn’t believe it!
Not only did I have working software, it was cross-platform and included
regression and performance test suites. That meant each time it made changes
to the code, I could check that everything it had written so far still worked and
that it hadn’t slowed down (or if it had, by how much).
That’s not to say everything went perfectly. There were occasional bugs,
misunderstandings and things that needed fixing. I still had to test the software,
make decisions about how it should work and check that the results were
correct. But the sheer amount of useful work the AI could do with relatively
little supervision was astonishing.
This is absolutely revolutionary. Some may call it “vibe coding” but it is
like having a team of remarkably capable programmers at your disposal for
a few dollars per day. The $30 plan has its limits, but bump that up to, say,
$150 or $300 and those limits become much less restrictive. Some users reckon
Anthropic’s competing Claude Code is even better! I will definitely have to
investigate that.
For electronics enthusiasts, the implications are enormous. If you know what
you want to achieve but lack the time or programming expertise to implement
it, these tools make previously impractical projects achievable – I would even
say easy.
I’ll write a more detailed article explaining what I discovered and how to set
this up yourself in an upcoming issue. Until then, consider me gobsmacked!
by Nicholas Vinen
1 Huntingwood Dr, Huntingwood NSW 2148
54 Park St, Sydney NSW 2000
2
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
MAILBAG
your feedback
Letters and emails should contain complete name, address and daytime phone number. Letters to the Editor are submitted on the condition that
Silicon Chip Publications Pty Ltd has the right to edit, reproduce in electronic form, and communicate these letters. This also applies to submissions to “Ask Silicon Chip”, “Circuit Notebook” and “Serviceman’s Log”.
So long and thanks for all the fish
I have enjoyed being the Serviceman since late 2011.
Recently, I’ve had some health challenges that have affected
my ability to deliver columns. While none of them are
overly serious, combined they have made it very hard for
me to write very much. I recently had some corrective surgery, leaving me immobile for the next few months. The
recovery time is notoriously long and painful.
At the same time, my servicing business is naturally winding down (sadly, as clients die off; another long-time client
passed away just last week) and my days as a serviceman are
coming to a close. I’m 64 and retiring officially in July next
year. Even without these health issues, which are resolved
to a point, the writing (mine at least) was on the wall.
I have really enjoyed being a part of the Silicon Chip
family for all this time, and it has been such a dream and
an honour. I hate to say it, but I think it is time to pass the
mantle to someone else now. If I had my way, I’d write for
as long as I possibly could, but I don’t feel I can continue.
All the best with the magazine in the future.
Dave Thompson, Christchurch, New Zealand.
Comment: we’re sad to see Dave go but it’s understandable under the circumstances. It has been a joy to work with
him and read his columns over the last 15 years.
The development of EDM machining in the 1970s
Your July 2020 article on Subtractive Manufacturing
(siliconchip.au/Article/14495) mentioned Electrical Discharge Machining (EDM) starting on page 19. I have some
personal experience in developing that technology.
Cliff Naden was the owner of Eroda Machinery Ltd (not
to be confused with a later company called Eroda Tools),
He drove around in an impressive vintage Bentley. In 1973,
he visited me at home. I was self-employed designing electronic units, repairing and installing sound systems etc. He
asked if I could design a spark erosion machine for him.
Back
Issues
The UK
’s
Circui
t
Pr
Electractical
onics
fT sTa
r TE r
premier
electron
ics so
and
Underst Surgery
using anding and
gyrato
rs
Make
computin
g ma
ker ma
Mrom
Finishin
h Mic
E Th
g
gazine
light con the PicoMite ite E ‘sP La
troller
T’ aT
Audio
softwar smart
e
Designi Out sw iTc
h- on
discreteng a practic
!
al
audio
op am
a Mi
p
GPS-Sy
Analog nchronise
ue Cloc d
k
it wit Ta
6- d Ec
WIN
M
mTo ICRO A
CH
u
Dev ch AR IP
elo
10
adE
pm 00
LLionPico r Es
Kit ent
PrEc
MisE
VaLU
ite smis Ta n c
Es Tolight
E b ox
finECont art
YoUr
TUnErolle
dEsiG
r
ns
WIN!
Microch
ip
Inte
grat
Graphic ed
sE
s and
Touch
M
CuriTE
osity
sT – Pa
luation
bUEva
iLd
Kit an
rT 3
d Us
siL ic
E oU
on ch
r ULTiM
Ec kE
aT E
r
Jump
start
Mini LE
Driver D
Egg Tim
er – eg
to pe
gcellen
rfectio
t brea
n!
kf
ast, tim
Compl
PLUS!
APRIL
13 Cover.in
dd 1
ractica
lelec
practic
ed
etin
g
WideTechno
range the
Tal
inter
Cool Bea k – My tru
face, Ohmmet
th, you
ne
r truth techn
Net Wo ns – Arduin
er
and AI o tal t work, cir
o Boo
rk – Rou
k, pic
ters, pow tcamp: new
n’ mix cuit Surg
www.e
bo
er sup
Sep 202
ery, re
lectron
plies, TEM ards update
3 £5.9
adout,
publish
9
!
U and
ing.co
09
more
m
9 7726
<at>p
32 5730
30
alelec
APRIL 201
3
£4.40
tronics
Practical Electronics is the UK’s premier electronics, computing & maker magazine.
Each month has a wide variety of electronics projects suiting beginners and experts alike. It
also includes many different features on topics like audio, radio, computers and more.
14/02/2
013 10:33:4
7
24 YEAR COLLECTION OF PRACTICAL ELECTRONICS
Every issue of Practical Electronics published from January 2000 to December 2023.
Covering 288 magazines & over 20,000 pages of content; see siliconchip.com.au/Shop/3
PDF Download
SC7650 ▸ $165
The articles can be downloaded per month, year or the
entire block. Some of the excellent columns of Practical
Electronics include Audio Out, Circuit Surgery, Techno Talk
and more. The download size is approximately 5.4GB.
4
Silicon Chip
PDFs on USB
SC7645 ▸ $180
plus postage cost
Supplied on a 32GB Silicon Chip branded USB flash drive.
Purchasing this also gives
you access to the download
version shown opposite.
Australia's electronics magazine
siliconchip.com.au
I was 23 years old, and it was the first time I had heard
of such a machine. I had no idea what he was on about.
Evidently, the few spark erosion machines in general use
at that time in the UK were manufactured by a precision
Swiss company called Agie (there were earlier ones, but
they were not precise, did not have accurate servos and
were not fully solid state controlled), so it was a new field.
I said yes and created what became one of the first general-
use precision spark erosion machines to be designed and
built in the UK in 1973/4.
It started with the massive transformers and power supplies carefully smoothed. I then designed and built the
electronic circuit boards, the metal racks, controls etc at
home in my attic and in the kitchen.
My two prototypes were put together at premises Naden
had hired from an old mill company in Holmfirth. It was
a long distance from home, and I went there nearly every
day of the week from around 7am to 11pm for about a year
until the breakup of my marriage. Being out working every
day was not something my wife could cope with.
Unfortunately, I had a £3000 bank overdraft (which was
a lot of money in 1973!) due to the need to buy the parts in
advance, ready to make six more machines once the prototypes had been proven, as per the contract.
The reason I had to order parts in advance was because,
at that time, there was a 3-9 month wait for heavy-duty
electronic components due to the industrial strife at the
time – the three-day working week. It was a difficult time
as there was no electricity for hours and days at times, constant strikes and unrest.
I could not travel to Holmfirth to build the six machines
immediately because I needed to be at home to look after
my two children. I asked for a delay of 4-6 weeks before
starting to allow me time to find some help at home. Eroda
would not agree to the delay and insisted on immediate
results, then the bank said they would foreclose.
6
Silicon Chip
A week later, Mr Naden ‘kindly’ offered to give
me £3000 for the parts at cost, providing I gave
him all of my diagrams and plans. They therefore got my designs for nothing, and I received
no additional compensation for all my work for
that year.
My first prototype was taken by Eroda Ltd to
Burnley Engineering, which was owned by RollsRoyce. It was used to make titanium fan blades
out of solid titanium slabs of metal for use in the
newer version of their RB211 jet turbine engine.
Rolls-Royce had run into severe financial difficulties in 1971 and had to be taken over by the
UK government. The RB211 needed to work, but
the intended Hyfil carbon-fibre composite fan
blades failed bird-strike testing, so they needed
to switch to titanium.
The problem with titanium is that it is
extremely strong and very difficult or virtually
impossible to machine. Spark erosion overcame
these problems, and their RB211 engine with
these new titanium blades cut by spark erosion
helped to save the company in 1973/74.
The RB211’s success was crucial to the company’s recovery. The engine became the foundation
for the highly successful family of Rolls-Royce
three-shaft engines that followed.
My second prototype machine was used by Ford to erode
microscopically thin holes in the suspension ball joints (in
their Capri cars) to allow hot gases to escape but not grease,
which had to remain inside. Some ball joints had failed
while driving due to the build-up of trapped air pressure,
which could fracture the joint.
Eroda Machinery then started the manufacture of my
machines and ended up employing 250 people at Silkstone near Barnsley. I received no further payment for my
designs or development work.
About 10 years ago, my original chewed-up pencil drawings and schematics were found by a Sheffield company; I
have kept them as a proud yet sad memory of engineering
from the 1970s! I have enclosed a scan of the first brochure.
The happy ending is that I met a lovely lady, Sue, in
the late 1970s and we have now been happily married for
50 years.
Dr Stephen H. Alsop, Sheffield, England.
Repair services wanted
I’ve been a long-time subscriber to your wonderful magazine, even before you offered a subscription service. I
eagerly checked the newspaper stands for the next one. I still
look forward to each month in anticipation, albeit in digital
form now, as my lovely wife has forced me to downsize.
Anyway, I wanted to ask if your office or any of the
contributors to the Serviceman’s Log or other articles to
your magazine are able to repair a Synology NAS model
DS2411+. I have many PDF articles and various personal
scanned documents I want to retrieve from the unit. This
is the consequence of de-cluttering. If you can recommend
or suggest somebody, it would be greatly appreciated.
I would also like to find someone to repair a Nakamichi SR-3 receiver. The speaker hospital (they also do stereo repairs) did an initial repair, which worked, but when
I brought the unit home, I found that the manual station
Australia's electronics magazine
siliconchip.com.au
Blackmagic Studio Camera
The ultimate live production
camera in a revolutionary all-in-one design!
Introducing the world’s most advanced self contained studio
camera! Blackmagic Studio Cameras have the same features
as large studio cameras, miniaturized into a single compact
and portable design. Advanced features include talkback, tally,
camera control, built in color corrector, Blackmagic RAW
recording to USB disks and much more! You can even add a focus
and zoom demand for lens control!
Powerful Broadcast Connections
Revolutionary Studio Camera Design
USB Expansion Port for Accessories
The distinctive Blackmagic Studio Camera has the benefits of a large
studio camera because it’s a combination of camera and viewfinder all
in a single compact design. The camera is designed for live production
so it’s easy to track and frame shots with its large 7” viewfinder. The
touchscreen has menus for camera settings, and knobs for brightness,
contrast and focus peaking.
The Blackmagic Studio Camera features a high speed USB-C expansion
port that allows you to record to external disks or connect accessories.
Plug in an external USB flash disk and the camera can record high quality
12-bit Blackmagic RAW files for later editing and color correction. Or
plug in the optional focus and zoom demands so you can adjust the
lens from the tripod handles just like a large studio camera!
Get Cinematic Images in Live Production
The amazing 4K sensor combined with Blackmagic generation 5 color
science gives you the same imaging technology used in digital film
cameras. Plus, when combined with the built in color corrector you
get much better images than simple broadcast cameras. The color
corrector can even be controlled from the switcher. The resolution of
3840 x 2160 allows both HD and Ultra HD work.
www.blackmagicdesign.com/au
Blackmagic Studio Cameras have lots of connections for connecting
to both consumer and broadcast equipment. All models feature
HDMI with tally, camera control and record trigger, so are perfect for
ATEM Mini switchers! The advanced Blackmagic Studio Camera 4K
Pro model is designed for broadcast workflows so has 12G-SDI,
10GBASE-T Ethernet, talkback and balanced XLR audio inputs.
Blackmagic Studio
Camera 4K Plus G2 ....... $2,115
Blackmagic Studio
Camera 4K Pro G2.....$2,945
Learn More!
selection worked perfectly but the auto-scan feature just
cycled through the entire FM band; it never locked onto
a station.
I took it back to them to fix, and it ended up sitting with
them for two years. The other day, I asked for its return.
I kindly asked for its status on numerous occasions, and
they went through their list of canned replies. I suspect
they probably took an initial look, decided it was too hard,
and were probably going to work on it during their quiet
period, but eventually did nothing.
Can you recommend somebody, possibly one of your
contributors, who might be up to the challenge? It’s a beautiful piece of equipment that I would love to get back into
full operation.
Nick Lutovski, Oatley, NSW.
Technical eBook/book publisher wanted
Do any of you or your readers know of any book publishers or publisher’s agents who produce eBooks of technical information? I have a manuscript for an educational
book relating to electronics and information technology.
I have aimed it at the eBook market by making each subject with a maximum of 180 words per page and diagrams
that are designed to fit a tablet or smartphone screen. As a
printed book, it could be fit into an outline that would fit
neatly into the side pocket of a backpack, toolbox or briefcase.
The content is over 150 separate subjects, each 1-2 pages,
with chains of subjects for more in-depth topics. For example, the one-page topic, “A one-bit digital adder” is on the
end of a string of 25 other topics, which starts at “Electrical units”.
Early subjects are not beyond the understanding of an
interested 10-year-old. Later descriptions of doped silicon
would require at least some understanding of the periodic table, and later again there are subjects dealing with
machine language, the primary data register, the significance of op codes and how they are handled by the internal architecture of the CPU.
I hope the book will be a teach-yourself companion for
the long term as the student grows in understanding. The
topic is obviously specialised, but it has worldwide significance. In Australia, every high school has up to half a
dozen or so students who are thirsting for this information and not getting it from the present education system.
I am willing to negotiate internationally if the best agent
is somewhere overseas.
Jim Sinclair, Noarlunga Downs, SA.
Comment: we can’t think of an obvious publisher to direct
you to, but keep in mind that these days self-publishing
is quite possible, through companies like Amazon (Kindle
Direct Publishing). eBooks can also be distributed through
Google Play Books, Apple Books, Kobo Writing Life and
Barnes & Noble Press.
On Improvised Electronics article and SEMs
The two-part article “Improvised Electronics” in the
September & October issues is a fantastic look at electronics over the years (siliconchip.au/Series/462). It brings
back memories of my experience going back to when I first
built my first crystal set radio in the 1960s. The part of the
discussion on vacuum pumps and, at the very end, about
someone building an electron microscope, brought back
memories of working in R&D until 1998.
8
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
I serviced and maintained a Jeol JSM-6400 scanning
electron microscope (SEM). There were two units I was
aware of, one at my place of work and the other at Latrobe
University in Melbourne. This instrument operated under
high vacuum, as did the Dynavac vacuum coater used for
sample preparation.
The actual vacuum systems comprised a two-stage oil
vacuum pump as the backing pump down to about 10-4
torr, followed by a diffusion pump, which could bring the
vacuum down to 10-10 torr. Where the source filament was,
it apparently had a further sputter-ion pump.
On more modern machines, the diffusion pump was
replaced by molecular turbine vacuum pumps to ensure
no oil vapour in the vacuum. They have the same vacuum
end-point limit and both require a backing pump, as they
are damaged if they start at atmospheric pressure.
As well as being used to coat the samples in carbon,
aluminium or gold for use in the SEM, it was also used
to put the aluminium reflective coating on an 8-inch
(~20cm) parabolic piece of glass to make a reflecting
telescope mirror.
Regarding the person that built his own electron microscope, I can only say it would have its challenges. I know
that aligning the optics of the Jeol SEM was a great challenge for me; I felt good when it was done.
For those interested, Jeol Ltd has a PDF of “Scanning
Electron Microscope A to Z”, which explains a lot about
the instrument (siliconchip.au/link/acdg).
Wolf-Dieter Kuenne, Bayswater, Vic.
Making cathode ray tubes (CRTs) was mentioned in the
articles about Improvised Electronics (September-October
2026; siliconchip.au/Series/462). The following is a fascinating documentary about the history of Tektronix. Early
in their history, they decided to make their own CRTs.
The video is titled “The Fascinating Story of Tektronix,
the Oregon Oscilloscope Giant That Gave Silicon Valley Its
Eyes”. See it at: https://youtu.be/AHRzqulqKa4
Paul Howson, Warwick, Qld.
Thirdly, the efficiency of these linear chargers is not
that great, but if a USB lead containing the normal size
power wires is used, at 1A, more energy is lost as heat
than is stored in the cell. To correct this, I make cables
using recovered USB plugs fitted with only power wires
of either 7 × 0.2mm or 16 × 0.2mm size depending on the
length of the cable.
On another topic, in one of my earlier letters, I stated
that home solar-generated energy could not be transmitted
upstream into the higher voltage networks. Another reader
corrected me, and I agree that it can, but only under certain
conditions. Since then, I have thought about how energy
could be distributed under all conditions both from phase
to phase and to adjacent low-voltage networks.
Three-phase motors have the natural ability to transfer
energy from one phase to another. Therefore, an appropriately sized motor could be used to ensure the maximum
use of solar energy within a single low-voltage network.
City centres and network areas, which are dominated by
high-rise units, need more energy than can be supplied by
any local solar installations, and almost certainly, they will
be surrounded by low-voltage networks, which contain
mainly houses and plenty of solar installations.
If, at the boundaries of the networks, a motor in one network is coupled (in phase) to a motor in the adjacent network, then power should transfer from the suburban network to the city or high-rise network.
I do not know if this system has been considered by academics and engineers, and whether it has been rejected.
However, I wanted to mention it because, in principle, it
is a very simple system using readily available motors and
could be easily tested without excessive cost.
If energy transfer by this method was proven to be economic, the next step would be to combine the two motors
into a single unit in the same manner as rotary converters that were used to generate high-voltage DC from
low-voltage DC. This would eliminate two bearings, the
coupling, and a lot of assembly work, resulting in more
efficient assembly.
George Ramsay, Holland Park, Qld.
Li-ion charger modules and mains power distribution
Educational electronics game available on Steam
History of Tektronix documentary
Some time ago, I bought a quantity of the little Li-ion
charger modules and I have used several them.
For those with a micro-USB socket, I have discovered
that the socket is of a very weak design. If more than a gentle push is used to insert the plug, the ‘guts’ of the socket
are pushed out. The mini-USB socket has a folded tab to
retain the moulded centre, but the micro-USB socket has
only two small side tabs, and the mating plastic shears off
when a little force is used.
The best way to use the micro-USB modules is to solder
5V power wires to the PCB.
Secondly, I should have known better, but I never gave
a thought to the heat generated by the charger modules.
It was only while studying the datasheet for the IC that I
realised that up to 2W needs to be dissipated when the cell
voltage is 3V and just above. Without some heatsinking,
the module gets quite hot.
The IC has overheating protection built in, but it would
be much better to provide means to dissipate that heat. If
the IC does overheat, it may either shutdown or charge at a
lower average current, so the charging time will be longer.
10
Silicon Chip
I saw your Pinball Machine series (June-October 2026;
siliconchip.au/Series/460) and thought I would share that
I made a puzzle game called BitWise. You wire up circuits on a board and a simulator runs them. It starts with
a battery and a lamp, and later you build a small computer. There’s no answer key, so a weird build passes if
it does the job.
It’s out now on Steam for Windows and Mac. See: https://
store.steampowered.com/app/5195360/BitWise/
Faint Current, via email.
Comment: we tried the game, and it seemed quite well-
designed and educational.
Analog community TV repeater story
In 1984, I donated a colour TV to a Baptist fundraising
auction. A couple of weeks later, the purchaser asked if I
could get it going for him. He said it’s at Whanarua Bay. I
asked: where is that?
He said it was up the east coast of New Zealand, 300km
from Taupo. I could take the family and stay in his house
there, so off we went. I found the reception was just about
Australia's electronics magazine
siliconchip.com.au
ATEM Television Studio
Portable all-in-one live production switcher with
streaming and network storage for DaVinci editing!
ATEM Television Studio is a professional live production switcher
built into a broadcast control panel so it can be used for high
end work while being extremely portable. You get a powerful
switcher with 8 standards converted SDI inputs, aux outputs,
4 chroma keyers, 2 downstream keyers, SuperSource, 2 media
players and lots of transitions! There are even ISO models that
record all 8 inputs for editing!
Connect up to 8 Broadcast Cameras
Easy to Use and Fast to Learn
Built In Fairlight Audio Mixer
There’s never been a switcher that’s easier to use, as you simply press
any of the program row buttons on the front panel to cut between video
sources. You can select from exciting transitions such as dissolve, or
more dramatic effects such as dip to color, DVE squeeze and DVE push.
You can even add a DVE for picture in picture effects with customized
graphics. Then you can live stream the results!
With a built in Fairlight audio mixer, ATEM Television Studio makes it
possible to do complex live sound mixing. The internal mixer features
enough channels for all SDI inputs, as well as extra channels for the
XLR, RCA and MADI inputs! Each input channel features the highest
quality 6 band parametric EQ, compressor, limiter, expander and
noise gate as well as level and pan controls.
Self Contained Broadcast Quality Switcher
The ATEM Television Studio combines a switcher and control panel
into the same unit, so it’s extremely portable. The front panel includes
buttons for selecting sources, triggering transitions and setting up
video effects. You also get a built in t-bar for manual transition control!
The front panel even has an innovative audio mixer control area with
live metering on a dedicated LCD.
With 8 SDI video inputs, ATEM Television Studio lets you connect
multiple cameras for different views of the performance. Cameras
with SDI outputs, such as Blackmagic Studio Camera or Blackmagic
URSA Broadcast G2, offer better low light performance and use
professional lenses. All video sources will re-sync to the switcher, so
setup on location is easy!
ATEM Television Studio HD8 ...................$4,209
ATEM Television Studio HD8 ISO........ $4,675
ATEM Television Studio 4K8.....................$7,765
Learn more at www.blackmagicdesign.com/au
Learn More!
nonexistent from a Band 1 transmitter on Mt Te Aroha,
185km away, across the sea. When the tide was out, you
got sound and a vague, snowy picture. With the tide in, it
all disappeared.
There was a band 3 transmitter on Mt Edgecumbe, 103km
away, obstructed by hills from the bay. I flippantly said if
we put a repeater on the high hill behind the bay, we might
get a reasonable signal. My eager friend offered to test for a
signal. I gave him a little one-inch TV I had. A few weeks
later, he said he had gotten a picture on it.
So during the next holidays, I did the 30-minute climb
up the hill with an aerial and field strength meter and got
an 80μV signal. I figured that the noise floor would be so
low that it would be usable with a stack of two Hills PA16
phased-array aerials. NZ on Air, a government body, provided $100 per resident, but a signature was required from
every resident.
My willing friend went and got them, so we had $4200;
not quite enough, but with donations, we had enough.
Then came the day of the installation. The whole community came to help. A resident with a helicopter carried
the deep-cycle battery, poles, solar panels and building
materials. Later, we were told there were “weed” patches
on the hill and every time we went up, binoculars were
trained on us.
Because we were repeating and not translating frequencies, the transmitting and receiving aerials had to be
separated by 100m, so a ditch was dug and a cable buried between the aerial sites. The 1W transmitter gave a
three-channel service. A resident 4km away had a good
picture. Subsequently, selective fading occurred due to
the signal path being mostly along a seashore with hot and
cold air layers.
I installed three Hills single-channel amplifiers, which
had AGC. That improved reception a lot. Freeview digital
satellite service has now replaced this analog technology.
Ian Foster, Taupo, New Zealand.
Litz wire can improve switch-mode efficiency
I have the following to add regarding the letter on skin
effect in the July 2026 issue (“The skin effect can be a major
efficiency penalty”, Mailbag, p4 & p6).
It is not uncommon to use Litz wire in SMPS design up
to a few megahertz; its performance above this starts deteriorating. It is commercially available, although for very
small runs you might roll your own. Self-fluxing insulated
copper wire is best used for easiest soldering.
Personally, I have used it in SMPS design up to a few
MHz and in a system to generate a 1.2MHz 1500V regulated
sinewave to drive a laser electro-optic modulator.
One of the dangers of using simulators is that they are
only as good as the models used for the components; even
the best wide-band models are a compromise. Increased
computational time will be the penalty for complex component models, and SPICE vendors may pride themselves
on simulation speed.
In that respect, magnetic core models can be particularly
troublesome. Simulators can help to point you in the right
direction if you know where the limitations lie, but as the
author found out, in the end you have to build, test and
iterate the design.
SC
Richard Gabric, Oxford, New Zealand.
Versatile
Battery
Checker
This tool lets you check the condition of most
common batteries, such as Li-ion, LiPo, SLA, 9V batteries, AA, AAA,
C & D cells; the list goes on. It’s simple to use – just connect the battery to the
terminals and its details will be displayed on the OLED readout.
Versatile Battery Checker Complete Kit (SC7465, $65+post)
Includes all parts and the case required to build the Versatile Battery Checker, except the optional
programming header, batteries and glue
See the article in the May 2025 issue for more details: siliconchip.au/Article/18121
12
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
MACHINE TOOLING
5 Inserts Per Pack
Grade KC5410 - Aluminium
Made in Germany for exceptional
quality and precision
5 types
PRICED FROM
$98.00
Round
M35 High speed steel with 5% cobalt
TiCN Coating for increased wear resistance
Precision ground thread
DIN 371 Standard
Spiral Flute Taps
M35 High Speed Steel with 5% cobalt
TiCN Coating for increased wear resistance
Precision ground thread
DIN 371 Standard
BSPT Taps
M35 High Speed Steel with 5% cobalt
TiCN Coating for increased wear resistance
Precision ground thread
Sold individually
ER20
ER25
ER16
M35 High speed steel with 5% cobalt
TiCN Coating for increased wear resistance
Precision ground thread
ISO 529 Standard
FOR MORE
PRODUCTS VISIT
MACHINERYHOUSE.COM.AU/
SIC2611
Trigon / Diamond
ER32
ER40
6 sizes: 1.0 ~ 1.5mm pitch
PRICED FROM
$17.60
PRICED FROM
M6
M12
M8
M14
M10
M16
M6
M12
M8
M14
M10
M16
6 sizes: 1.0 ~ 1.5mm pitch
$17.60
4 sizes
PRICED FROM
$31.90
1/8" x 28
Hand Tap Sets - 3 Piece
Triangle
$27.50
High Precision Collet made from
hardened spring steel
Manufactured to DIN standard
Made in Taiwan
Individual Gun Taps
Diamond (Rhombic)
5 sizes
PRICED FROM
Individual Collets
Rhombic
PRICED FROM
$27.50
1/4" x 19
1/8" x 28
1/2" x 14
10 sizes: 0.5 ~ 2.0mm pitch
M3 ~ M20
SYDNEY
NEW STORE
BRISBANE
MELBOURNE
PERTH
ADELAIDE
(02) 9890 9111
(07) 3715 2200
(03) 9212 4422
(08) 9373 9999
(08) 9373 9969
1/2 Windsor Rd,
Northmead
680 Boundary Rd,
Richlands
4 Abbotts Rd,
Dandenong
11 Valentine St,
Kewdale
11/20 Cheltenham Pde,
Woodville
Specifications & prices are subject to change without notification
10_SIC_231126
Turning Inserts - Carbide
HIGH QUALITY TOOLING
FOR GREAT PRECISION
Part 2 by Dr David Maddison, VK3DSM
Image source: www.pexels.com/photo/vintage-computermotherboard-with-cpu-and-ram-slots-37256941
THE HISTORY OF
AMD
trials and tribulations
We started this series last month by describing how AMD was
started by a few disgruntled Fairchild employees (like Intel...)
initially operating from a garage. It grew quickly, to the point of
building clones of most of Intel’s processors. We continue from
1993, with the Pentium on the horizon.
T
he period covered in the first article in this
series was roughly 1969 to 1995. During
that time, its main products were
copies of other company’s designs,
usually under license, sometimes via
reverse engineering. However, AMD
did design its own chips, some of them
quite complex.
One such example was the Am2900
chipset from the mid-1970s, a powerful (for the time) processor solution
that was used in minicomputers and
mainframes. From that, its ambition
only grew, and AMD would soon start
producing RISC processors that were
14
Silicon Chip
well ahead of their time, such as the
Am29000.
The Am29000
By 1993, AMD had produced the
Am386 (1991) and Am486 (1993). To
understand the next part of the AMD
journey, we need to take a step back for
a moment. By 1975, the microprocessor market was split into two philosophies: cheap and slower single-chip
CPUs like the Am9080/Intel 8080, or
expensive but faster/more powerful
multi-chip processors like the Am2900
family.
Australia's electronics magazine
The multi-chip Am2900 family used
bit-slice technology to build powerful processors for markets where
cost was no object, such as high-end
minicomputers, military hardware
and graphics applications. Chips like
the Am9080 were used for hobby and
consumer PCs.
Systems built using AMD’s Am2900
chips ran extremely fast (for the time)
at around 20-40MHz, compared to x86
processors like the 8080 (1974) and
8085 (1976), which ran at 2-3MHz.
Because they were so fast, they
could even run emulators to mimic
the 8085, which could run up to ten
times faster than the actual 8085 chip.
As processor technology advanced,
AMD moved away from the multichip bit-slice concept to produce the
single-chip Am29000, also known as
the 29k (see Figs.13 & 14). Like the
Am2900, it was a RISC processor.
When it was introduced in 1987/1988,
it was one of the fastest 32-bit processors in the world, running 3-5 times
faster than CISC chips of the time.
For example, the first version of the
Am29000 could achieve 17-23MIPS
(million instructions per second),
while CISC processors like the Intel
80386 or Motorola 68020 would run
at only 3-5MIPS.
Although the Am29000 was not an
x86 processor, for reasons that will
be explained later, this chip allowed
AMD to compete with the new Intel
Pentium (released in 1993) which Intel
refused to license to AMD (along with
the 80386 and 80486; AMD reverse-
engineered them as mentioned previously).
In the 1990s, x86 processors were
dominant in the PC market even
though the 29k was significantly faster.
The 29k beat Intel in the embedded
processor market for such devices as
HP Color LaserJet printers and graphics terminals, but as the 29k was not
an x86 processor, how could it be used
in the PC market if it didn’t ‘speak’ the
x86 language?
Despite being faster, the 29k was losing money. Therefore, in 1995, AMD
officially cancelled the entire 29k family of processors and moved the engineering team to its PC division. Since
AMD already had an advanced, superfast RISC processor in the form of the
29k, it would have been foolish to
waste the design effort, so they recycled elements of the design into what
they called the K5.
siliconchip.com.au
This was to be AMD’s next big processor, designed to compete against
Intel’s Pentium (1993), specifically the
P5 and P54C versions. “K” in K5 stands
for the fictional element Kryptonite,
which is Superman’s only weakness
(by analogy, Superman was Intel).
The “5” is a reference to the Pentium,
which was so-named because it was
next in the series after the 286, 386
and 486.
Developing the K5
The architecture of the Am29000
(29k) released to customers in 1988
was completely unrelated to the Intel
x86 architecture. The 32-bit Am29000
was a RISC processor. However, AMD
needed something to compete with
the Intel Pentium, which was released
in 1993.
To build a competitive processor,
they took elements of an unreleased
version of the Am29000 codenamed
“Jaguar”, such as the mathematical
engines and out-of-order execution
logic, and attached an x86 decoding front-end, which translated x86
instructions into 29k-style RISC
instructions on-the-fly (which they
called RISC-ops).
For the back-end, they used the
physical execution units, arithmetic
logic units (ALUs) and maths coprocessors from the 29k. The K5 contained
five ALUs from the Jaguar microarchitecture.
To handle speculative out-of-order
execution, where the processor can
execute instructions ahead of time
based on what it predicts the software
will do next, it relied on the Reorder
Buffer (ROB) from the Jaguar, with
minor modifications to keep track
of x86 registers rather than 29k registers. The floating-point unit (FPU)
was taken from the Am29050 to be
used in the K5.
The K5 also adopted the concept of
the branch target cache (BTC) from the
Am29000, which minimised processor
pipeline stalls by storing the targets of
predicted program branches (eg, conditional jump instructions).
The BTC is a tiny, high-speed memory buffer that instantly feeds the first
few instructions of a detour in code
(such as an if/else statement that forces
the computer to break from a linear
sequence to a different set of instructions), giving the slower main memory enough time to catch up without
stalling the processor.
siliconchip.com.au
RISC vs CISC
RISC stands for Reduced Instruction Set Computer, while CISC stands for
Complex Instruction Set Computer. In the 1970s and 1980s, these represented
two different approaches to designing processors, although the distinction has
become much less clear over time.
The idea behind RISC processors was to use a relatively small set of simple,
regular instructions. These were often encoded using fixed-length instruction
words, making them easier to decode and pipeline efficiently.
In contrast, CISC processors used larger and more complex instruction sets,
often with variable-length instructions and many addressing modes. Some
instructions could do quite a lot of work, but might take multiple clock cycles
to complete. This was considered programmer-friendly, especially when more
software was written directly in assembly language.
Arm and RISC-V are modern examples of RISC-derived instruction set
architectures. The main surviving CISC architecture in general-purpose
computers is x86, as used by Intel and AMD processors.
However, modern Intel and AMD x86 processors do not execute most x86
instructions in a simple, literal way. Instead, they decode them into internal
micro-operations, which are then scheduled and executed by a highly optimised
processor core.
These internal operations are often described as RISC-like, although they
are private implementation details rather than a separate RISC instruction set.
The AMD K5 was an early example of this approach.
Over time, the distinction between RISC and CISC has blurred. CISC
processors such as x86 adopted many RISC-like implementation techniques,
including pipelining, register renaming, out-of-order execution and internal
micro-operations.
Meanwhile, RISC architectures became more complex, adding features
such as compressed instruction encodings, SIMD/vector extensions, memoryordering features, privilege modes and specialised instructions. As a result, RISC
and CISC now mostly describe the software-visible instruction set architecture,
not a simple division between how processors work internally.
Figs.13 & 14:
the AM29000
chip and
die. Sources:
https://w.wiki/
RQYm
https://w.wiki/
RQYn
Australia's electronics magazine
December 2026 15
Fig.15: the AMD K5 block
diagram. Note the four
instruction decoders to
convert x86 instructions to
RISC-like instructions.
The K5 was designed to be pin-
compatible with the Pentium; a user
could simply unplug their Pentium
from a Socket 5 or Socket 7 motherboard and drop in a K5.
However, the K5 turned out to be a
disappointment. It suffered from serious design flaws: lower-than-expected
clock speeds and poor performance
compared to Intel’s Pentium and Pentium Pro.
The K5’s fatal flaw
Original source:
https://w.wiki/RQYe
In essence, the K5 was less of a Pentium clone and more of an Am29000
masquerading as a Pentium.
Independent Architecture and the
GHz race (1996-2003)
After the legal battles of the 386/486
era, AMD made a strategic decision
to stop relying on reverse engineering and commit to designing its own
next-generation x86 processors from
the ground up. This led to the K5,
K6, and K7 (Athlon) processor families that would ultimately transform
the company.
The 32-bit K5 (Figs.15-17) was
released in 1996. It was AMD’s first
independently designed x86 microprocessor, intended to compete with
Intel’s Pentium.
AMD’s core RISC architecture was
highly efficient, but it struggled to meet
the clock speeds of Intel chips, which
the marketplace viewed as a competitive feature. Because of the high efficiency of the RISC architecture compared to CISC, it is not necessary to
have the same high clock speeds for
equivalent performance.
However, to overcome this marketing deficiency, AMD introduced
the concept of PR or Processor Rating numbers. For example, an AMD
K5-PR100 ran at a physical clock speed
of only 66MHz but AMD claimed it
was equivalent to an Intel Pentium
running at 100MHz, hence 100 in the
name of the K5. AMD also sold the
PR100 at much higher prices than the
100MHz Pentium.
The K5 had superior raw integer processing performance to the Pentium,
which made it better for running typical office software, but it had an inferior
FPU, which is necessary for complex
3D mathematical operations as used in
games. It happened that, just after the
K5 was launched, 3D games like Quake
were released, which ran much better
on the Pentium’s superior FPU.
Thus, the Pentium retained the loyalty of gamers and power users, for the
time being at least.
The chip taught AMD painful but
valuable lessons about the extreme
difficulty of designing competitive
x86 processors. Nevertheless, as
AMD’s first fully in-house designed
x86 CPU, it was an important milestone for them.
Introducing the K6
AMD recovered quickly with the
32-bit K6 (1997), which was based on
technology acquired from NexGen –
see Figs.18 & 19.
AMD made a strategic acquisition
by purchasing NexGen in 1996. It was
a small but innovative semiconductor company founded in 1986. NexGen had developed one of the most
advanced x86 processor designs of
the era with its Nx586 and Nx686
architectures, which used a RISC-like
internal core that translated complex
Fig.16: the AMD K5
physical package.
Image sources.
Source: https://w.
wiki/RQYf
Fig.17: the AMD
K5 die. It’s
significantly more
complex than
its predecessors.
Source: https://w.
wiki/RQYd
16
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
x86 instructions into simpler micro-
operations, much like the K5.
NexGen extensively patented this
concept, releasing its Nx586 in 1994,
before the K5. Struggling with the disappointing performance of its own
K5 processor, AMD acquired NexGen
for approximately US $850 million
in stock.
The acquisition proved to be a turning point. AMD scrapped much of
the troubled K5 project and adopted
NexGen’s superior microarchitecture,
using the Nx686 as the foundation for
the K6 processor, adding a floating-
point unit and adapting the chip to
fit standard motherboards. It then
launched the K6 in 1997.
Many of NexGen’s top engineers also
joined AMD, bringing critical expertise that would later contribute to the
successful Athlon (K7) and subsequent
generations. The NexGen acquisition
gave AMD both advanced technology
and vital engineering talent at a crucial moment in its history.
The K6 was a much stronger performer than the K5, finally giving AMD
a credible competitor to Intel’s Pentium II (1997). It introduced features
such as MMX (multimedia extensions)
support and a larger internal memory cache.
While still behind Intel in raw
single-threaded performance and manufacturing process technology, the K6
allowed AMD to regain market share
through aggressive pricing and excellent floating-point performance.
The K6 series proved that AMD
could produce competitive processors, but it also highlighted that the
company was still playing catch-up
to an extent.
The K7
A breakthrough came with the 32-bit
K7, codenamed Athlon, launched in
Complicated Athlon naming schemes
AMD’s K7 chips were sold over 1999-2003, while the K8 chips were sold during
2003-2007. The 32-bit K7 was named AMD Athlon, with later models being
Athlon XP. They competed against the Pentium III and early Pentium 4.
For the K8, AMD decided to keep the trusted Athlon name, which had gained
great respect and market success, while adding 64 to indicate this was the
world’s first mainstream 64-bit processor. It was named Athlon 64 and later
Athlon 64 X2 (dual-core).
K7 and K8 were internal names, not official names, while K5 and K6 were
official names; the “K” series was discontinued after the K8. Anther name used
for the K7 is the Duron, which was the budget version, launched in 2000. It had
a smaller L2 cache to reduce its cost.
The K8 used the names Opteron for its enterprise/server model, which
competed against the Intel Xeon; Athlon for mainstream models, which
competed against the Pentium; and
Duron or Sempron for the budget
versions, which competed against the
Intel Celeron. The Turion 64 was used
for laptops and competed against the
Pentium M. See Table 4 overleaf.
Fig.20: an AMD Athlon (Thunderbird)
PGA socket CPU. For this generation,
they abandoned the heat spreader for
better thermal contact between the
die and the heatsink. It worked, but
unfortunately it was possible to crack
the die if too much or uneven clamping
force was applied. Source: https://w.
wiki/RQZG
1999 – see Figs.20 & 21. The Athlon
represented a complete architectural
overhaul and was AMD’s most ambitious project to date.
Its key innovations included a completely new superscalar, out-of-order
execution microarchitecture with nine
execution units; a revolutionary EV6
bus (originally developed by Digital
Equipment Corporation [DEC]) instead
of Intel’s GTL+ bus, allowing much
higher bandwidth and easier multiprocessing; and excellent floating-point
performance.
The original “Classic” Athlon processor was in the Slot A format, like
Intel’s Pentium II (see Fig.22).
In March 2000, AMD achieved a
historic milestone when it released
the Athlon 1GHz. It was the first x86
processor in the world to reach 1GHz
(1000MHz or one billion instruction
clock pulses per second). From August
1999 until January 2002, it was the
fastest x86 chip in the world. This was
a major psychological and marketing
victory over Intel.
The Athlon not only matched but
often exceeded Intel’s Pentium III
(1999) performance, particularly in
gaming and content creation. For the
first time in its history, AMD was
widely regarded as having the faster
desktop processor. The success of the
Figs.18 & 19: an
AMD K6-233APR
processor and die.
The metal can
on top is a heat
spreader that also
physically protects
the silicon die
underneath. Most
modern CPUs use a
similar approach.
Sources: https://w.
wiki/RQZC
https://w.wiki/
RQZD
siliconchip.com.au
Australia's electronics magazine
December 2026 17
Athlon showed that AMD was no longer just a cheaper alternative; it led
the market in performance, at least
for a while.
This coup was achieved through a
combination of talented individuals,
hard work, careful planning and calculated risks. The relative success of
the K6, thanks to the acquisition of
NexGen, generated the breathing room
and cash flow to attempt something
more ambitious.
AMD had its own technical experts,
the people they brought on from NexGen, and also benefited from talent
connected with DEC/Alpha, including
Dirk Meyer and Jim Keller. The K7 they
developed had strong out-of-order execution, a much better floating-point
unit than AMD’s earlier chips and a
bus architecture influenced by DEC’s
Alpha EV6 bus.
Intel was also heading toward the
Pentium 4/NetBurst era, where it
would chase clock speed very aggressively, with limited success. During
1999-2003, Intel was not making flawless architectural choices, and AMD
exploited that gap.
AMD used partnerships to compensate for being smaller. Jerry Sanders’
“virtual gorilla” strategy said that since
AMD could not match Intel’s internal
resources, it could partner its way into
competitiveness.
For example, they partnered with
Motorola to develop a copper interconnect process. Athlon was among
Fig.21: the AMD K7 “Athlon” architecture, again a step up in complexity but
bringing significant performance benefits. Source: https://w.wiki/RQZE
Fig.22: the original Athlon processor, the Athlon Classic in Slot A configuration.
The chips on both sides are L2 cache memories. Like Intel, AMD later abandoned
the slot form factor due to frequency limitations. Source: https://w.wiki/RQYg
18
Silicon Chip
Australia's electronics magazine
the first commercial processors to
use copper interconnect technology,
helping AMD push frequency and
efficiency.
Wired reported in August 1999 that
AMD claimed the Athlon 600 was
about 9% faster on average than a
600MHz Pentium III and 21% faster
in graphics tests. Custom PC’s retrospective on the Athlon 600 backs up
those figures.
Innovation leadership (2003-2007)
Encouraged by the Athlon’s success, AMD went further into the highend market. In 2003, the company
launched the Opteron processor, the
first implementation of K8 architecture
(see Figs.23 & 24).
The Opteron introduced AMD64
(later known as x86-64), a 64-bit extension to the x86 instruction set that
AMD developed.
Basically, AMD decided to unilaterally extend the 32-bit x86 instruction
set to 64 bits, allowing vastly larger
amounts of memory to be addressed
linearly (the 32-bit limit was 4GiB)
while maintaining full backward compatibility with 32-bit software. This
proved to be highly successful, as Microsoft, the Linux kernel developers
and other operating system vendors
quickly added support.
Most importantly, Intel itself eventually adopted AMD64, initially
calling it EM64T and later Intel 64,
rather than continuing with its own
incompatible 64-bit architecture (Itanium/IA-64). By 2006, Intel had fully
embraced AMD64, making it the universal standard for 64-bit x86 computing.
AMD64’s success gave AMD a temporary technological lead, and it also
ensured the long-term survival of the
x86 architecture into the 64-bit era.
Unlike Intel’s earlier failed Itanium
architecture, AMD64 offered full
backward compatibility with 32-bit
software while adding true 64-bit
addressing and performance. The
Opteron quickly gained traction in the
server market, particularly for scientific computing, databases and high-
performance clusters. It gave AMD its
first serious inroads into the enterprise
server segment, a market Intel had long
dominated.
While the 1GHz K7 Athlon relied
on a traditional Front Side Bus (FSB)
to access RAM via a separate memory
controller chip on the motherboard,
siliconchip.com.au
Table 4: a comparison of some AMD K7/K8 variants and competing parts from Intel
AMD processor
Year
Target market
Intel competitor
Outcome
Athlon (K7)
1999
High
performance
Pentium III Slot 1
Athlon Slot A (cartridge), 500-1000MHz.
Faster than the Pentium III.
Athlon (K7)
2000
High
performance
Pentium III
Socket 370
Athlon Slot A, equivalent in performance
to the Intel model. AMD crossed the 1GHz
barrier one day before Intel.
Athlon XP (K7)
2001
Budget
desktops
Early Pentium 4
Popular socketed version.
Athlon MP (K7)
2001
High-end
workstations
and servers
Early Xeon
processors
based on the
Pentium III and
early Pentium 4
Designed to support dual-CPU
motherboards for workstations.
Duron (K7)
2000
Budget
desktops
Celeron
Duron was superior to Celeron.
Athlon 64 single-core (K8)
2003
High
performance
Single-core
Pentium 4
AMD had the superior processor.
Athlon 64 FX Ultrapremium single-core (K8)
2003
Gamers,
overclockers.
Pentium 4
Extreme Edition
The FX was the pinnacle of single-core
gaming performance at the time.
Opteron (K8)
2003
Servers/data
centres
Pentium 4 Xeons
AMD had superior architecture and
performance in this era.
Athlon 64 FX dual-core
and Quad-FX models (K8)
20032006
High
performance
Core 2 Extreme
6800, Core 2
Quad
The Quad FX needed a special motherboard
to run two dual-core FX chips at the same
time; Intel’s Quad model was a single chip.
Sempron (K8)
2004
Budget
desktops
Celeron D (based
on the Pentium
4)
A competitive “value” battle.
Athlon 64 X2 early dualcore models (K8)
2005
High
performance
Pentium D 800
and 900 series
“Presler”
AMD’s dual-core chip was superior to Intel’s
dual-die approach for the Pentium D.
Turion 64 (K8)
2005
Laptops/mobile
Pentium M
Turion was 64-bit vs the 32-bit Pentium M,
but the Pentium offered superior battery life.
Athlon 64 X2 later dualcore models (K8)
20062009
High
performance
Core 2 Duo
To remain competitive, AMD pushed the K8
to 3.2GHz (Athlon X2 6400+).
Figs.23 & 24: the dies of a four-core Opteron “Barcelona” (left) and six-core Opteron “Instanbul” (right) processor.
Sources: https://w.wiki/RQZH & https://w.wiki/RQZJ
siliconchip.com.au
Australia's electronics magazine
December 2026 19
the K8 removed this bottleneck by
building the memory controller
directly into the CPU and using AMD’s
“HyperTransport” technology for fast
communication with the cores.
Internally, the K8 was an evolution
of the K7. Its performance gains came
mostly from architectural improvements, such as the integrated memory
controller, HyperTransport system bus
and AMD64 extensions. By reducing
memory latency and improving the
way the processor communicated with
the rest of the system, the K8 could
often do more useful work per clock
than earlier x86 processors.
Early Slot-A versions of the K7/Athlon used separate Level 2 cache chips
mounted inside the processor cartridge, rather than placing the cache on
the CPU die itself. Later K7 versions,
starting with Thunderbird, moved the
Level 2 cache onto the processor die,
where it could run at the full processor speed. The later K8/Athlon 64 and
Opteron processors also had on-die
Level 2 cache.
In addition, later versions of the K8,
like the Athlon 64 X2, had two processor cores on a single piece of silicon.
The Athlon 64 X2 was not the first
consumer dual-core CPU, as Intel’s
Pentium Extreme Edition 840 and
Pentium D reached the desktop market slightly earlier. However, AMD’s
implementation was widely regarded
as the cleaner and more effective
design, and it was AMD’s first native
dual-core desktop processor.
The K5-to-Opteron era (1996–2003)
was a major transformative period in
AMD’s history. The company completed its transition to a genuine innovator. Although the K5 and early K6
had difficult teething problems, the
engineering and other lessons learned
during those years directly enabled
the spectacular success of the Athlon
and Opteron.
By 2003, AMD had proven it could
design world-class processors, competing fiercely with Intel on both performance and price.
AMD’s entry into the graphics
market
AMD’s entry into the graphics market began with another of its strategic
moves. In July 2006, AMD acquired
ATI Technologies for US$5.6 billion.
This was one of the largest acquisitions in the semiconductor industry
at the time. The deal was intended to
transform AMD from a CPU specialist
into a full-platform provider, capable
of competing with Intel across both
processors and graphics.
The CEO at the time, Hector Ruiz,
believed the future of computing lay
in “heterogeneous integration”. This
is the combining of CPU and GPU
capabilities on the same chip or package, where both processors can work
together on general-purpose computing tasks, not just graphics.
By owning ATI’s Radeon graphics
technology, AMD hoped to accelerate the development of integrated
“Fusion” processors (later called
APUs, accelerated processing units)
that would merge the CPU and GPU
on the same chip.
ATI acquisition
By acquiring ATI in 2006, AMD
gained several critical assets:
1. The well-established Radeon
graphics brand.
2. Advanced GPU architectures,
including the highly regarded R300
family (Radeon 9700/9800) that had
pioneered programmable shaders,
with excellent image quality.
3. Substantial expertise in high-
performance graphics for both consumer gaming and professional workstations (FireGL/FirePro).
4. Early work on combining graphics with other processing functions.
This acquisition transformed AMD
from a CPU-only company into a true
CPU + GPU competitor almost overnight. While the integration proved
difficult and expensive in the short
term, which contributed to AMD’s
financial struggles in the following
years, it laid essential groundwork for
later successes, including Ryzen APUs
and the Instinct line of AI accelerators
(described later).
The ATI acquisition was therefore not merely about entering the
graphics card business. It was a longterm strategic gamble that the future
of computing would involve tightly
coupled CPU and GPU silicon working together. This vision would take
more than a decade to fully mature,
ultimately becoming one of AMD’s
greatest strengths.
Intel was increasingly integrating
basic graphics directly into its processors (Intel GMA, Graphics Media
Accelerator) during 2004-2006, aiming
to control the entire PC platform and
reduce the need for discrete graphics
cards. But for AMD, CEO Hector Ruiz
believed the future of the company
and computing in general was with
heterogeneous computing.
Heterogeneous computing
After acquiring ATI, AMD began
working toward processors that
combined x86 CPU cores and
Fig.25: the ATI Radeon
HD 2900 XT launched
in May 2007, and was
one of the first cards to
be released after AMD
had acquired ATI in
2006. Source: www.
vgamuseum.info/index.
php/home/item/666-atiradeon-hd-2900-xt
20
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
Radeon-
derived GPU cores on the
same die (APUs).
This fitted into the broader idea
of heterogeneous computing, where
different types of processor cores are
used for the tasks they handle best. A
CPU is well-suited to general-purpose
sequential code and operating-system
tasks, while a GPU can process many
similar operations in parallel.
Programming systems such as
OpenCL (Open Computing Language)
were designed to let software use this
mixture of CPUs, GPUs and other processors in a more portable way.
This vision took time to mature.
NVIDIA’s CUDA software platform
became dominant in GPU computing, helped by strong hardware, early
developer support and a growing software ecosystem. OpenCL remained
important, but it did not become
the main programming model for
high-performance GPU computing
in the way its supporters (including
AMD) once hoped.
Even so, AMD’s Fusion/APU thinking was important. It helped establish
the idea that the GPU should was a
major computing resource inside the
system.
Dark years & the Bulldozer
architecture (2008-2016)
Up to 2007, AMD was in a very
strong position with the success and
popularity of the Athlon series of processors, culminating in the Athlon X2,
successfully competing with Intel’s
offerings. Then things took a turn for
the worse during the so-called Dark
Years, roughly between 2007-2016.
While the Athlon X2 wasn’t discontinued until 2009, sales had already
slumped by 2008, and the later X2s
were marketed as budget processors.
Despite strong momentum up to
that point, the company started to
face a series of strategic, technical
and financial difficulties that lasted
nearly a decade and brought it close
to collapse.
The acquisition of ATI was seemingly the correct move based on the
future of heterogeneous computing,
but it added significant debt to the
company. Also, integrating two large
engineering organisations with different cultures proved more difficult
than expected. The development of
the first “Fusion” APUs took longer
than planned and consumed substantial resources.
siliconchip.com.au
On top of those considerable challenges came the global financial crisis of 2008. PC and server demand
weakened sharply, hurting both AMD
and Intel. AMD was more vulnerable because it carried a heavier debt
load, had fewer financial reserves, and
was less dominant in the high-margin
server market. The company reported
a large loss in 2008 and was forced to
cut costs aggressively.
In October 2008, AMD announced
plans to split off its manufacturing
operations into a new company, which
was formally launched as GlobalFoundries in March 2009. This marked
AMD’s transition away from being an
integrated device manufacturer, like
Intel, toward becoming a fabless chip
designer.
From this point on, AMD would
design processors but rely on external foundries, initially and primarily GlobalFoundries, to manufacture
them. The move reduced AMD’s capital burden, but it was not an instant
escape from manufacturing problems.
AMD remained closely tied to
GlobalFoundries through wafer-
supply agreements, so GlobalFoundries’ process delays and yield problems would later cause difficulties of
their own. In the longer term, however, becoming fabless gave AMD
more flexibility, eventually allowing it
to take advantage of TSMC’s (Taiwan
Semiconductor Manufacturing Company) increasingly advanced manufacturing technology.
At the same time, two of AMD’s
major technical initiatives from this
era, Bulldozer and Fusion, produced
disappointing or mixed results. Fusion
helped establish the APU concept,
but Bulldozer failed to deliver the
expected performance and efficiency
gains, leaving AMD to flounder.
Bulldozer
AMD suffered a major technical
setback with its first completely new
CPU architecture since the K8, the
Bulldozer architecture (October 2011).
It introduced a modular design in
which two integer cores shared a single
floating-point unit and other resources
(Fig.26). The flagship of the Bulldozer
architecture upon its release was the
FX-8150 on a single 315mm2 silicon
chip. The multi-chip 16-core Opteron
“Interlagos” had two 8-core chips
in one CPU package (AMD kept the
Opteron name over several processor
generations).
The goal was to improve multithreaded performance and power
efficiency while reducing die area. In
practice, however, the design delivered disappointing results. Single-
threaded performance was often lower
than the previous generation of processors.
Fig.26:
an AMD
Bulldozer
architecture
die (32nm),
possibly the
FX-8150. Each
Bulldozer
module had
two integer
cores sharing
a single
floating-point
unit. Thus,
four modules
gave eight
cores. Source:
https://w.wiki/
RQYj
Australia's electronics magazine
December 2026 21
Technical problems included that
branch prediction was weak, cache
latency was high, and the chip struggled to reach competitive clock speeds.
The FX-8150, despite having eight
cores, frequently underperformed
Intel’s quad-core Sandy Bridge processors in many consumer workloads.
Power consumption was also higher
than expected.
AMD attempted improvements to
Bulldozer with Piledriver (2012) and
Excavator (2015), which brought some
improvements in efficiency and clock
speed, but they did not close the performance gap with Intel’s Haswell and
Broadwell processors.
As a result, AMD lost considerable
market share in both the desktop and
server segments. By 2013, Intel held
roughly 80-85% of the x86 microprocessor market, while AMD’s share fell
into the low teens.
Due to the Bulldozer, Piledriver and
Excavator failures, AMD’s financial situation became critical. They reported
net losses in multiple years between
2011 and 2015. The company’s stock
price dropped below US$2 in 2015;
market analysts questioned whether
AMD could even survive as an independent company.
Multiple staff layoffs occurred, and
AMD was forced to restructure several
times. Manufacturing dependence on
GlobalFoundries (the former AMD
spin-off) also became a disadvantage,
as GlobalFoundries lagged behind
TSMC and Intel in process technology
during this period.
Fusion APUs – an idea before
its time?
While Bulldozer struggled, AMD
pushed forward with its Fusion processors. This was the direct result
of the 2006 ATI acquisition and part
of AMD’s heterogeneous computing
strategy. Fusion was the marketing
name, but AMD gave them the technical name of APUs, hence they were
called AMD Fusion APUs.
However, the Fusion name had to
be phased out starting in 2012 due to
trademark conflicts, leaving the processors to be called APUs. As a result,
they phased out the name Fusion System Architecture (FSA) and started to
use the term Heterogeneous System
Architecture (HSA).
Fusion aimed to combine CPU and
GPU cores on the same die, AMD’s
vision for heterogeneous computing
22
Silicon Chip
at the time (before moving to the chiplet concept).
These early APUs offered very good
integrated graphics and power efficiency, but their CPU performance
(with Bulldozer-derived cores) was
mediocre. They were decent for budget laptops and desktops but couldn’t
compete effectively with Intel’s stronger CPU cores at the time.
Despite these difficulties, AMD
maintained the vision of investing in
its heterogeneous computing strategy.
The company released its first APUs
(later rebranded as Ryzen with Graphics), such as the processors with the
internal development names of Llano
(2011), Trinity (2012), Richland (2013),
Kaveri (2014) and Bristol Ridge (2016).
These generations were all sold
under the name A-Series, with different designations under that.
Transition to chiplets
The progression from the Fusion
concept to a chiplet architecture in
2017 represents one of the greatest
transitions in semiconductor history.
It marked AMD’s shift from a failed
‘all-in-one’ monolithic processor philosophy (CPU, GPU and other parts all
on one piece of silicon) to a modular
‘Lego-brick’ style strategy that fundamentally saved the company.
The successful production of
chiplet-based CPUs allowed for cost
reduction, better yields, process node
optimisation (only using expensive
nodes for parts of the processor) and
excellent scalability.
Throughout the Dark Years, the
Fusion strategy relied on monolithic
dies. As processors get larger, the
chance of a manufacturing defect
increases exponentially. If one tiny
part of a massive monolithic chip is
broken, the entire expensive chip has
to be thrown away (although, in some
cases, sections can be disabled and it
can be sold as a less capable part for
less money).
The scaling problem
Under Fusion, AMD had to spend
immense resources designing completely different monolithic chips for
laptops, desktops and servers. Short of
cash, AMD simply couldn’t afford the
research and development to compete
with Intel across all these markets.
A crucial leadership transition
occurred in 2014. Dr Lisa Su, who had
joined AMD in 2012 as head of global
Australia's electronics magazine
Fig.27: current (since 2014) AMD CEO
Dr Lisa Tzwu-Fang Su. She joined at
the tail end of the ‘dark years’ and has
overseen AMD’s partial overtake of
Intel, thanks to the Zen architecture.
Source: https://w.wiki/3q$z
business development, was promoted
to CEO in October 2014. Su was a
highly respected engineer with a strong
technical background and immediately
began a major restructuring.
She simplified AMD’s product roadmap, reduced operating expenses and
refocused the company on engineering execution. Most importantly, she
initiated the development of a completely new CPU architecture called
Zen, which aimed to address the fundamental weaknesses of Bulldozer.
By 2015-2016, AMD was operating
in survival mode. The company had
managed to preserve its graphics business and made progress on APUs, but it
still remained far behind Intel in CPU
competitiveness.
Observers at the time gave AMD
few chances of a meaningful recovery.
The period from 2007 to 2016 tested
the company’s resilience and forced it
to confront deep structural problems
in design philosophy, manufacturing
strategy and organisational efficiency.
These difficult years also created the
conditions for renewal. The painful
lessons from Bulldozer, the integration challenges from the ATI acquisition, and the financial pressure from
that and the global financial crisis ultimately pushed AMD toward a disciplined, engineering-focused approach
that would define its later success
under Su’s leadership.
Infinity Fabric Architecture
AMD started developing Infinity
Fabric technology around 2012-13,
unveiling it in 2017. They realised
siliconchip.com.au
Fig.28: how Infinity Fabric is used to connect the core complexes (CCDs) and I/O die (cIOD) in a modern AMD processor.
Source: www.igorslab.de/wp-content/uploads/2022/09/zen4guide_slide_4.jpg
that Zen would require a scalable,
modular connection methodology to
make it work. The actual engineering
work was done by Jim Keller and Mark
Papermaster, with Dr Su, then Senior
Vice President (2012-14), providing
executive leadership, engineering
insight and corporate strategy.
Infinity Fabric (IF) is the internal
‘data highway’ for all modern AMD
chips. It is AMD’s proprietary highspeed interconnect technology, introduced in 2017 with the first Zen-based
Ryzen processors (see Fig.28).
It enables modular chiplet-based
designs by providing a means to connect multiple independent silicon
dies, such as CPU and GPU cores, I/O
components, cache memories, memory controllers, AI accelerators and
others, to communicate with each
other at high speeds. It is the main
reason AMD’s chiplet designs are so
effective.
Apart from communication within
the processor module, it also allows
system-wide communication with
peripherals like PCIe cards and USB
devices. It is even scalable to data
centres and exascale supercomputers.
This communication protocol
allows the use of multiple chiplets
in processor designs, which can be
‘mixed and matched’ to produce different variations of processors without having to design new chips from
scratch.
AMD developed Infinity Fabric
to replace HyperTransport, which it
introduced in the early 2000s. While
HyperTransport was advanced for the
siliconchip.com.au
time, it was not flexible enough for the
new chiplet-based design philosophy
AMD wanted to implement.
Infinity Fabric was built to support:
• Multiple chiplets connected
together on the same package
• High-bandwidth and low-latency
communication
• Scalability across desktop, server
and data centre products
Infinity Fabric is a combination of
physical hardware and software-like
communication protocols.
The Physical Layer defines the
wires, circuits and interfaces embedded within the chip to transmit electrical signals. These can be copper or
aluminium wires directly on the silicon chip, on-package traces embedded
in the package substrate that carries
the chiplets, or off-package lanes comprising physical tracks in the motherboard or between CPU or GPU sockets
on a server.
The Control Layer defines the rules.
This is the communication protocol
defining how packets are formatted,
addressed and verified. It also manages cache coherency to ensure that
different parts of the system have the
Table 5: AMD CEOs
CEO
Years in charge
W. J. “Jerry”
Sanders III
1969-2002
Hector Ruiz
2002-2008
Dirk Meyer
2008-2011
Rory Read
2011-2014
Dr Lisa Su
2014-present
Australia's electronics magazine
same idea of what’s in each memory
location despite changes not propagating instantly.
The Control Layer also uses embedded sensors across the silicon chip to
dynamically adjust clock speeds, voltages and data routing based on temperatures and workload demands. The
two main electronic protocols of the
Control Layer are:
1. The Infinity Fabric Data Path
(IFD) is the high-bandwidth pathway
responsible for moving data between
components. It uses high-speed SerDes
(Serialiser/Deserialiser) for converting
between parallel and serial formats. In
Zen-based processors, these links typically run at very high frequencies, up
to 32GT/s (gigatransfers per second) in
recent generations.
2. The Infinity Fabric Control Path
(IFC) handles configuration, synchronisation, power management and coordination between the different dies
and subsystems.
For AMD’s heterogeneous computing philosophy, Infinity Fabric is a key
enabling technology. It allows CPU
cores, GPU cores, I/O dies, memory
controllers and other functional blocks
to exchange data efficiently, both
within a single chip package and, in
some systems, between separate processors or accelerators.
Today, Infinity Fabric is one of
AMD’s most important architectural
advantages. It helped AMD scale EPYC
server processors to high core counts
using chiplets, while also supporting
advanced products such as the AMD
Instinct MI300 series, where CPUs,
December 2026 23
GPUs and high-bandwidth memory
can be tightly coupled inside a single package.
The architecture of Infinity Fabric is
shown in Fig.29. The following abbreviations are used in the diagram:
• 32G SerDes: refers to processors
(like the 4th Gen EPYC series) that support IF lanes/links operating at speeds
up to 32Gbps.
• 2P: two-processor (dual-socket).
A server board setup containing two
physical AMD EPYC CPUs linked via
dedicated IF lanes.
• G Links: dedicated PCIe I/O lanes
originating from the processor that are
explicitly designed to connect GPUs,
storage devices and networking cards.
• P Links: dedicated PCIe I/O lanes
optimised for single-socket (1P) servers. A CPU part number ending in “P”
(eg, EPYC 9654P) is designed to use
all available SerDes links as PCIe I/O
rather than inter-CPU IF.
• D5: a DDR5 (5th generation double data rate) random access memory
(RAM) module.
• 3L/4L: 3-link/4-link specifies how
many physical IF connections are
bound together between computing
resources (such as multiple graphics
dies or separate server dies) to multiply the total bidirectional bandwidth
(up to 400Gbps).
• Fabric + IO: the central I/O die
and Infinity Fabric infrastructure used
to connect the CPU cores, memory
controllers, PCIe/CXL interfaces and
inter-socket links.
• 1DPC: one DIMM per channel,
meaning that each Dual Inline Memory Module has a dedicated communication channel rather than sharing
one between two.
AMD has also worked with industry
partners such as Broadcom to extend
this approach beyond direct chip-tochip links. Broadcom has discussed
support for AMD’s XGMI/Infinity Fabric traffic across future PCI Express
switching technology, including PCIe
Gen 7-era links.
This is associated with AFL, or
Accelerated Fabric Link, a proposed
way to carry AMD fabric traffic, including XGMI, across switched physical
links rather than only direct connections.
Other interconnect
technologies
NVIDIA’s closest competing technology is NVLink, used to connect
GPUs and other processors with a high
bandwidth.
Meanwhile, the broader industry is
developing open interconnect standards such as CXL (Compute Express
Link), UCIe (Universal Chiplet Interconnect Express) and UALink (Ultra
Accelerator Link).
CXL is aimed mainly at coherent
communication between processors,
accelerators and memory devices.
UCIe is designed to connect chiplets
inside a package. UALink, or Ultra
Accelerator Link, is an open accelerator interconnect intended for direct
high-speed communication between
large numbers of AI accelerators in
data centres.
The UALink effort is backed by a
wide group of companies, including
AMD, Apple, AWS, Broadcom, Cisco,
Google, HPE, Intel, Meta, Microsoft
and others. Its goal is to create an
open alternative to proprietary accelerator interconnects such as NVIDIA’s
Fig.29: the Infinity Fabric architecture as used in the AMD EPYC 9004 Genoa. Source: www.servethehome.com/amdepyc-genoa-gaps-intel-xeon-in-stunning-fashion/amd-epyc-9004-genoa-infinity-fabric-overview
24
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
NVLink, reducing dependence on any
single vendor’s ecosystem.
Table 6: AMD’s current products
More on heterogeneous
computing
Product
Heterogeneous?
Features
Ryzen AI
Yes
Ryzen AI 300 series is CPU + GPU +
NPU
Ryzen Desktop/
Mobile APUs
Yes
With Radeon Graphics, CPU + GPU
Instinct MI300/
MI350 series
Yes
Multi-chip CPU + GPU + HBM (High
Bandwidth Memory)
EPYC (many
models)
Highly
High-core count CPUs with integrated
graphics and AI acceleration
Versal Adaptive
(SoCs)
Highly
Arm CPU cores + GPU-like AI engines
+ FPGA fabric
Kria SOMs
Yes
Embedded modules designed for
heterogeneous ‘edge’ AI
Radeon RX
7000/8000 series
No
Discrete graphics cards (shared
driver/software ecosystem with
Ryzen)
Some EPYC and
Ryzen chips
No
Some CPUs lack integrated graphics
AMD later developed the heterogeneous computing idea in several directions: APUs for mainstream PCs and
consoles, Radeon and Instinct accelerators for non-graphics workloads,
and increasingly advanced interconnect technologies like Infinity Fabric.
Chiplets are another major part of
AMD’s modern approach. Rather than
building every processor as one large
monolithic die, AMD can combine
several smaller, specialised dies inside
one package. This improves manufacturing yield, allows different parts of
the processor to be made on different
process technologies, making it easier
to scale core counts.
Intel is also taking a chiplet-like
approach now, albeit much later than
AMD and using different technologies
(we described their system in detail in
the April 2026 issue).
The same broader move toward
heterogeneous and scalable computing is also visible in AMD’s Instinct
MI-
series accelerators. These high-
performance data-centre GPUs are
designed primarily for workloads such
as AI, scientific simulation and high-
performance computing, rather than
conventional graphics output.
So while Intel and AMD both offered
integrated graphics, AMD’s Fusion/
APU strategy was different in emphasis. Intel’s integrated graphics were initially about platform integration and
cost reduction, while AMD wanted
the GPU to become a more powerful
and programmable partner to the CPU.
That vision took many years to
become practical, but it now sits at the
centre of AMD’s processor, accelerator
and chiplet designs.
Today’s computer graphics
market
The computer graphics market
today is dominated by two primary
players: NVIDIA, the clear leader,
and AMD, with Intel trailing as a distant third through its Arc GPU range.
Unfortunately, Arc’s future is looking
uncertain at the moment.
Modern graphics processing units
(GPUs) do far more than simply rendering images on a computer monitor.
They have become the basic hardware
for artificial intelligence and machine
siliconchip.com.au
Table 7: technical terms for modern processor chip designs
Term
Meaning
Usage
APU (Accelerated
Processing Unit)
AMD’s official term for a processor that
tightly integrates CPU cores and GPU
cores (often with an NPU as well) on the
same chip/package.
Primarily
AMD
products
Heterogeneous
processor
Any chip that combines different types
of processors (CPU, GPU, NPU, DSP
[digital signal processor] etc).
Generalpurpose
Heterogeneous
SoC
A System-on-Chip (SoC) that contains
different processing elements.
Generalpurpose
Integrated
Accelerator
Processor
Emphasises the presence of specialised
accelerators (especially artificial
intelligence).
AI chips
learning because they have thousands
of simple processing cores that excel
at performing highly parallel mathematical operations, particularly matrix
multiplications and floating-point
arithmetic.
These operations form the core
workload of neural networks during
both AI training and inferencing tasks.
This massive parallelism gives GPUs a
huge performance advantage over traditional CPUs for AI tasks.
Modern processors increasingly
integrate powerful GPU cores directly
onto the same package or die, making
it less meaningful in many cases to
think of the GPU as a completely separate unit.
While AMD’s current strategy is
based on the idea that future performance gains will come from tightly
integrating different types of processing engines (CPU cores, GPU cores,
Australia's electronics magazine
AI accelerators etc), they still maintain plenty of traditional products
(especially high-end discrete CPUs
and GPUs) to serve specific markets,
generating revenue while the heterogeneous vision matures.
Table 6 indicates the mix of heterogeneous and traditional products. NPU
stands for Neural Processing Unit;
AMD’s name for their NPU architecture is XDNA. For general terminology, see Table 7.
Next month
We will start the third and final part
of this series next month by looking at
the Zen series of chips that has allowed
AMD to take over a significant portion
of the desktop and server CPU market,
as well as make inroads into the mobile
space. We will also give more details on
chiplets, 3D V-Cache, Ryzen AI, Versal,
SC
Radeon/RDNA GPUs and more.
December 2026 25
TECHY GIFTS THEY'LL
WANT TO UNWRAP FIRST?
MARK YOUR CALENDAR:
Thu 3rd to Thu 24th of Dec, 2026
ULTRA FAST
PRINT SPEED
UP TO 500MM/S
350°C HOTEND TO PRINT
TOUGH MATERIALS
MB4029
BUY 2 FOR
$
25
MB4007
SAVE 15%
Waterproof
Smart Fitness Band
Step counter, smartphone
push notifications & more.
QC3113 RRP $14.95EA
$
$
30
SAVE<at>$280
SAVE 20%
FROM BLACKOUTS
TO THE BUSH.
GREAT SOUND
TWS Bluetooth
Earbuds
*
309
MB4093
BUY 2 FOR
NOW FROM
LARGE BUILD
VOLUME UP TO
256 X 256 X
256MM
FIVE MODELS TO POWER
EVERYTHING FROM OVERNIGHT
TRIPS TO JOBS ON SITE.
$
®
Charging case with
digital display.
AA2246 RRP $19.95EA
*True Wireless Stereo
SHOOTS WATER
UP TO 6M
Selected Power Stations
NOW ONLY
UNWRAP
ENDLESS CREATIVITY
699
Centauri Carbon V2
Filament 3D Printer
SAVE $100
Power everything from phones to tools with
USB, 12V and pure sine wave 230VAC outputs.
MB4007 - MB4093
TL5104
BUILT-IN
FPV CAMERA
360° RGB SPINNING LIGHTS
160
BUILT-IN
GAMES
2-WAY VOICE
COMMUNICATION
NOW ONLY
$
1995
$
.
SAVE 10
$
ONLY
3495
.
Electric Water
Gun with LED Nozzle
Available in blue or yellow. Ages 6+.
GT4361 While stocks last
Mini Retro
Game Console
Available in blue, red or
purple. Ages 6+. GT4400
Portable Fridge/Freezer
SALE
NOW ONLY
$
39
95
.
SAVE 10
$
EASY ONE KEY
TAKE-OFF &
LANDING
Mini R/C RGB
Tornado Drone
Ages 8+. GT4918
$
LCD REMOTE
WITH LIVE VIDEO
ONLY
9995 1:20 Scale FPV R/C Car
While stocks last
.
Get a live view as you drive.
250(L)mm. Ages 6+. GT4303
NO ICE. NOICE!
NOICE PROICE
While Stocks Last
^Selected Models
THE PERFECT GIFT, WRAPPED IN A CARD!
*Conditions apply - see page 2 for full T&Cs
1
Prices shown in $AUD, and correct at time of publication but are subject to change. Jaycar reserves the right to change prices if and when required.
KJ9112
WHERE CREATIVITY
MEETS CURIOSITY
KJ9203
12 KITS TO CHOOSE FROM
SAFE, NON-TOXIC
3D PRINTING FOR KIDS
Low Temperature
95 PCL 3D Printing Pen
NOW ONLY
KJ9230
KJ9209
$
34
SAVE $12
Child safe. 2 speed
settings.Rechargeable.
TL4580 While stocks last
ONLY
Walking Robots 3D
Printing Pen Stencil
.
KJ9207
CREATE
WORLDS BLOCK
BY BLOCK
4 SETS TO
CHOOSE FROM
KJ9116
KJ9205
FROM
3495
$
.
ANY 2 FOR
$
Wooden Construction Kits
Space shuttle, steam engine, Christmas
candy stand & more. Ages 14+.
KJ9082 - KJ9240
PRECISION-CUT
WOODEN PIECES
30
SAVE 20%
Magnetic
Building Blocks
$
20mm blocks magnetically attach
together for fun creations. Ages 3+.
KJ9203 - KJ9209 RRP $19.95EA
.
ANY 2 FOR
KJ9124
DIY Bonsai Construction Kits
DRAW YOUR VERY
OWN VIDEO GAMES
3995 Hydraulic Cyborg Hand Kit
SAVE $ 10
Fun to build and use. 229 pieces.
Ages 10+. KR9266
Fish GE4320
Jellyfish GE4322
AVAILABLE IN
2 COLOURS
$
ONLY
4495
$
.
95
30 activities. Over 100 example games. Free iOS™
and Android ® App. Ages 6+. KJ9415 Tablet not included
5995
EA
Mini Arcade
Claw Machine
7-in-1 Board Game Set
Play chess, checkers, Chinese checkers, backgammon,
snakes & ladders, noughts & crosses & ludo. GA0049
Fill the machine with your
favourite treats or toys. 8 capsules
& 24 coins included. Ages 3+. GT4305
GIFTS FOR CURIOUS MINDS UNDER $25
$
ONLY
.
.
Pixicade® Draw
Your Own Video Games Set
EA
Aquarium
RGB Lamps
108-PAGE GAME
DESIGN GUIDEBOOK
ONLY
49
2995
NOW ONLY
$
.
Detailed Sakura & Evergreen Pine designs.
Ages 14+. KJ9122 / KJ9124 RRP $19.95EA
$
ONLY
.
115 PCS
KJ9122
SAVE 20%
$
GE4322
30
Requires TL4580 3D Printing Pen
(sold separately)
GE4320
$
Design & build a walking
robot. Ages 6+. KJ9400
COLOUR-CHANGING
RGB LIGHTS
PICKS UP
LARGE OBJECTS
263 PCS
2495
ADJUSTABLE
LAUNCH ANGLES
LAUNCH UP TO
4 ROCKETS IN
ONE STOMP
ONLY
495
.
3D Rotating
Puzzle Maze Cube
Ages 8+. KJ9108
$
ONLY
11
95
.
Build Your Own Solar System Kit
Ages 8+. KJ8994
$
ONLY
1995
.
Fossils & Minerals Kit
Ages 6+. KJ9115
$
ONLY
2495
.
Rocket Launcher
Includes 8 rockets. GT3078
TERMS & CONDITIONS: Prices valid from 3/12/26 to 24/12/26. Stock may be limited on sale items. No rain checks. Savings on Original RRP (ORRP). Prices are correct at time of publication but are subject to
change. Jaycar reserves the right to change prices if and when required. Page 1: MULTIBUYS: 2 x QC3113 for $25. 2 x AA2246 for $30. Page 2: MULTIBUYS: Any 2 of KJ9203 / KJ9205 / KJ9207 / KJ9209 for $30.
Any 2 of KJ9122 / KJ9124 for $30. SUPPLY CHAIN DISRUPTION: We apologise for factors out of our control which may result in some items not being available on the advertised on-sale date of the catalogue.
For full gift card T&Cs see www.jaycar.com.au/giftcards. For full T&Cs see www.jaycar.com.au.
2
FROM BRIGHT IDEAS TO
FINISHED PROJECTS
NOW FROM
$
17 95
.
SAVE 10%
TL6424
Selected 1kg Filaments
TL6400 - TL6530
TL6411
TL6400
HUGE RANGE OF
COLOURS & MATERIALS
NOW ONLY
NOW ONLY
$
$
314
'CREATE' THIS
CHRISTMAS
DESIGNED FOR
INTRICATE DETAILS
Mars 5 Resin 3D Printer
Centauri Carbon V2
Combo Filament
3D Printer
TL5010
Featuring 6.6" 4K mono LCD with COB
light source, automatic levelling & smart
mechanical sensor. TL4830 While stocks last
$
SEE IN-STORE FOR MORE OFFERS
ON RESIN 3D PRINTERS
FROM
149 Filament Dryer Boxes
TL5010 / TL5012
Speed, stability and simplicity in
one premium unit. Print in up to
four colours without manual spool
swapping. TL4986
1
ROTARY TOOL
110 PCS
NOW ONLY
350°C
HOTEND TO
PRINT TOUGH
MATERIALS
SAVE $100
SAVE $35
$
899
TL5012
MULTICOLOUR CREATIONS WITH
SEAMLESS 4-COLOUR PRINTING
POWER THROUGH SANDING,
SHAPING, POLISHING & DETAILING
4995
POLISHER
.
SAVE 15%
XC0400
2
DRILL, SAW, SAND,
POLISH, CARVE OR GRIND
12V Rotary
Tool Kit
TD2451
BELT SANDER
4.3" COLOUR
SCREEN
$
NOW ONLY
$
99
99
SAVE<at>$110
SAVE $ 40
UP TO
600X ZOOM
720P Digital
Microscope
XC0434
NOW FROM
Wireless Weather Stations
1
Rechargeable. LED
illumination. QC3193
2
$
MEASURES & RECORDS
TEMPERATURE,
HUMIDITY,
RAINFALL & MORE.
5.5" Display XC0400
NOW ONLY $99 SAVE $20
Colour Display XC0434
NOW ONLY $199 SAVE $110
$
24
NOW ONLY
$
95
.
.
SAVE $12
10W Mini
Rechargeable
Glue Gun
Suits 7mm glue sticks. TH2024
3
2950
48
PCS
Micro Belt Sander
with Rotary Tool & Polisher
Supplied with an extensive range of sanding
& polishing accessories. TL8119
MAGNETIC STORAGE
Screwdriver Set
Includes tamperproof
type bits. TD2134
29
50
.
SAVE 15%
Cat III
Multimeter
QM1527
IDEAL ENTRY-LEVEL
SOLDERING STATION
NOW ONLY
$
NOW ONLY
$
129
SAVE $20
HANDY GIFTS UNDER $40
ONLY
NOW ONLY
3995
.
EASY
TO USE
NONCONTACT
VOLTAGE
SENSOR
SAVE 22
$
ADJUSTABLE
TEMPERATURE
UP TO 450°C
48W Soldering Station
Lightweight iron with anti-slip grip. TS1620
Prices shown in $AUD, and correct at time of publication but are subject to change. Jaycar reserves the right to change prices if and when required.
DC9014
FOR THE ROAD & WHERE
IT TAKES YOU
DC9091
WIRELESS
QI CHARGER
DC9049
10%
$
OFF
99
SAVE $ 50
12V 1000A
Jump Starter and Powerbank
DC9014 - DC9098
INCLUDES DUCTING & ADAPTORS
399
Portable Air Conditioner
Powerful, ultra-quiet, and designed for convenience wherever
you go. Mains powered. GH1574 While stocks last
Whopping 1000A cranking current. MB3986
2
NOW ONLY
SAVE $ 200
BRIGHT
LED TORCH
WITH SOS
.
Selected
Radios and Accessories
$
DUAL
USB PORTS
NOW ONLY
MIGHT JUST SAVE YOU
FROM SPEEDING FINES
5100 BTU
COOLING
MASSIVE
12000MAH CAPACITY
5"
7"
LA9025
TOUCH
SCREEN
1
LA9036
NOW FROM
$
89
95
.
SAVE $ 10
NOW ONLY
NOW ONLY
$
GPS Speedometers
149
$
SAVE $ 20
Add GPS accuracy to any vehicle.
OBD-II Head-Up
1 LA9036 NOW $89.95
Rechargeable
2 LA9025 NOW $129
ADD APPLE CARPLAY® OR
ANDROID AUTO™ TO YOUR CAR
Wireless Smart Monitor
169
SAVE $ 80
REVERSING MADE EASY
Wireless Reversing Camera Kit
Built-in speaker, FM radio, auto-dimming,
and reversing camera input. QM4002
Easy to install and use. 2.4GHz digital
signal for crystal clear picture. QM3842
STAY COOL THIS SUMMER
HUNT FOR COINS, RELICS,
JEWELLERY, GOLD, SILVER, ETC.
VIEW METER
WITH POINTER
LED LIGHT
NOW ONLY
$
5995
$
SAVE 10
SAVE 10
$
ONLY
399
NOW ONLY
.
$
Portable Rechargeable Camping Fan
100Ah 12.8V Lithium Deep Cycle Battery
Integrated powerbank to keep your devices
charged. Up to 24hr run time. 4hr timer. GH1295
Half the weight of lead-acid equivalent. SB2215
HOT DRINKS ON THE ROAD TRIP
.
$
6000+ CYCLES
7995
WATERPROOF
SEARCH COIL
ADJUSTABLE
LENGTH
Metal Detector
With auto tune.
QP2302
GIFTS FOR THE GETAWAY UNDER $50
AIR VENT & SUCTION
MOUNTS INCLUDED
PERFECT FOR
CHRISTMAS TRIPS
NOW ONLY
NOW ONLY
$
NOW ONLY
29
95
.
SAVE $10
SAVE $15
95
.
SAVE 15%
1L 12V Kettle
GH1386
39
$
$
12V Dual 4" Fans
GH1403
$
NOW ONLY
49
95
.
SAVE $ 12
.
Phone Holder
HS9062 While stocks last
4995
15W WIRELESS
CHARGER
12V Portable Stove
Cook & warm up food. 3L capacity.
YS2811
Prices shown in $AUD, and correct at time of publication but are subject to change. Jaycar reserves the right to change prices if and when required.
4
Advanced
Cable Analyser
by Richard Palmer
Faulty cables
are the bane of
electronics and
audio enthusiasts
alike. While the
most common faults result
from wires parting company with
connectors and shorts between pins, highresistance and intermittent connections can also cause havoc.
E
ven the most basic cable testers cost
over $100 despite supporting a limited range of connector types – some
you will use and possibly many that
you will not.
Accommodating all potential cable
connectors is impossible in the limited space available on a reasonably
sized case. Looking at audio connectors, while RCA connectors come in a
single size, there are three-, four- and
five-pin XLR connectors and even
more DIN variations. There are three
common sizes of TRS jacks with two,
three or four pin variants. The list
goes on.
Rather than specifying which connectors should be used, in this project
we have provided space on the sides
and lid of the enclosure for constructors to select and lay out their preferred
30
Silicon Chip
plugs and sockets. If you run out of
space to mount less-often used connectors on the case, adaptor leads can
be made and plugged into the onboard
8P8C connectors (often incorrectly
called RJ45).
Locating high-resistance connections is difficult with low-
current
testing, as small voltages are induced
between the cable ends that are
hard to measure accurately. This
unit tests cables at 120mA, allowing
cable resistances between 0.2W and
15W to be measured with reasonable
accuracy.
As each pin combination is tested
for less than one millisecond, battery
life is not unduly affected by the substantial test current.
At the completion of the test cycle,
a map of the cable, including bridged
Australia's electronics magazine
Features & Specifications
▶ Comprehensively analyses cables
with up to eight wires
▶ Measures cable resistance
▶ Displays a cable map, including
bridges between pins
▶ Results are visible in full sunlight
▶ Continuous test mode for
intermittent faults
▶ Comparison against a reference
cable
▶ A selection of test connectors
▶ USB or battery powered
▶ Test connector expansion port
pins and cable resistance, appears on
the display, then the unit switches off
to conserve battery power. Since it’s
an e-paper display, the image remains
even when the electronics are powered
off. Being reflective rather than backlit, it’s legible under a wide range of
lighting conditions, including direct
sunlight.
Intermittent cable connections can
be frustrating to trace. A long press on
the start button puts the unit into continuous test mode, where all pin combinations are scanned ten times per
second. If a substantial change in readings occurs, the unit stops scanning
and shows the offending connection.
Cables can also be tested against
a known-good cable. Connecting
the good cable followed by two long
presses on the start button stores the
reference configuration. Short presses
make a comparison between the cable
currently connected and the reference
map, displaying any differences.
When the unit’s USB cable is
plugged into a computer, the connection maps are also sent to the computer
in text format.
Improving a traditional design
Commercial cable testers generally
connect LEDs to the pins on one connector and pass a current through each
pin, lighting any LEDs where there is
continuity with the other connector
(Fig.1). The process is usually driven
by a rotary switch or simple electronic
sequencing of the test voltage.
Any cable shorts or bridges between
pins will light additional LEDs, but
only when there is continuity to the
driven pin. If there is no wire between
the pins, either due to design or a poor
connection, pin-to-pin bridges will not
immediately be detected.
siliconchip.com.au
The open circuit will be detected,
but the tester will not reveal all faults
simultaneously. You would have to
fix the open-circuit fault, then re-test
the cable to find the short-circuit fault.
Moderate cable resistances will not
visibly dim the LEDs, so it is not possible to detect high-resistance connections with this method.
To provide more information
on cable resistance and pin-to-pin
bridges, the common pole of the switch
can be connected to a constant current
sink (Fig.2) along with a low-value
resistor on the high-side of each pin.
If the current is known, the wire resistance can be calculated by measuring
the voltage at each end of the cable
and applying Ohm’s Law.
A substantial test current helps to
generate a measurable voltage drop
even when the cable resistance is low.
However, pin-to-pin bridges where
there is no cable connection will still
remain undetected.
If a resistor of a substantially higher
value than any expected cable resistance is connected between each pair
of A and B pins (Fig.3), a small current will flow even when there is no
cable connection. If there is no cable
connection, the B-side pin will float
high unless driven low.
If there is an A-side bridge on an
unconnected pin, the voltage on the
A-side undriven pin will drop slightly
with the driven A-side pin. For a
B-side bridge, the undriven pin’s B
voltage will also be pulled down.
Where there is a B-side bridge and a
cable connection to both pins, it is not
possible to determine whether there
is also an A-side bridge, as both the
Fig.5: the TLC59108 functional block diagram.
Source: Texas Instruments
A pins will be at the same potential
whether there is a bridge or not. This
brings us to the final configuration for
this project (Fig.4) where each A-side
pin also has a switch.
With this configuration, if the
A-side switch for the driven pin is
opened, the A-side voltage for any
bridged pins and the driven pin will
remain high. A mixture of bridged
pins and crossed-over wires makes
analysis more complex; the approach
taken is explored below.
TLC59108 LED driver
This design is made simpler by
using a specific IC, the TLC59108 constant current sink LED driver (Fig.5).
It is designed to provide a controlled
current sink for eight LED strings. It
has a 20V maximum LED source voltage, allowing it to power up to eight
LEDs in each string at programmable
currents up to 120mA.
Each string has an independent 8-bit
pulse-width modulation (PWM) controller, operating at 97kHz, to control
LED brightness. There is also a master
controller that can dim or blink all the
LED strings together. The PWM controllers can be disabled, allowing on/
off operation of each channel at a single, programmed current.
The programmed maximum sink
current is a combination of the current flowing through the external current programming resistor, REXT, and
multiplier settings in one of the control registers. 256 different current
values from 4mA to 120mA can be
programmed.
The chip is controlled over a twowire I2C serial bus and requires no
external components other than resistor REXT. There is a sister chip, the
TLC59108F, which does not offer the
current programming facility; instead,
it relies on a series resistor in the LED
Figs.1-4: (1) common commercial cable testers comprise a battery, LED, series resistor and a switch pole for each
connector pin. (2) Replacing the battery with a constant current source allows cable resistance to be calculated from the
connector voltages using Ohm’s Law. (3) Adding a resistor with a value much higher than the cable resistance between
connector pins enables bridged pins to be detected on the low side. (4) Adding high-side switches enables detection of
high-side bridges when low-side bridges are also present.
siliconchip.com.au
Australia's electronics magazine
December 2026 31
string. The F variant is not suitable for
this project.
Circuit details
The full circuit is shown in Fig.6.
The three main elements in the design
are the test path, the controlling microprocessor and the display.
Following the test path from the
3.3V rail, the A-side switches are
implemented using P-channel Mosfets Q3-Q10. They were selected for
their low on-resistance (<0.1W) and
low gate-source threshold voltage (Vgs,
-1.5V). Each Mosfet is controlled by
an output from IC1, an eight-channel
32
Silicon Chip
PCA9554 (or PCA9554A) I2C bus
expander, which provides individual
control of each channel.
The 10W high-side resistors lead
to the A-side test connector, and the
150W resistors bridge to the B-side test
connector.
The values of the two resistors are
compromises. The high-side resistor
value needs to be high enough that
there is a measurable voltage across it
when there is a cable connection, yet
low enough that the voltage across it is
much less than 3.3V when the 120mA
test current passes through it.
The value of the bridging resistor
Australia's electronics magazine
needs to be substantially greater than
any expected cable resistance, while
still generating a measurable voltage
across the high-side resistor for an
open-circuit connection. The voltage
across a 27.5W resistance will be 3.3V
at 120mA. So, with 10W series resistors, the maximum measurable cable
resistance is 17.5W.
In practice, it is a little less, as the
Mosfet and IC4 have a small series
resistance and the 150W parallel resistors also draw a small current.
The B-side pins are connected to the
eight individually controllable constant current sinks of the TLC59108
siliconchip.com.au
Scope 1: the yellow trace of the LED
pin voltage shows the test and display
cycle. The green 3.3V reference
voltage is stable throughout.
Fig.6: under the Pico’s control, Mosfets Q3-Q10 apply voltage
to the pins of the connector at one end of the cable while the
TLC59108 LED driver (IC4) sinks current via the pins at the
other end. The voltages at each connector pin are sensed by
the microcontroller’s A0 and A1 ADC pins via 8:1 analog
multiplexers IC2 & IC3, which are sequenced in unison.
LED driver (IC4). The TLC59108’s individual, group dimming and flashing
capabilities are not required for this
application; each channel is simply
switched on or off at the programmed
sink current. Only one channel is
driven at a time.
A pair of 8:1 analog multiplexers,
IC2 and IC3, select the A and B connector pins to be measured by the microcontroller’s internal analog-to-digital
converter (ADC).
74LV4051 and 74HC4051 chips
both have an average series resistance
of around 100W with a 3.3V supply,
which varies by up to 100W across
siliconchip.com.au
their rail-to-rail input range. This does
not cause any significant measurement error, as the resistance is much
smaller than the Pico ADC’s 100kW
input impedance.
CD4051 multiplexers are not a good
substitute, as their series resistance
is five times higher at a 3.3V supply
voltage.
The choice of microcontroller came
down to cost, ease of construction and
programming. For ease of programming, the Raspberry Pi Pico 2 wins
hands down as it can be programmed
on any major computer platform without additional software or hardware.
Australia's electronics magazine
It costs only a few dollars more than a
bare PIC or ATmega chip and already
has the necessary ancillary components on board.
Most of the processing is undertaken
on CPU 0 of the RP2350 chip, with the
buzzer beeps and LED flashing handled by CPU 1 so they don’t delay test
and display processing.
The Pico module has an onboard
buck/boost switch-mode regulator
that can generate 3.3V from a wide
range of input voltages (1.8V to 5.5V).
This allows significant flexibility in
powering the unit from a battery or
USB power. The Pico datasheet indicates that up to 300mA can be drawn
from this supply, well in excess of the
150mA consumed when sinking a single channel at 120mA.
The 3.3V supply is used as the test
voltage as well as the ADC reference,
automatically scaling voltage measurements if the test voltage varies. Scope
1 shows that the regulator copes well
with the 120mA test current, with its
output voltage varying by less than the
10mV switching noise.
The Pico’s ADC has a nominally
twelve-bit resolution. In practice, the
accuracy is limited to around ten bits
or 1024 (210) steps. A 3.3V reference
voltage yields a step size of 3.2mV
(3.3V ÷ 1024). when sinking 120mA,
this step corresponds to a resistance
resolution of 27.5mW.
The three 1nF capacitors on the analog microcontroller inputs provide a
low-impedance source for the ADC’s
sample-and-hold circuitry as well as
filtering out any high-frequency noise
that may be present.
The 8P8C connectors, CON6 &
CON7, are in parallel with the A and
B pin headers. As well as enabling the
testing of network cables, they are a
December 2026 33
convenient connection point for test
harnesses.
The e-paper display is controlled
via a three-wire SPI serial interface,
similar to that required by TFT displays. The additional BSY pin is
polled by the Pico so that new commands or data are not sent while the
display is updating. This is necessary
as display update and erase cycles can
take a second or two to complete.
The piezo transducer is driven by
a 50% duty cycle PWM signal from
the GPIO9 digital output pin on the
microcontroller via a 100nF capacitor. A passive device is used, as two
different tones are required. The LED
connected to digital output GPIO2
provides a visual indication of test
progress.
The Pico has an internal means of
sensing when power is present at the
USB connector. This signal determines
several differences in operational
behaviour between USB and battery
power. The input voltage level is also
read by ADC channel A2 via a voltage
divider to estimate the remaining battery capacity if one is used.
A key challenge in laying out the
main PCB was minimising the resistance of the traces from
the A-side Mosfets
and source
resistors to the A-side cable connectors
and from the B-side connectors to IC4
to minimise resistance in the test path.
To simplify construction and provide a decorative faceplate for the
project, the enclosure’s lid is replaced
by a second PCB on which switch S1,
the display and LED are mounted, as
well as some of the test connectors.
Battery life
As the cable tester is designed to be
a toolbox item, it needs to be battery-
powered, so its running time on a set
of AA cells was a key design criterion.
To reduce overall power consumption,
we looked for components with low
current requirements and sought to
minimise the current drain while the
unit was idling.
The three main consumers of current in this design are the microcontroller, the display and the TLC59108
current sink. The multiplexers and
bus expander draw minimal current.
The Pico 2 consumes 18mA when
running at 150MHz, 9mA at 50MHz
and 2mA when in sleep mode. As the
computing power of the Pico is well in
excess of our needs, it runs at 50MHz
to minimise its operating current.
While putting the microcontroller to
sleep would have provided an additional benefit, the options available are not sufficiently flexible
to make this attractive.
For TFT displays, the backlight LED can consume up
to 40mA at full intensity,
which can be reduced by
dimming the LED using
a PWM signal. Small
OLED displays average around 20mA.
E-paper displays,
o nce updated,
consume no current.
As retaining
the display
as long as
Photo 1: I
mounted
RCA, barrel,
banana and
XLR sockets
on my unit
but you could
use a different
combination.
34
Silicon Chip
is needed to identify and rectify any
cable faults is a useful feature, e-paper
becomes an obvious display choice.
The TLC59108 current sink (IC4)
draws nearly 30mA of idle current.
Even in its lowest power mode, when
reset is held low, the chip consumes
only 2mA less. This is not particularly
significant when driving eight LED
strings at 120mA each, but it is more
than an acceptable constant battery
drain for extended idle use.
As alternative approaches for creating the current sink and low-side
switching functions added significant cost or circuit complexity, we
sought to reduce its average power
consumption.
Instead of dealing with the microcontroller and IC4’s idle currents separately, we switch off power to the
entire circuit at the end of each test and
display cycle using P-channel Mosfet
Q1. The reverse leakage current of D1
and the off-state current of Q2 are both
in the nanoampere range. A battery
power switch (S2) is also provided to
prevent accidental toolbox triggering.
Tests run in single-shot mode; this
is a very effective strategy, as the active
cycle is only a few seconds long. For
continuous and comparison modes, a
reference cable map needs to be stored
between test cycles, and the current
sink IC needs to remain active, so the
available power savings are limited.
USB power is advised when undertaking cable test sessions that use these
modes for an extended period.
The minimum voltage at which
the unit will function properly is primarily determined by the battery discharge curve. Alkaline cells are 90%
discharged at 1V. The Pico’s onboard
regulator is specified to operate from
as low as 1.8V, and Q1 was selected
to switch on fully with a similar Vgs.
To provide some headroom for the
regulator and Q1, the unit shows 0%
battery remaining and switches off
with a battery voltage just below 3V.
Thus, a three-cell alkaline battery is
ideal. AAA alkaline cells will deliver
around 1000mAh when discharged
to 1V, which should allow many hundreds of five-second test cycles at an
average 50mA current (see siliconchip.
au/link/accr).
The test cycle
When the start button (S1) is
depressed, P-channel Mosfet Q1 conducts, connecting the battery to the rest
Australia's electronics magazine
siliconchip.com.au
of the circuit. Once the microcontroller
has booted, PWR goes high, sustaining
power by switching on NPN transistor Q2 until the signal is released after
the completion of the testing cycle,
switching off Q1 and shutting everything down.
The Pico’s GP12 pin has its internal
pull-up resistor enabled, so the input
is normally high when the microcontroller is powered on. The pull-up
resistor also provides bias to S1 once
PWR has been pulled low. Diode D1
ensures that PWR and START are
isolated so that S1 can still be sensed
when PWR is low.
The test current is passed through
P-channel Mosfets Q3-Q10 and the
10W resistors on the high side, through
the cable being tested, to a channel on
the TLC59108 constant current sink
(IC4). During testing, IC4’s pins and
the Mosfets are enabled as required,
and the voltages at each of the A and
B pins are selected by multiplexers IC2
and IC3 for measurement at the ADC
pins on the microcontroller.
While the two multiplexers have
separate channel select pins on the
microcontroller, in practice they are
always sequenced together. Mosfets
Q3-Q10 conduct when their gates are
pulled low by IC1.
In single-shot mode, IC4’s outputs are enabled in sequence at their
120mA maximum current setting, with
all A-side Mosfets enabled. At each
step, the multiplexers cycle together
through each channel, and the A and
B voltages are read. The driven channel A-side Mosfet is switched off and
a second pair of readings taken.
Over the 30ms test cycle, a total
of 256 measurements are stored for
later analysis. In most cases, only the
readings with all the A-side Mosfets
enabled are required for analysis.
If Va is close to 3.3V and Vb is low
when the corresponding B pin is
driven, the connection is open. For
a pair of connected pins, Va and Vb
will be close in value. The resistance
is calculated by applying Ohm’s Law
to the voltage difference, as the current is known.
A cable map entry is created for each
pair of connected pins, flagged with a
“C” (for connected) and the resistance
recorded if it is greater than 0.2W. If
the resistance is more than 15W, an
open circuit is assumed. Crossed wires
and multiple connections are detected
and logged.
siliconchip.com.au
Parts List – Advanced Cable Analyser
1 double-sided 80 × 62mm green PCB (main) coded 04101261
1 double-sided 89 × 152mm blue PCB (lid) coded 04101262
1 Raspberry Pico 2 microcontroller module with header pins
1 1.54-inch 200 × 200-pixel Waveshare monochrome e-paper display with
cable [Core Electronics WS-12955]
1 8-pin 0.1-inch/2.54mm header for display (CON4) [Core CE08367]
2 8P8C PCB-mount sockets without LEDs (CON6, CON7) [Amphenol 54602908LF, Oupiin 8949-F188-06BNA, Altronics P1438, Core PRT-00643] ●
2 20-pin 0.1-inch/2.54mm header sockets for Pico 2 [Core CE09119]
1 3V passive piezo transducer (PB1)
[Altronics S6109, Jaycar AB3442, Core ADA1740]
1 12mm tactile switch with 3.5mm or higher round stalk (S1)
[Jaycar SP0609, Core COM-12993]
1 miniature slide switch (S2)
[Altronics S2010, Jaycar SS0852, Core CE03160]
1 3×AAA flat battery holder (BAT1)
[Altronics S5040, Jaycar PH9272, Core FIT0626]
3 AAA alkaline cells
1 micro USB Type-B to Type-A cable (for programming and USB operation)
multiple pairs of test connectors (constructor’s choice)
various colours of hook-up wire (lengths vary with connector choices)
● Jaycar PS1478 is not suitable
Semiconductors
1 PCA9554(A)D I2C bus expander or TCA equivalents, SOIC-16W (IC1)
[Mouser 595-PCA9544ADWR]
2 74HC4051 or 74LV4051 8:1 analog multiplexer ICs, SOIC-16 (IC2, IC3)
1 TLC59108IPWR (not F variant) constant-current LED driver, TSSOP-20
(IC4) [Mouser 595-TLC59108IPWR]
9 AO3401(A) P-channel Mosfets or similar, SOT-23 (Q1, Q3-Q10)
[Core COM-29463, Mouser DMP1045U-7]
1 BC817 or similar NPN transistor, SOT-23 (Q2)
1 3mm red LED, bezel optional (LED1) [Altronics Z0200, Jaycar SL2615]
1 BAV70 or similar dual common-cathode signal diode, SOT-23 (D1)
[Altronics Y0099]
Hardware
1 UB1 Jiffy box
3 M3 × 15mm tapped male/female spacers with nuts
3 M3 × 6mm countersunk head machine screws
2 M2 × 6mm panhead machine screws [Altronics H3101, Jaycar HP0390]
4 8mm clear round rubber bumpers [Bunnings]
1 8cm length of double-sided mounting tape
Capacitors (all SMD 2012/0805 X7R or X5R 50V SMD ceramic)
3 1μF
5 100nF
3 1nF
Resistors (all SMD 2012/0805 ⅛W ±1% unless noted)
1 100kW
3 3.3kW
1 33kW
9 150W
1 22kW
8 10W SMD M3225/1210 or M3216/1206 ¼W ±1% SMD
3 10kW
[Altronics R7972]
Bridge detection follows. There
are several different cases for bridged
pins: A- or B-side bridges, where the
A and B pin pairs are connected or
not connected. Each requires specific
treatment.
Where there is a cable connection
from any A-side pin to the driven
B-side pin, any non-driven pin developing a significant voltage across its
Australia's electronics magazine
10W resistor it is flagged as having an
A-side bridge to the connected A-side
driven pin.
A B-side bridge will cause a much
smaller voltage to be created across
the 10W resistor, which is ignored. If
the bridge is on the B-side, Vb on the
bridged, the undriven pin will also fall
and be flagged with a “B”.
Where there is no cable connection
December 2026 35
to the driven pin, an A-side bridge will
cause a smaller drop in Va for each of
the bridged pins. If the bridge is on
the B-side, Va for the bridged pins will
drop when either B-side pin is driven.
These are also flagged with a “B”.
Where there are both A-side and
B-side bridges on the same pin pairs,
the second set of readings, with the
A-side Mosfet open, come into play. If
the process above finds a B-side bridge
for a pair of pins, if the alternative Va
is high, the pair is marked as an A-side
bridge, but only if the reversed pair of
pin numbers has already been flagged
as a B-side bridge.
Where there is a bridge between two
connected pins, there’s no electronic
means of discovering which end of a
low-resistance cable has the bridge. In
this case, both A and B side bridges are
electrically identical. The unit may
flag these cases as crossover connections or B-side bridges.
In the rare case that the unit detects
a connection and cannot determine the
type, it displays a dotted line bridge
on the B side and marks the value
with a “?” plus the two pin numbers
on the right-hand side of the display.
In both these cases, the eyeball is the
best test instrument to make the final
determination.
The resulting cable map is displayed,
with resistances greater than 1W shown
as dotted lines. At the end of the cycle,
the unit switches off to save power,
leaving the cable map displayed. If the
latest test cycle found no connections
or bridges, the display is cleared.
In continuous mode, the cable map
from the first cycle is saved, and the
maps from subsequent cycles are compared to it. If a difference is found –
a missing, extra, or higher resistance
connection or bridge – testing ceases
and only the difference is displayed
on the screen. The type of difference,
a missing or extra connection, is noted
on the bottom line of the display.
In comparison mode, the initial
cable map is stored as a reference, and
the unit makes one comparison scan
for each short button press. Results
are displayed as for continuous mode.
If there has been no activity for one
minute, the unit flashes the LED and
beeps once as a reminder that it is still
switched on.
Construction
The Advanced Cable Analyser
is built on a double-sided PCB that
measures 80 × 62mm and is coded
04101261. The lid is also a PCB, coded
04101262, measuring 89 × 152mm.
The PCB mounts on 15mm-long
tapped spacers with the component
side down in the bottom-right corner
of the Jiffy box so that the 8P8C and
Pico USB connectors can be accessed
from outside the case. The screws
through the bottom of the box should
be countersunk to avoid bumps in the
instruction card.
The box is much larger than the
PCB to allow for cable connectors to
be mounted on the sides, top and lid.
The display, battery holder, power
and start switches will be mounted on
the underside of the lid. A pin header
is used to connect the e-paper display
as it comes with a suitable cable. The
rest of the off-board wiring (S1, buzzer
and battery) is soldered directly to the
PCB. The Pico is mounted on header
sockets.
Before mounting any components
on the main PCB, pare off the vertical
ribs in the case next to where the Pico’s
USB connector will lie with a chisel or
sharp knife. There is adequate clearance on the bottom edge of the case
with the PCB ribs left in place, as the
8P8C sockets protrude 5mm past the
edge of the PCB.
Place the blank PCB component-side
down in the bottom right corner of the
case, then mark and drill the 3mm
mounting holes for it – see Photo 3.
Now refer to the PCB overlay diagram, Fig.7, to see which components
go where on both sides of the board. On
the main PCB, mount the SMD parts
on the Pico side first, followed by the
19 that are on the opposite side of the
board, including one more IC and one
transistor.
Various techniques can be used,
including hand-soldering, hot air
reflow, oven reflow, or hotplate reflow
(probably only suitable for parts on one
side, though).
Hand-soldering with a standard
temperature-controlled soldering
iron is possible as long as you
use a good-quality flux paste. We
Photos 2 & 3: the bare PCB is used to
mark the mounting holes in the case.
The ribs next to the Pico have been
pared off to allow access to the USB
connector. Only the headers for the
display are required; the wires for
the battery, buzzer, LED and switch
connections can be soldered directly
to the PCB pads.
36
Silicon Chip
siliconchip.com.au
Fig.7: fit the components to both sides of the board like this. We’re showing pin headers for CON1-CON5, which could
have DuPont wires plugged into them, but you could also solder wires directly to the pads (except CON4). Don’t use
headers for CON8 and CON9; it’s better to solder those wires directly to avoid extra resistance in the measurements.
recommend having some solder-wicking braid and flux cleaner (eg, pure
alcohol) on hand, plus fine-tipped
tweezers.
Even with hand-soldering, there are
various techniques that can be used.
You would usually start by placing
an IC in its correct orientation, tacking one pin, then verifying that all its
other pins are centred on their pads. If
not, remelt that initial solder joint and
gently nudge it into position. Repeat
until you are happy, then add flux
paste to all the pins.
One option is then to add some solder to the clean tip and drag it along
one edge of the IC (not the one with
the tacked pin!) so that solder flows
onto all the pins and pads, aided by
the flux paste. Another is to add a
small amount of solder to the clean
tip and touch it to the end of one pin,
observing that solder flows onto the
pin and pad, then repeating that for
the other pins.
Once all pins of an IC have been soldered, check under magnification for
solder bridges between them, or for
any joints where solder has not flowed
properly onto both the pin and the pad.
If you find any, add a little extra flux
paste to the pin(s) and reflow them. If
that doesn’t fix it, apply braid to the
effected pin(s), press down with the
soldering iron and slide it away after
a few seconds.
The TLC59108 chip has closely
spaced pins and you probably won’t be
able to avoid bridging them but luckily
it’s quite easy to clean them up with
some extra flux paste and solder wick.
Just be careful not to pull the wick too
hard until the excess solder has melted
siliconchip.com.au
and flowed into it, at which point it’ll
be easy to slide away from the pins.
When all the solder joints on a chip
look good, clean off the flux residue and
proceed to the next one. Other SMD
parts like transistors, diodes, resistors
and capacitors can be soldered similarly; they just have fewer pins.
Next, solder in the pin sockets for
the Pi Pico 2, followed by the 8P8C
sockets on the top of the PCB, and the
8-pin display header on the underside.
After that, add the rest of the components on both sides of the PCB.
Mark the positions of the 8P8C
connectors on the bottom right of the
case using Fig.8 and drill 6mm holes.
Position the PCB on its spacers and
with the Pico removed, enlarge the
8P8C connector holes with a file.
Mount the Pico and repeat for the
USB connector, using Fig.9 as a guide.
Start enlarging the hole from the
8P8C end, as the connector will move
slightly toward that end once the hole
is completed. There is no need to leave
extra space around the USB connector,
as a standard USB-micro plug is long
enough to connect properly through
the side of the case.
Fig.8: make
the openings
for the 8P8C
connectors by
drilling 6mm
pilot holes
and enlarging
them with a
file. These
diagrams are
views from
the outside of
the case.
Fig.9: after
the 8P8C
openings
have been
made, mount
the Pico and
position
the USB
connector
hole. Drill
a 2mm hole
and enlarge it
with a file.
Australia's electronics magazine
December 2026 37
Before mounting the Pico 2 module
on the PCB, load the firmware, which
you can download from siliconchip.
au/Shop/6/3646
Hold down the white button on the
top of the Pico while plugging it into a
USB port on your computer and copy
the firmware onto the RP2350 drive
that should appear on your computer’s file explorer or equivalent.
Once that completes successfully,
disconnect the Pico from the computer
and plug it into the PCB.
Connect the e-paper display cable
to CON3 on the underside of the main
PCB (see Table 1). The Waveshare
module comes with flying leads terminated in female DuPont connectors.
Testing
Reconnect the USB cable to the computer and the e-paper display should
show a welcome message after a few
moments. Its black border can be used
to align the display when it is mounted
on the underside of the lid (see Photo
4). If you have a USB power monitor,
the standing current should be around
35mA when powered via USB.
Once the connections have been validated, a light run of hot glue across the
DuPont connectors will convert them
into a more secure block. Ensure that
no glue gets into the metal parts of the
connectors by aligning all the locking
tabs on one side and gluing the other.
On the prototype, the heat from a
second gluing run along the latch side
Table 1: Waveshare 1.54-inch
e-paper connections
PCB
Signal
Colour
1
Vcc
Grey
2
CS
Orange
3
GND
Brown
4
SCK
Yellow
5
TX
Blue
6
DC
Green
7
RST
White
8
BSY
Purple
of the connectors caused all the plastic locking tabs to deform, requiring a
final run of hot glue along the top of
the connector to secure the pins. Try
to avoid that! You could also superglue them together.
With the unit connected to USB
power, briefly short the pins of the
start connector, CON2. The display
should flash several times and then
clear. If a cable map is displayed with
connections or bridges, check for solder bridges between the pins on the
three ICs.
Further assembly
Before mounting the tactile switch,
LED and e-paper display, lay out and
make the holes for any connectors you
want on the lid for testing cables. Refer
to the photos of my prototype for ideas.
To assist, a grid has been printed on
the underside of the lid PCB. The PCB
and display take up the bottom 65mm
of the case. Any space above that is free
for test connectors. A-side connectors
should go on the left-hand side of the
layout, to match the cable map on the
display and the 8P8C connectors.
The lid that came with the case can
be used for testing your connector
layout, as it is surplus to the project’s
needs. If your LED has a chrome bezel,
drill out the hole in the lid to 6mm.
Remove the e-paper’s mounting
screws, spacers and screen protection film. Position the display over
the square cut-out in the lid PCB with
the connector end toward the bottom
of the case. Bend four 1cm lengths of
solid hook-up wire into L shapes, then
mount the display temporarily with
a strip of electrical or masking tape
across its back (Photo 5).
The connector on the display goes
toward the bottom of the case. As
mentioned earlier, there is a border
on the splash screen to aid alignment.
When you are satisfied with its position, thread in each of the mounting
wires and solder them to both the PCB
and mounting holes on the display,
re-checking alignment after each wire
is soldered. Trim the mounting wires
close to the display.
Carefully bend the pins on the tactile switch so they are facing towards
the button – see Photo 6. If needed,
bend the ends back down so that
they are flush with the top face of the
switch. Tape the switch temporarily
Photo 4: the splash screen has a border that is useful for
aligning the display in the lid cut-out.
Photo 5: the display and start switch taped in place on the
underside of the lid, ready for soldering.
Photo 6 (inset): the pins on the tactile switch need to be bent
upward so it can be mounted on the back of the lid PCB.
38
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
in place, checking the clearance of
the central button after each pin is soldered. Solder two 15cm lengths of flexible hook-up wire to the switch pads.
Mount the LED with 15cm flying
leads connecting to the main PCB. Hot
glue will secure a standard LED. The
power switch mounts on the back of
the lid PCB using 2mm screws.
The battery box is secured to the
underside of the lid with two strips
of double-sided mounting tape. Careful positioning of the battery holder is
required, as it needs to avoid the box’s
mounting pillar at one end and the display connector on the PCB at the other.
The wires can point in either direction.
The red wire goes to the switch and
the black wire directly to the negative
terminal of CON1 on the main PCB.
Another length of hook-up wire connects the switch to the positive terminal of CON1.
The LED wires connect to CON3 on
the underside of the main PCB. Connect the two flying leads from the start
switch to CON9, also on the underside; similarly, the buzzer connects to
CON5. Polarity does not matter for the
start switch and buzzer. The buzzer is
mounted on the side of the case next
to the PCB – it can be secured with a
touch of superglue.
Once the test connectors are
mounted, wire all the pin 1s on the
A-side connectors together in a daisy
chain, leaving a 10cm tail on the closest connector to the PCB. Solder the
Table 2: Operation summary
Operation
Start
Stop
Single shot
Short press
Automatic
Continuous mode
Long press
Automatic on detection of
difference and power off
Comparison
Two long presses, short
press for each test
Long press
Clear screen
Short press with no
connections
Automatic
tail to CON8’s pin 1 and repeat for the
other pins on the A and B connectors.
Recheck for shorts between pins by
running a test cycle with no cables
connected.
Print the instruction card and trim
it on the grey border, rounding the
corners. Laminate the card and trim
with a 2mm clear border and rounded
corners. After mounting the PCB,
stick it to the bottom of the case with
double-sided tape and place the rubber bumpers in the corners (see the
photo below).
Operation
Table 2 outlines the unit’s basic
operation. Plug in a cable to be tested
– a short press on the start button will
initiate the testing cycle. One beep
indicates that a single test cycle is
being performed. When complete, the
number of connected pins and bridges
are displayed, as well as a resistance
estimate of paths with a resistance
greater than 0.2W.
Paths greater than 1W have dashed
lines. Connection resistances greater
than 15W are considered as open circuit – see Photo 7 & Screen 1.
The unit switches off at the end of
each cycle to minimise battery drain
while the cable map remains visible
on the e-paper screen. Clearing the
display is recommended at the end of
each session to prevent image ghosting. A short press with no cable connected will do the trick.
High-pitched beeps are emitted
during each test cycle. Low-pitched
beeps signal a change in the unit’s
state: powering on or off, a mode
change, the sixty-second no-activity
alarm, clearing the display, or switching off. The LED generally flashes
along with the beep.
To test for intermittent connections,
continuous mode is initiated with a
long press of the button. After two
beeps, the first scan is saved as a reference and compared with scans taken
ten times a second. The LED flashes
Photo 7: a test rig with resistors connecting between pins 1-7. The resistances
are 0.22W for pin 1, followed by 1.2W, 2.2W, 3.3W, 10W, 22W and 33W on pin 7.
siliconchip.com.au
Australia's electronics magazine
December 2026 39
Screen 1: the cable map displayed for
the Photo 7 test connections via the
8P8C sockets. The 33W path registers
as an open circuit.
Screen 2: a test with an A-side 1-2 bridge on connected pin 1, a B-side 3-4 bridge
with no cable connection, both A & B bridges on 5-6 with no cable connection
and A & B bridges on 7-8 with a connection on pin 7.
Screen 3: a long network crossover cable produces a more complex cable map.
Screen 4: in comparison and
continuous modes, only the
differences from the reference cable
map are shown on the screen.
Screen 5: the serial terminal display
for Screen 2 above.
Screen 6: the serial terminal display
for Screen 3 above.
Another long press exits comparison
mode, clears the display and switches
the unit off. If a substantial number of
cables are to be tested, consider powering the unit via the USB cable.
The remaining battery capacity is
shown in the bottom-left corner of
the display. If the battery voltage falls
below 3V, it automatically clears the
display and switches off.
After each minute of inactivity, the
unit will beep and flash the LED as
a reminder that power is still being
drawn.
Even though the circuit draws less
than one microampere after it automatically goes off, switching the unit off
using the power switch will prevent
accidental triggering.
The unit may be powered via the
Pico’s USB port, in which case it does
not switch off between runs as Q1
is not in circuit. The battery switch
should be in the off position when the
unit is USB-powered to avoid shorting
the battery to the USB supply, which
could be bad for both.
When connected to a computer via
the USB port, the unit sends the cable
map and other test information to a
terminal program such as Arduino’s
Serial Monitor or PuTTY at 19,200
baud in N81 serial format.
The format of the cable map is suitable for importing into a spreadsheet:
a header with the field labels followed
by the cable map items. “C” denotes
a connected pin, while “A” and “B”
indicate bridges on the respective connectors. Lower-case letters indicate
missing or extra connections in comparison or continuous modes.
rapidly, once for each scan. If a difference is found, the unit switches
off with the first difference displayed.
The bottom line of the display provides an indication of whether the difference is an extra or missing connection. A short press cancels continuous
testing and switches the unit off.
To allow multiple cables to be tested
against a single reference, plug in
the reference cable and use two long
presses on the button to save the reference map and cause the unit to enter
comparison mode. It will then display
the reference cable map.
Each short press then compares the
current cable with the reference, and
the first encountered difference (two
beeps) or an OK message (one beep)
is displayed. A long press clears the
display and switches off.
Comparison mode does not switch
off between runs; the 35mA idle current will drain the battery if the unit
is left switched on for long periods.
40
Silicon Chip
Australia's electronics magazine
Conclusion
Faulty cables should cause a little less pain with this cable analyser,
which can do much more than simple
go/no-go testing.
A bonus is that the constructor gets
to choose the connectors that they
most commonly use, and the 8P8C
sockets are available to create simple
and easy plug-in adaptors for less-used
SC
connectors.
siliconchip.com.au
Gifts &
GADGETS
Finish up 2026 with sleigh loads of tech gift ideas for the
whole family. Sale ends December 31st.
altronics.com.au
SAVE $80
199
$
C 5162
SAVE $50
139
$
Why pay
$300 or
more?
C 5161
SAVE 15%
75
$
X 0254 20W
SAVE 15%
SAVE $30
rn
Just wave to tu
it on and off!
55
$
99
$
Dynalink® Boom Box & Wireless PA Systems
X 0253 10W
T 1348
With magnetic
swivel mount feet
Tool
box mu
st
have!
Great sound for tunes anywhere you go this festive season with
wireless microphone for PA use. 3-8 hours use per charge.
ONLY
29.95
$
Blast away mess with the
mighty turbo jet fan!
Powerful Swivel Workshop Lights
Includes vacuum attachment - great for car
dashboards. Super powerful brushless DC
fan. USB rechargeable.
Great for the car or workshop, these handy
magnetic lights feature adjustable swivel feet and
easy USB C recharging.
RC
choppers
are back!
XY3054A Yellow
XB3054A Blue
Much loved Xmas
gift idea is back in stock for 2026!
Durable, lightweight and USB rechargeable
(3xAA requires for controller). 8-15m range.
STOCKING
STUFFER!
Two Roasts,
One App.
A 1102B
SAVE 19%
X 0231
BARGAIN!
2 for
$
30
Be cool this summer!
This handy portable personal fan
keeps you cool all summer. USB
rechargeable. 6 speeds.
24
$
SAVE 22%
27
$
D 2201
Mini Bluetooth
Audio Adaptor
Transmits or receives
audio. Up to 10 hours
portable use per charge.
Bluetooth 5.4.
Magnetic Phone
Charger
Provides 15W wireless charging
to keep your phone juiced up
on the road. Best paired with a
QC3.0 charger.
Bluetooth dual probe
meat thermometer
- track temps, set
targets, and get realtime updates right
from your phone.
SAVE $40
89
$
X 7019*
Great gift
idea!
50 Years Of Building.
Your electronics supplier since 1976.
Build It Yourself Electronics Centre®
Learn & HAVE FUN!
Makerzoid® 200
in 1 STEM Set.
Build up to 200 different
projects or create your
own! Great for inspiring
kids to invent and design
with staged learning from
the basics up to advanced
Scratch programming. Each
set contains hundreds of
blocks, plus multifunction
sensors, programmable
motor and host controller.
Compatible with big brand
name blocks. For ages 6
and up.
K 8673
SAVE $60
199
$
Makerzoid® 100 in
1 STEM Set.
SAVE $50
129
$
SAVE 20%
79
$
K 8671
Smart Robot STEM Set
This 300pc set contains the parts for 72 different
project designs which can be built using Makerzoids
3D app tutorials.
A perfect introduction into
developing kids’ coding and
self-learning skills. Through
the ability to build of up to 100
different robotic projects from
simple to complex, kids are
able to progress their STEM
skills up with staged lessons.
Each set contains hundreds
of blocks, plus multifunction
sensors, programmable motor
and host controller.
K 8672
SAVE $30
SAVE 20%
109
$
44
$
K 8670
Superbot STEM Building Set
A 400pc set allowing kids to build up to 26 different
projects, and create designs of their own.
K 1141
Build a cyborg hand!
This unique build it yourself kit fits over your own
hand like a glove and simulates full joint movements
of a human hand.
K 2206
K 2204
SAVE 20%
SAVE $30
49
79
$
$
30 In 1 Electronics Lab Kit
60 In 1 Electronics Lab Kit
Build a radio, burglar alarm, water detector, Morse
code practice circuit, simple computer circuits, etc.
Build a compass, rain detector, invisible power radio,
electro magnet. No special tools or soldering required.
K 1149
SAVE $10
40
$
Build &
rebuild it 12
ways!
12 In 1 Solar & Hydraulic Kit
A huge parts kit which can be built and rebuilt into
12 solar powered designs. Hours of fun for kids
aged 8+ (or younger with adult help).
X 3090
SAVE 24%
33
2 for
$
$
30
X 3062
The Amazing Hover Ball!
We dare you to find more fun for under
$20 this Xmas. Hugely popular every festive
season, they scoot across hard floors for your
very own family world cup! Requires 4xAA
batteries. Ages 4+
RC Drifting Motorcycle
SAVE $15
30
$
Ride like a MotoGP pro with this USB rechargeable bike,
only requires 2xAA batteries for the controller (not
included). Blasts around smooth surfaces for hours of
fun. Ages 3+.
SCAN TO FOLLOW US!
Stay up to date on latest releases,
exclusive specials and news on
our socials.
S 4815A AA 24pk
29.95
$
S 4810A AAA 24pk
Stock up on batteries for Xmas toys!
Quality Procell brand (made by Duracell) .
Like our service?
Review your store on Google.
Every review helps us serve you better.
Tech
UPGRADES.
HiFuture® Passion X
The premium home speaker you can take on the go!
Delivering amazing sound with a touch of luxury - all wrapped in a leather textured finish
with metallic trim. With a 8000mAh battery on board you can take it anywhere with you or
move it outdoors whenever you entertain. Delivers up to 100W output power with BT5.4.
Rugged
portable
design!
SAVE $70
329
$
C 5065 Gold
C 5068 Black
C 5163
HiFuture® 100W Portable Event Speaker
A beefy 100W party speaker with Bluetooth 5.4 plus instrument and
mic inputs. Great for small bands, sports clubs and more! Rugged build
quality, RGB lighting and adjustable karaoke controls.
C 9022B
+A1111
199
$
Perfect for the car or caravan! Powered off internal rechargeable
battery, your vehicle battery or mains plugpack.
A 1101A
SAVE 24%
15
$
Bluetooth Aux
Adaptor Cable
Bring high-fidelity
wireless audio to
any vehicle or device
with a 3.5mm input.
USB powered.
SAVE 20%
89
48
$
FutureTour X ANC Headphones
Dynalink® 14” Go-Anywhere Digital TV
C 9044B Black
C 9044W White
SAVE $44.95
With internal battery
- use it anywhere!
SAVE $90
299
$
Great for
fitness!
Includes bluetooth plane
adaptor valued at $34.95
S 8864A
SAVE $60
Featuring hybrid active noise cancellation
(ANC) that dynamically silences ambient
noise. Premium build and feel with playback
time up to a whopping 35 hours!
$
FutureMate Over Ear
Headphones
Latest design over ear wireless (OWS)
headphones - drivers sit on your ear rather
than in it for all day comfort.
SAVE $60
SAVE $100
179
299
$
$
A 4201
C 5205
2x50W RMS
Bluetooth Stereo Amplifier
Boomin’ 200W RMS 10” Sub!
Freestanding
Device Mounts
D 2221
SAVE 24%
SAVE 19%
D 2220
30
$
20
$
Handy all metal
desktop stands for
your phone or tablet.
Use as a second
screen while you
work.
D2221 suits devices
4.7”-13”.
D2220 max device
size 10”.
Stream audio directly from your device to
your speakers in the study or entertaining
area. 3.5mm and RCA inputs. Internal
headphone amplifier. Includes power supply,
banana speaker plugs & 3.5mm to RCA cable.
This stunning active home cinema subwoofer
adds plenty of bass to any home sound
system. Features auto power on, level
control, crossover adjustment, and phase
reversal switch. Size: 442 x 246 x 410mm.
NO STRESS 30 DAY RETURNS!
GOT A QUESTION?
Not satisfied or not suitable? No worries! Return it in
original condition within 30 days and get a refund.
Ask us! Email us any time at:
customerservice<at>altronics.com.au
Conditions apply - see website.
Gift Ideas & MORE
SAVE 20%
68
$
X 7060A
With
outdoor
sensors!
SAVE $70
229
X 7063A
$
Live and local weather monitoring.
Indoor / Outdoor Weather Station.
Track your local conditions with this sleek, 9.5” slimline weather station.
Provides real-time data via a high-precision wireless sensor (range 30m).
Tracks temperature, humidity, air pressure & moon phase. Each system can
connect up to 3 external sensors (X 7061A $19.95).
Our fully updated weather station displays your local weather data - great for
boaties & gardeners. Bright & clear base station provides readings for indoor/
outdoor temperature, humidity, air pressure, rainfall, wind speed and direction.
You can even connect it to wi-fi. 100m sensor range.
Portable Digital
Microscope
Camera
T 2125
SAVE 28%
Repair
faster!
50
$
A compact design with a
4.3” colour screen, ideal
for classrooms, hobbyists,
and PCB inspection. USB
connectivity & microSD
card slot.
SAVE 17%
82
Lightweight with USB
recharging for portability.
$
T 2128C
SAVE $30
Micron® USB Lithium Rotary Tool Set
89
$
Drills, cuts, sharpens, cleans, polishes and engraves
most surfaces, this rotary tool is ideal for enthusiasts,
hobby makers, or just odd jobs around the house. 5
speeds from 5000 to 25000RPM. USB C recharge.
X 4300B
X 7014
SAVE 17%
29
$
Compact
folding
unit!
35 in 1 Motorised Screwdriver Set
This handy set uses both 4mm and 6.3mm bits
making it suitable for a wide range of tasks. From
small items like watches & phones, to large items
like toys. USB C rechargeable.
X 0232
SAVE 19%
20
Buy one
for the
dog!
$
X 0226A
X 0187
SAVE 19%
12
$
Dual Weather Monitor
Indoor/outdoor monitor
provides temp, humidity, clock &
calendar in one handy readout.
Folding LED Lantern Light
Ideal for camping and more! Innovative
design includes a central LED lamp with
fold-out side panels. USB C rechargeable.
Mini RGB Clip On Light
USB rechargeable multi colour,
straps or clips to just about anything.
6 light modes and 12 colours.
NEW!
24.95
$
Point light where you need it!
This flexi torch features a super strong
magnetic base and built in hook. USB
rechargeable. 5 light modes.
Sale Ends December 31st 2026
Shop in-store at one of our 11 locations around Australia:
WA » PERTH » JOONDALUP » CANNINGTON » MIDLAND » MYAREE » BALCATTA
VIC » SPRINGVALE » AIRPORT WEST QLD » VIRGINIA NSW » AUBURN SA » PROSPECT
Or find a local reseller at: altronics.com.au/storelocations/dealers/
Shop online 24/7 <at> altronics.com.au
B 0012
© Altronics 2026. E&OE. Prices stated here in are only valid until date shown or until stocks run out. Prices include GST and exclude freight and insurance. See latest catalogue for freight rates. *Devices for illustration pursposes only.
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
▶
siliconchip.com.au
▶
▶
I decided to power the circuit from a
Current
Limiter
▶
▶
Design decisions
Inrush
▶
T
his Inrush Current Limiter is
designed to ‘soft start’ large power
amplifiers and power supplies,
although it would likely have application in other places, such as starting
modestly sized motors.
The principle is very simple: insert
a fixed resistance in series with the
mains input to the transformer for
the first second (or so) of operation,
then short it out with a power relay.
This allows the inrush current to be
reduced before the relay removes the
resistance from the circuit and allows
the full current required for operation.
We also provide an open-collector
output to act as a disable input for an
amplifier during power-on, minimising the current draw from the amplifier while the capacitor bank charges.
For switch-mode power supplies, it
is common to use a negative temperature coefficient (NTC) thermistor as
an inrush current limiter. These have
a resistance of a few tens of ohms initially, but as current flows through
them and they heat up, the resistance
drops to a fraction of the initial value,
to around 1W. The advantage of these
is simplicity – all you need to do is
insert the device in series.
However, for a serious power amplifier, we would need several of these in
parallel. Even so, we would need to
accept a voltage drop across them, and
at high currents, these devices run hot
– between 150°C and 200°C.
Furthermore, if you interrupt power
for a short time, they won’t have time
to cool down, and they don’t help with
that inrush current on the second and
subsequent switch-on events.
One way of avoiding the voltage
losses and dissipation of NTC thermistors used for inrush current limiting is
to put a relay across them and switch
it on after a delay, shorting the thermistor(s) out. That also has the benefit of
letting them cool down sooner.
However, once you’re doing that,
you have the option to use wire-wound
power resistors instead, bringing the
further advantage that their effectiveness no longer depends on temperature. So if you switch the unit off and
on rapidly, they will be just as effective each time, although they will get
hotter than they would from a single
switch-on event. So they need to have
a high power rating.
By Phil Prosser
While testing a
power supply with a 600VA toroidal
transformer and a substantial capacitor bank, I frequently
tripped the circuit breaker on my switchboard. This simple
circuit limits the inrush current for this type of power supply,
saving the hassle and making power-up much less eventful!
▶ Limits mains current draw to about 7A when an appliance is switched on
▶ Rapid detection of mains interruption, ensuring reliable soft starting
▶ The control circuitry is isolated from the mains
▶ Programmable 1-10s delay with a resistor (default is 1s)
▶ Open-collector output for disabling speakers or other loads while starting up
▶ Not affected by rapid on/off cycling
▶ Maximum load current: 8A
Australia's electronics magazine
December 2026 45
standalone transformer on the board. It
may seem extravagant to add an extra
transformer when we are likely limiting the inrush of a large transformer,
but we do not want to be subject to
the regulation of the large transformer.
If there is a heavy load on that transformer, there is the potential for its output to droop, and we don’t want that
affecting the operation of the Inrush
Current Limiter.
Furthermore, our onboard power
supply has been sized so that if the
mains input drops significantly below
220V AC, it will continue to operate
under both start up and mains fault
conditions; including brownouts.
Given that the source impedance of
a mains socket is not zero, the appliance will have a cord (possibly a long
one), and may even be plugged in via
an extension cord, when drawing tens
of hundreds of amps, it is not unlikely
for the voltage seen at the socket to
be substantially lower than nominal.
If the mains input is interrupted
briefly, an NTC inrush limiter will not
work. Our circuit detects a brief loss
of mains, switches the limiter in and
drives the open collector disable active.
This adds a couple of parts to the circuit
but it is an important feature.
This circuit has parts connected
directly to the Active mains conductor. To ensure maximum safety, the
transformer isolates all but the relay
and 10W resistors from the mains.
Because it’s floating, we can safely connect the ground of the Inrush Current
Limiter to an external circuit, allowing
Why large amplifiers draw so much current at switch-on
There are a couple of reasons why large linear power supplies, like those used
in high-power Class-A or Class-AB amplifiers, draw such a high current initially.
1. Charging the capacitor bank. When the reservoir capacitors are fully discharged, they initially present a very low impedance to the rectifier. The charging
current is limited mainly by the transformer winding resistance and leakage
inductance, rectifier and wiring resistance and the ESR of the capacitors. In a
large amplifier, the initial current can easily reach tens of amps.
2. Transformer magnetising inrush. Toroidal transformers, which are commonly used in high-power amplifiers, can draw an exceptionally large magnetising current when first energised. Transformer core flux depends on the integral
of the applied voltage, so switching on near a mains-voltage zero crossing can
produce a flux excursion much greater than occurs during normal operation.
Any remanent magnetisation left in the core can make this worse, potentially
driving the core into saturation. The resulting current can briefly reach hundreds of amps in a large transformer.
Switch-mode power supplies can also have high inrush currents. Small EMI
suppression capacitors connected across the mains can produce a very brief
current spike as they charge, although this is generally not the main source
of inrush.
More importantly, most mains-powered switch-mode supplies rectify the
mains to produce high-voltage DC, with a substantial electrolytic capacitor
connected to the DC bus. When the supply is first switched on, this capacitor
is discharged and can draw many tens of amps through the bridge rectifier
unless some form of inrush-current limiting or controlled pre-charge is provided.
the open-collector limit active output
to be used.
The delay required to reduce inrush
current is quite short. Usually, a handful of cycles is all that is required, so
a few hundred milliseconds. The recommended configuration implements
a one-second delay.
The delay is controlled by a single
resistor value, which is 10kW for the
default delay of one second. Changing
this to 2.2kW will make it more like
200ms, while increasing it will make
the delay much longer, up to 10 seconds with 100kW. However, if increasing the delay, you need to consider
the thermal limitations of the current-
limiting resistors, discussed later.
The open-collector output uses a
BC548 NPN transistor and can sink
up to 20mA reliably.
The maximum load current is
defined by the PCB; the PCB traces are
7mm wide, so they can safely handle
8A continuously. The relay itself is
rated at 16A; you have to derate that
for inductive loads like transformers, but since it’s switching when the
voltage across the 10W power resistors is already quite low, it isn’t being
worked hard.
Circuit description
Scope 1: the yellow trace is the unregulated DC supply voltage across the 470μF
capacitor, while the cyan trace is the negative end of the relay coil. The relay is
energised close to one second after power is applied.
46
Silicon Chip
Australia's electronics magazine
The circuit of the Inrush Current
Limiter is shown in Fig.1.
At power on, RLY1 is off, so mains
current and voltages comes in via the
ACTIVE IN terminal of CON2, through
the three 100W resistors in parallel,
then to the ACTIVE OUT terminal to
power the main toroidal transformer
and whatever else runs from mains
in the amplifier. Only the Active current flows through this board; Neutral
and Earth will be wired to the power
socket externally.
The Neutral connection on CON2
siliconchip.com.au
Fig.1: most of the circuit is involved in controlling RLY1 so that it switches on
one second after power is applied, shorting out the resistors that the Active
current otherwise has to flow through, between pins 1 & 2 of CON2. It’s also
designed to switch RLY1 off quickly after mains power is lost.
is to power this board, and there is
no Earth terminal, as the circuitry is
floating.
The Inrush Current Limiter circuit
is powered from the mains input via
a 3VA transformer, bridge rectifier and
470μF filter capacitor, producing a
roughly 17V unregulated DC supply.
That is dropped to a regulated 12V
DC by REG1, which powers the rest
of the circuit.
The rest of the circuit is essentially
a time delay so that NPN transistor
Q5 switches on roughly one second
after mains power is applied, energising RLY1 and shorting out the 100W
resistors. When that happens, mains
current to the appliance flows in via
ACTIVE IN, through the relay contact(s) and out through ACTIVE OUT,
bypassing the 100W resistors.
When mains power is switched off,
the circuit is designed to switch RLY1
off immediately so that the 100W resistors are back in-circuit in case power is
switched on immediately again.
Timer
Initially at power-on, the 100μF
capacitor will be discharged and thus
diode D7 will be reverse-biased. Q3’s
base is pulled up to +12V by the 100kW
resistor, so Q3 is off and so are Q5 and
Q6. The relay is not energised. The
100μF capacitor starts charging via
siliconchip.com.au
its series 10kW resistor (or whatever
value you’ve chosen).
The 100μF capacitor has to charge
to about 8.2V before anything happens. At that point, there’s enough
voltage to forward-bias diode D7 and
transistor Q3’s base-emitter junction.
This happens at 8.2V due to the voltage across 6.8V zener diode ZD1 plus
D7’s forward voltage (~0.7V) and Q3’s
base-emitter junction forward bias
voltage (~0.7V).
At that stage, Q3, Q5 and Q6 switch
on, the relay is energised and the
inrush current limiting resistors are
shorted out by the relay – see Scope 1.
Note that during this process, transistor Q1 is switched off and so it has
no effect on the charging of the 100μF
capacitor. We can verify this by looking
at the rest of the circuit. When there is
an AC output voltage from transformer
T1, periodically NPN transistor Q4
will switch on.
When it’s on, it discharges the 100nF
capacitor at its collector, so that capacitor can never charge sufficiently to
switch off PNP transistor Q2 due to the
time constant with the 100kW resistor
(100kW × 100nF = 10ms, but it needs
several time constants to charge sufficiently).
Thus, Q2 is switched on the whole
time that mains is applied to the circuit, and it pulls up the base of PNP
Australia's electronics magazine
transistor Q1, preventing it from
switching on.
About 75ms after mains power
is removed, with Q4 off, the 100nF
capacitor at the base of Q2 charges
to about 11.5V, switching Q2 off and
Q3 on and rapidly discharging the
100μF capacitor. Q3, Q5 & Q6 then all
immediately switch off and the relay
is de-energised, with D5 clamping its
back-EMF voltage spike.
Timing
The time constant for the 100μF
capacitor charging and thus the relay
switching is one second for R = 10kW
and C = 100μF.
Since this is linearly proportional to
R, all you really need to know is that
doubling the value of R doubles the
time delay, and halving it will halve
the time delay. Suitable values range
from about 2.2kW (<250ms) to 100kW
(5-10s).
Transformer options
For transformer T1, we’ve chosen a
small PCB-mounting 3VA transformer
from Altronics. Using this results in
minimal mains wiring, so the result
is neat and tidy. We’ve slotted the secondary holes so a standard 25mm-pitch
PCB-mounting transformer can be used
instead, such as Myrra’s 44199, available from element14.
December 2026 47
Parts List – Inrush Current Limiter
1 double-sided 110.5 × 88.5mm PCB coded 10110261
1 2-way polarised header, 2.54mm pitch (CON1; optional)
1 3-way 5/5.08mm mains-rated terminal block (CON2) [Altronics P2039]
1 PCB-mounting safety fuse holder (F1)
[Altronics S5987, Mouser 693-0031.3553]
1 200mA fast-blow M205 fuse (F1)
1 255VAC 16A SPST PCB-mount relay, 12V DC coil (RLY1)
[Altronics SA4198 or Hongfa HF14FW/012-HST(136)]
1 3VA 6+6V output mains transformer (T1) [Altronics M7012A, Myrra 44199]
1 micro-U TO-220 heatsink and mounting hardware (optional; see text)
[Altronics H0630]
1 100mm length of 1.6mm diameter heatshrink tubing
1 21 × 25mm piece of insulating material (eg, acrylic or Presspahn)
2 12mm nylon tapped spacers
4 M3 × 6mm plastic/nylon machine screws
M3 mounting hardware to suit application
Semiconductors
1 7812 12V 1A linear regulator, TO-220 (REG1)
3 BC558 30V 100mA PNP transistors, TO-92 (Q1-Q3)
3 BC548 30V 100mA NPN transistors, TO-92 (Q4-Q6)
1 6.8V 400mW/500mW axial zener diode, DO-35/DO-41 (ZD1)
4 1N4004 400V 1A power diodes, DO-41 (D1-D4)
5 1N4148 75V 150mA signal diodes, DO-35 (D5, D7-D10)
Capacitors
1 470μF 25V/35V electrolytic capacitor
1 100μF 25V electrolytic capacitor
3 100nF 50V/63V/100V ceramic or MKT capacitors
Resistors (all ¼W ±1% axial unless noted)
2 100kW
8 10kW (7 if using a timing resistor value other than 10kW)
3 100W 10W ±5% wirewound [Altronics R0435]
Fig.2: assembly is straightforward and all parts are through-hole types. Start
with the lower-profile components and work your way up to the taller ones,
watching the orientations of the diodes and transistors. Change the resistor
labelled in red for a different timeout (higher value = longer, lower = shorter).
48
Silicon Chip
Australia's electronics magazine
If you need to use a transformer
with flying leads, such as Altronics’
M2851A, we have included holes in
the PCB that allow you to use cable ties
to secure the wires to the PCB.
We are also using a PCB-mounting
safety fuse holder, with a 200mA fuse
for the 3VA transformer. There is room
for a small flag heatsink on the regulator, but this would only be required
if your transformer outputs more than
12V AC.
Construction
The Inrush Current Limiter is built
on a 110.5 × 88.5mm PCB that’s coded
10110261. Use the overlay diagram,
Fig.2, as a guide during construction.
Start by fitting all the small resistors. We are recommending a 10kW
timing resistor (value shown in red).
This gives a one-second soft start,
which worked well for our large test
amplifier. If you are not using the
open-collector output to disable your
speakers or load, a delay of 200ms (use
2.2kW) would be worth considering.
Follow by fitting the diodes. Do
not mix up the sole zener diode, the
small 1N4148 signal diodes and larger
1N4004 parts. Make sure the cathode
stripes all face as shown in Fig.2.
The three-way terminal block needs
to be mains-rated; Altronics Cat P2039
is suitably rated for this application.
Mount it now, plus CON1 if you will
be using it.
Now install all the transistors. There
are three each of the NPN (BC548)
and PNP (BC558) devices, so don’t
mix them up. Other transistors from
this series, like the BC547/BC557 and
BC549/BC559, could also be used.
Make sure the flat faces are orientated as shown. Follow these with
the voltage regulator, capacitors and
fuse holder.
The heatsink for REG1 isn’t required
if you’re using the recommended
transformer, but it wouldn’t hurt if
you fit it.
Finally, load the relay, the 10W
resistors and the transformer. Add
short lengths of heatshrink tubing to
these resistor leads before you solder
them so that once the board is running,
there is no possibility of you touching
these legs.
Testing
Connect a DC power supply set to
15V DC or more, with its negative output connected to the GND test point
siliconchip.com.au
Figs.3 & 4: a suggestion for how to wire the unit up. It’s pretty simple. We haven’t shown anything
connected to CON1 here; many applications won’t require it. Cut the shield shown here from a small piece
of insulating material like acrylic or Presspahn and attach it with plastic tapped spacers and plastic machine screws so
you can’t accidentally come in contact with the live screws on the top of terminal block CON2.
in the lower left of the board and the
positive to the “+UNREG” test point,
which is next to the 470μF capacitor.
Apply power and check the voltage
on the 12V test point in the top-right
corner of the board. This should be
12V ±0.5V. The relay will not click in
as there is no AC present.
If it doesn’t work:
• Is the 12V rail OK? If not, is the
DC input OK?
• If the output is low, is the regulator hot? If so, something is likely
shorted to ground.
• Is the 470μF capacitor the right
way around?
• Are the transistors and diodes in
the right way around?
Tack a piece of hookup wire to TP1,
which is just near the top left of the
transformer. With the DC power supply on, touch this to the +UNREG test
point. This should enable the AC presence detection system and the relay
should click in after about a second. If
this works, everything is good.
Once the board operates from the
DC supply as above, you are ready to
install it. Insert a 200mA M205 fuse
into the fuse holder. Wire the Inrush
Limiter into your circuit as shown in
Fig.3, ensuring that all the mains wiring is secured using cable ties.
Cut the shield shown in Fig.4 from
a sheet of Presspahn, acrylic or similar insulating material, drill two 3mm
siliconchip.com.au
holes as shown and mount it on nylon
tapped spacers using plastic/nylon
screws via the two holes in the PCB.
This covers the terminal block so you
can’t come into accidental contact
with the live screws on the top.
Apply power and check that the
relay clicks in after about a second.
If it does not, power it down, wait for
all capacitors to discharge and investigate.
Usage
quickly as we could, and repeating.
These twelve cycles took about 90
seconds. At the end, after discharging
and unplugging everything, the resistors were at about 50°C.
This is an unrealistic test, as if your
speakers are disabled at power off, the
main rails will not collapse in the few
seconds we took to discharge them
with a 4W resistor. So this suggests the
resistors will not live a stressful life
with normal usage. SC
We tested this with a 600VA
transformer loaded with
35,000μF capacitors on both
rails. Without the Inrush Limiter, the power board’s 20A
circuit breaker would kick
out with alarming regularity. With the Inrush Limiter in circuit, the start-up
was much more sedate; no
thumps from the transformer and everything was
controlled.
One concern is how
much dissipation occurs
in the 10W resistors.
We ran twelve cycles of
turning the circuit on,
then off, discharging
the main capacitors as
The assembled board.
The heatsink is optional.
Australia's electronics magazine
December 2026 49
Feature by Andrew Levido
Motor Control
Part 3: Stepping/Stepper Motors
This month, our topic is stepping (or stepper) motors. Unlike the DC & synchronous
motors we covered in the last few articles, the stepping motor rotates in a series
of discrete steps rather than continuously. This makes it very useful for applications
that require fairly precise position control without the need for feedback.
T
hese types of motors are found in many
low-cost positioning devices like 3D
or 2D printers, where the expense
of full-feedback servomotors is not
justified.
All the motors we have looked at so
far produce torque through the BIl relationship between flux density, current
and winding length. Stepper motors
are no different, although the mechanism is a bit more obscure.
Stepping motors generate torque
through a mechanism called variable
reluctance. You may recall that reluctance is the ‘resistance’ of a magnetic
circuit to magnetic flux.
The current in an electrical circuit tends to take the path of least
resistance because this is the lowest
energy state – it takes more voltage,
and therefore energy, to push a given
current through a higher resistance
than a lower one.
It is the same for magnetic flux –
it takes more magnetomotive force
(mmf), and thus more energy, to push
a flux through a higher reluctance path
than a lower one. Thus, flux tends to
seek the low-reluctance path.
Variable reluctance
This phenomenon can create a
torque in a variable reluctance motor,
like a stepping motor, through the
mechanism illustrated in Fig.1. In
this simplified example, a salient pole
stator carries three pairs of windings,
each set 120° apart. Opposite windings are wired such that their magnetic fields are orientated in the same
direction.
I have shown the top and bottom
poles energised in the direction that
makes the magnetic field point upward
through the rotor. The flux completes
its loop through the stator body, but I
have not bothered to show this to keep
the diagram uncluttered.
The rotor is made of laminated
transformer steel and also has salient
poles, but it has no windings – the
‘doubly salient’ characteristic is critical for torque production these motors.
If the rotor poles are fully aligned
Fig.1: stepping motors produce torque using variable reluctance. The torque
tends to push the rotor back into alignment with the stator, where the reluctance
is lowest.
50
Silicon Chip
Australia's electronics magazine
with the stator poles, as shown on the
left, the low-reluctance path ensures
the magnetic flux produced by the
stator is evenly distributed across the
poles and the flux densities on the
winding on the top left in which current comes out of the page and the
one on the top right where it goes in
are equal.
There are still BIl forces on each of
the energised windings (blue arrows),
but they are equal and opposite, so
they cancel out and the net reactive
torque on the rotor is zero.
If the rotor is turned by a slight angle,
as shown in the diagram on the right,
the area of the air gap between the
rotor and stator saliencies decreases.
The field lines bunch together because
they still want to take the lowest reluctance path, so the flux density in this
region increases.
The flux is no longer evenly distributed across the energised stator poles.
At the right-hand side of the top stator, the flux density is higher, so the
winding on the right sees a higher BIl
force than the one on the left. The same
thing happens on the other side at the
bottom pole.
The BIl forces on the windings are
therefore unbalanced, indicated by the
length of the blue vectors, resulting in a
net torque on the rotor in the direction
that tries to pull it back into alignment
with the stator. The opposite would
happen if we turned the rotor counterclockwise; the rotor always wants to
pull back to the neutral position where
the reluctance is minimum.
The same thing would happen if the
current in the windings were reversed.
The magnetic field through the rotor
would point down, but the net force
will always act to pull the stator back
siliconchip.com.au
Fig.2: when the current commutates from the U-phase to the V-phase, the rotor shifts by 30° to align with the V-phase pole
pair. Different pole configurations produce different step angles.
to the minimum reluctance, fully
aligned position.
If we now energise successive sets of
poles in sequence, as shown in Fig.2,
the rotor is pulled around so that the
nearest rotor poles align with the
newly energised stator poles. In this
motor with three stator pole pairs and
four rotor poles, the step angle is 30°.
The step angle (θ) is related to the
number of rotor poles (Pr) and the
number of stator pole pairs (Ns) by the
expression θ = 2π ÷ (Pr Ns) for radians and θ = 360 ÷ (Pr Ns) for degrees.
In the example in Fig.2, the phases
are energised in a U-V-W sequence
counter clockwise around the rotor,
and the stable rotor angle rotates clockwise. You can probably see that if the
phase sequence was U-W-V, the rotor
would move the other way.
A variable reluctance motor therefore rotates in a series of discrete angular steps each time the stator current
commutates from one winding to the
next. Between commutations, the rotor
is ‘locked’ at a stable angle and will
resist attempts to move it. This precise
rotor positioning and ‘holding torque’
is what makes these motors handy for
positioning applications.
Unfortunately, there is no neat
electrical model or simple equation
for torque vs angle or speed for these
motors because everything depends
on the magnetic design of the motor.
However, we can look at the general
shape of the torque-angle characteristic and see what we can learn from it.
Fig.3(a) shows three generalised
torque-angle curves, each in a different
colour according to which winding is
energised. Only one of these curves is
valid at any given time. The solid dots
represent the stable low-
reluctance
rotor angle where there is zero torque.
These dots will be 30° apart for the
motor in Fig.2, but could be at different angles for motors with different
stator and rotor pole configurations.
Imagine the blue phase is energised.
With no shaft load, the rotor is stable
at the angle indicated by the solid blue
dot. If some load torque is applied,
the rotor angle will move up or down
the curve to the point that the motor
and load torques are balanced. One
such point is indicated by the hollow
blue circle.
The angle of the rotor does change a
bit with applied torque, but the characteristic is fairly steep in real motors,
so the angle does not change a huge
amount.
If the current commutates from the
blue phase to the green phase, the
operating point shifts up to the green
curve along the vertical black line.
The torque initially increases significantly as the operating point rides up
the green curve, accelerating the rotor
toward the next pole.
Once the rotor angle passes the
green curve’s peak, the torque reduces
until the rotor angle reaches the next
stable angle indicated by the hollow
green circle. The rotor has skipped
ahead one step.
This will obviously only work if
the initial blue-phase operating point
is at or below the intersection of the
blue and green curves shown by the
horizontal dotted line. If the operating
point were above this ‘pull-out torque’
(Tpo), the green phase would not be
able to ‘catch’ the rotor and the step
would be missed.
Depending on the circumstances,
the rotor may or may not resynchronise, but if you are relying on the
stepper to arrive at a precise angle,
you will be out of luck. This ‘missed
step’ problem is probably the most
Fig.3: these charts together show the torque-speed and torque-angle characteristics of a typical stepping motor. There is no
simple equation or model to describe their behaviour. ω is proportional to the drive pulse frequency.
siliconchip.com.au
Australia's electronics magazine
December 2026 51
Fig.4: a hybrid stepping motor includes a permanent magnet in the rotor. This example has 50 rotor poles and four stator
pole-pairs for a step angle of 1.8°.
common difficulty people encounter
when using stepper motors for the
first time. It is made more difficult by
the fact that the pull-out torque is not
fixed for a given motor.
When the rotor transitions from
one stable position to another, it does
not do so linearly, or even in perfect
steps. The rotor and load’s moment
of inertia will take it a little past the
next equilibrium point where the net
torque will reverse, pulling it back. It
may take a few oscillations until the
rotor settles at its steady-state position,
as shown in Fig.3(b).
It takes a finite time to settle, and
the faster you move it, the more pronounced the overshoot will be. This
imposes some limitations on the stepping rate and therefore the rotor speed.
If you try to step too fast, the rotor does
not have time to ‘lock on’ to the stable
angle before it has to move to the next,
and a step may be missed, or synchronisation may be lost altogether.
The result of all this is that the pullout torque reduces as the step rate
increases, as shown in the blue curve
in Fig.3(c). Some version of this curve
is usually provided in the motor’s data
sheet. The horizontal axis is step rate,
sometimes called pulse frequency for
reasons we will get to in a moment,
or shaft speed (in RPM or radians per
second).
The pull-out torque drops off dramatically with step rate until it reaches
some maximum response frequency,
where even the unloaded rotor cannot keep up.
There is also a pull-in torque limit,
as shown by the red line on the graph,
although you will rarely find this mentioned in the data sheets. This is the
maximum load torque that the motor
can accelerate from stationary for a
given step rate.
It is not really relevant for most
applications, because you would normally ramp the step rate up from a low
value rather than banging it on full
bore on a stationary motor and load.
Hybrid stepping motor
That’s the theory out of the way. If
we want to get practical, we need to
Photo 1 (left): you
can clearly see the
offset between the
segments in this photo
of the rotor of a NEMA 17 sized 1.8°
stepper. This offset is critical for the
operation of hybrid stepper motors.
Photo 2 (right): this is the stator from
the same motor as above. The four
pole-pairs, each with six saliences,
are clearly visible.
52
Silicon Chip
Australia's electronics magazine
talk about hybrid stepping motors,
which are the types you will most
commonly encounter in the wild. The
ubiquitous “NEMA” type motors with
the square-ish body are hybrid types.
The most common variant has 200
steps per revolution (a 1.8° step angle)
and comes in a range of standardised
frame sizes.
The NEMA size designation is the
approximate width and height of the
motor frame in tenths of an inch. For
example, a NEMA 17 motor is roughly
1.7 inches (43mm) square. For more
on NEMA stepper motors, see our article titled “A quick primer on stepper
motors” (January 2019; siliconchip.au/
Article/11370).
To achieve a 1.8° step angle requires
lots of poles: 50 rotor poles and four
stator pole-pairs is a common combination. Such a large number of necessarily small rotor poles makes it difficult to get adequate flux density using
the stator windings alone. These stepper motors therefore have permanent
magnets in the rotors to provide some
additional flux – hence the “hybrid”
designation.
The elevation in Fig.4 shows a side
view of the rotor. The magnets are
positioned axially in the rotor, which
is divided into two sections. Each section has 50 poles, which look a bit like
the teeth of a gear. Critically, these sections are fixed to the rotor such that
the ‘teeth’ on one section are aligned
with the ‘slots’ on the other.
Photo 1 is a photo of a rotor and its
bearings from a NEMA 17 motor. The
rotor slots and the gap between the
rotor sections are filled with epoxy,
but you can clearly see the offset.
The rotor in this photo has two
siliconchip.com.au
sections. You can stack pairs of sections end-to-end on a common shaft
to achieve higher torque, but you need
a correspondingly longer stator. This
is why NEMA-style stepping motors
come in different lengths for each
frame size.
The stator of this motor (Photo 2) has
four pole pairs arranged in two phases.
Each pole has several saliences on the
same pitch as the rotor poles. You can
see how the stator and rotor interact if
you look at the cross-section diagram
in Fig.4. This cross-section is through
the stator and the north pole section of
the rotor. We will start by considering
what is going on when no windings
are energised.
The flux produced by the permanent magnet crosses the air-gap from
the north rotor section to the stator
poles, flows axially along the stator
and back across the air gap to the
south-pole rotor section. Crucially, the
offset between rotor sections ensures
the reluctance of the flux path through
each stator pole is about equal.
If the reluctance of each path were
not roughly equal, the permanent
magnet’s field would produce a strong
static torque that the torque produced
by the windings would have to overcome. If you have ever turned the shaft
of an unenergised hybrid stepping
motor, you know that you can do so
easily while feeling just a very light
‘detent’ torque.
We can understand how this works
by mentally following the flux path
through each stator pole-pair. In the
case of pole pair 1, the flux passes
from the north rotor segment through
the fully aligned poles on the north
end of the motor and back via the
fully unaligned pole pair to the south
rotor segment.
The same sequence happens for
pole-pair 3, except the unaligned segment is at the north end and the aligned
segment is at the south. The flux paths
through pole pairs 2 and 4 are through
two sets of half-aligned air gaps.
By sizing the teeth and slots just
right, the designers make the reluctance of all the paths almost identical. This clever design means the flux
produced by the permanent magnets is
available to assist the windings in creating torque when they are energised,
but does not produce much torque on
its own.
When the green winding is energised, the flux it produces adds to the
siliconchip.com.au
permanent magnet flux to create a
holding torque that locks the rotor in
the position shown in the cross-section
in Fig.4. When the current commutates
from the green winding to the blue
one, the rotor advances clockwise by
the 1.8° necessary to align pole-pair 2,
as shown in Detail A.
The directions shown are for positive current. If we had energised the
blue winding with a negative current, the rotor would have moved
1.8° anti-clockwise. To rotate clockwise, the phases would be energised
in the sequence +G, +B, –G, –B, and
to rotate anti-clockwise the sequence
is +G, –B, –G, +B.
In reality, we don’t usually energise
single windings in a hybrid stepper
motor unless we are microstepping
(more on that below), but it is easier
to understand the motor’s operation if
we make this simplification for now.
Driving stepper motors
The important thing to note is that
to use a permanent-magnet or hybrid
stepper motor, we have to be able to
drive current through the windings
in either direction. How we do this
depends on the configuration of the
motor windings.
A bipolar or 4-wire stepping motor
like that shown in Fig.5 has a single
pair of wires for each phase and therefore each requires full-bridge drivers so current can be pushed through
the windings in either direction. In
a unipolar or 6-wire stepping motor,
each winding is centre-tapped such
that current flowing from the common
wire to each end produces flux in the
opposite direction.
This means the power electronics
to drive 6-wire motors is simpler, but
it comes at the expense of reduced
torque for a given motor size, since at
any time only half of each winding is
carrying current.
You can operate a 6-wire motor as a
bipolar one by ignoring the common
terminal of each winding and driving
the ends of the winding with a fullbridge. That is why most off-the-shelf
stepper motor drivers use the fullbridge configuration.
Unlike most motors, the current
through a stepping motor is not
directly proportional to torque. These
motors, especially small ones, draw
about the same current whether stationary or moving, loaded or unloaded.
The only thing limiting the current
is the DC resistance of the windings,
so you almost always need to use a
current-mode controller to drive them.
These motors usually have both a
voltage rating and a current rating,
but it is the latter you really need to
pay attention to. You will burn out
the motor if you exceed this for any
length of time.
Stepping and microstepping
As mentioned above, we don’t usually drive a hybrid stepping motor
Fig.5: 4-wire or bipolar stepping
motors have one pair of terminals for
each phase and must be driven by
a full-bridge. Six-wire or ‘unipolar’
steppers have a centre-tapped
winding which simplifies the drive
requirements at the expense of lower
torque.
Australia's electronics magazine
December 2026 53
Fig.6: microstepping creates intermediate steps for increased angular resolution
and lower motor noise. The torque available at the incremental positions is
lower than that at the full-step angles, though.
using just one of the two phases. We
get maximum torque if we energise
both windings at the same time, which
shifts the stable rotor position halfway
between the stator poles. You can think
of this as adjacent pole pairs (say 1 &
2 in Fig.4) forming a single larger pole
if that helps.
You can see this in action in the simplified representation of a two-phase
stepper in Fig.6 (at upper left). This
motor has two rotor poles and two
stator pole-pairs, so nominally each
step is 90°. During each full step, the
red and blue phases windings are both
energised, and the stable rotor angles
fall between the stator poles.
Note that this is a stylised drawing
– a motor that actually looked like this
would not work!
In each stable full-step position,
both windings are energised with
100% of the nominal current, albeit
in different permutations of polarity,
54
Silicon Chip
and the torque is maximised. The
rotor will jump very positively from
angle to angle with a lot of overshoot
and ripple.
We have already seen that if we
energise a single winding, the rotor
aligns directly with that pole-pair.
We can use this to create intermediate
steps between each full step. This is
called half-stepping and is illustrated
at upper right in Fig.6. There are now
eight rotor positions instead of four,
and the step-angle is halved.
Still, there is a catch. In positions 2,
4, 6 & 8 where only one winding is energised, the holding torque is about half
that at the full-step angle. This lower
intermediate position torque is often
referred to as the “incremental torque”.
The advantage of half-stepping is
that the noise and vibration of the
motor are reduced, as is the magnitude of the angle overshoot each step.
You also get twice the angular resolution (so 0.9° in a 1.8° stepper) but half
the pull-out torque in the intermediate steps. There is no such thing as a
free lunch!
Because we (probably) have a
current-mode controller, we are not
limited to the three current levels we
have used so far (100% positive, zero
and 100% negative), so we could create more sub-steps by partially energising coils.
For example, we could create a stable position between steps 1 and 2 of
the half-step drive by partially energising the blue winding and fully energising the red winding. In this way, we
Australia's electronics magazine
could quadruple the number of steps
per revolution from four for full stepping to 16 for quarter stepping. The
first four steps of the 16-step sequence
are shown at lower left in Fig.6.
You can actually create any arbitrary
number of sub-steps, but powers of two
are the most common. Collectively,
these sub-stepping strategies are called
microstepping. Typical stepping motor
drivers capable of microstepping top
out at 32 sub-steps per full step. The
incremental torque is 1/8 of the full-step
torque, so this level of micro-stepping
is not all that useful for positioning
unless the load is very light.
The main use case for very small
micro-steps is to make the rotor movement smoother and quieter. This can
be useful since stepper motors tend to
operate at audio frequencies and they
can be loud. If you want proof, check
out the video clips online of people
using stepper motors to play music!
Getting practical
A good way to explore stepping
motors is to get hold of one and a lowcost driver module, then try them out.
The driver I bought (Photo 3) is a typical example. It is suitable for fourwire motors, can source and sink up
to 4.0A and operates from a 9-42V
DC supply. It is equipped with DIP
switches to select the motor current
and the step type.
It has opto-isolated logic inputs for
enable, direction and the step pulse. To
run the motor, you leave enable open,
so it is really a disable input.
siliconchip.com.au
Photo 3: a typical off-theshelf stepping motor
controller. This unit
is rated for 4.0A
and can be set
up for microstepping at
various
rates.
You can
also leave
the direction
input open if you only
need single-
direction rotation. The motor will take a single step
on every rising edge of the step input.
This direction/step pulse interface
is very common with stepper motor
drivers, and is the reason that a lot of
the stepping motor data sheets use the
term ‘pulse rate’ rather than step rate.
It is pretty easy to generate a suitable pulse train with a microcontroller using a timer, but you can use
more or less any source that can generate a 3.3V or 5V pulse train. I used
my signal generator with the output
amplitude and offset adjusted appropriately.
I hooked the driver up to a NEMA 17
350mA stepping motor. Scope 1 shows
the stator current waveforms at 500mA
per division for full-stepping at 100
pulses per second. The high-frequency
ripple on the current waveform is the
current-mode regulation in action.
I had the current limit set to 500mA
(the lowest setting), but with a 12V
supply, the motor current was roughly
350mA as expected. This was with no
load on the shaft, reinforcing the point
that variable-reluctance motor current
is not proportional to load as it is with
most other motor types.
Under these conditions, the shaft
was making very distinct steps and
was impossible to stall with my fingers.
Scope 2 shows the same waveforms for half-stepping mode. You
can clearly see the ‘shoulders’ on the
waveform where only one winding
is energised. The shaft rotation was
noticeably smoother, but I could make
the shaft skip intermediate steps with
my fingers.
Scope 3 is a close-up of the current waveform shape with 16× micro-
stepping enabled. The steps themselves are barely visible in the current
switching hash and the transitions
siliconchip.com.au
Scope 1: the
current waveforms
(500mA per
division) for a
NEMA 17 stepper
operating in
full-step mode
at 100 steps per
second. The hash
on the traces
is the currentmode controller
switching.
Scope 2: with
half-stepping, the
periods where
only one winding
is energised are
very apparent.
The torque
available during
these steps will
be half that at the
full-steps.
Scope 3: this
‘scope grab
shows the current
waveforms
with 16× microstepping. You can
barely make out
the steps. The
motor operation
was very quiet and
smooth, but the
impact on torque
was marked.
look more-or-less linear. The acoustic
noise made by the motor was much
lower with 16 micro-steps, but the very
low incremental torque between major
steps was very apparent, and it was
easy to stall the shaft with my fingers.
Conclusion
Stepping motors, particularly the
low-cost NEMA-style hybrid steppers,
are very useful tools if you are looking
for low-cost position control.
They are easy to drive using cheap
off-the-shelf modules or one of the
many stepping motor control chips
available from the usual suspects.
They play very nicely with microcontrollers, so they are a good choice for
many do-it-yourself projects.
Australia's electronics magazine
Their strengths are holding torque
and positioning, as long as you don’t
need to move too fast. They are not a
replacement for DC motors, whether
electrically or mechanically commutated, if you just need to turn a shaft.
They don’t run fast (even at moderate speeds they have terrible torque)
and they are not very efficient. If you
want it all – accurate positioning, high
speed and high torque, you probably
need a full-blown servomotor.
Nevertheless, if you can live with
the limitations I have mentioned (and
often you can), hybrid stepping motors
are a cost-effective option.
Next month we will take a look at
three-phase induction motors, focusSC
ing on field-orientated control.
December 2026 55
W
e described how this new
clock works in the main
article last month, including the assembly instructions. What we didn’t have space to
explain properly is the user interface,
which is accessed via a web page on
the local network. We also didn’t
discuss how the somewhat complex
firmware works. This article will go
over both.
#1 Status display
At the top of the web page, we show
various status fields so you can see
what the Clock is doing:
A. Time – this is the same time
shown on the clock face during normal operation
B. Time source:
↳ “NTP” means the clock is keeping time using its internal oscillator,
synchronised hourly with the selected
NTP server. It should remain accurate
within less than a second. The time
is considered valid as long as it was
successfully synchronised within the
last four hours.
↳ “GNSS” means it’s receiving valid
UTC time from the GNSS module and
converting that to local time using the
time zone rules. This is considered
valid as long as valid time data was
received from the module within the
last three seconds.
↳ “Internal” is the same as NTP
except that it will no longer try to keep
it synchronised. This can be used if
NTP and GNSS are not available, but
you’ll have to correct it manually if it
drifts over time.
C. Last NTP sync – the time on the
clock face the last time NTP synchronisation was successful.
D. GNSS status:
↳ “No data” means the Pico W isn’t
receiving valid NMEA serial data from
the module.
↳ “Receiving, no fix, N satellites”
means the module is responding but
56
Silicon Chip
Ne o
The key to this new Clock is the web
interface, shown in Screen 1 (reproduced from last month). It provides
all the status and configuration details,
apart from what’s on the clock face.
The only additional controls are two
pushbuttons on the front of the clock;
their functions can be configured, as
described below. Let’s now go through
the fields one-by-one so we can explain
how they work.
P
Detailed usage instructions
R
l
e
ix
E
L
B
G
This new
clock uses
addressable
RGB LEDs
and has builtin Wi-Fi and
NTP support; plus,
it can connect to a
GNSS/GPS module.
After describing it
last month and providing
assembly details, we now have
full usage instructions and go into
depth on the firmware that runs it.
Australia's electronics magazine
siliconchip.com.au
ED ‘Ana 2 by N
l o g ic
’
Par
t
V
las
l
ho
C
i ne
n
ock
doesn’t have enough data to provide
an accurate UTC time yet
↳ “Valid time, no fix, N satellites”
means the module can’t provide a
location, but it is giving an accurate
UTC time.
↳ “Valid fix, N satellites” means the
module is providing both the accurate
UTC time and a valid location fix
E. Time zone – the currently calculated time zone based on the rules
below. This is what’s used to convert
the UTC time to local time for display
on the clock face.
F. Wi-Fi – the name of the Wi-Fi
network the unit is connected to, or
the access point name it is currently
providing
G. IP address – the unit’s current IP
address on the Wi-Fi network.
H. Wi-Fi signal – the signal strength
of the current Wi-Fi network
I. Auto brightness – if auto brightness is enabled, this is the current
brightness percentage determined
using the rules set below
J. LDR – the raw LDR data. This
should change if you cover the LDR or
shine a bright light on it. It’s updated
a couple of times a second.
K. Saved settings – indicates whether
there are any changes to the settings
that haven’t been written to permanent
(flash) memory
L. Firmware – the current firmware
version and how many times it has
been updated
M. Uptime – how long the Clock has
been running since it was last powered up
#2 Wi-Fi settings
A. Mode – if Mode is set to Connect
to network, the Clock will attempt
to connect to the network with the
given SSID and password the next
time it boots or the Reconnect button
is pressed. Otherwise, it will create its
own network using the given SSID,
and you will need to use the specified
password to connect to it (the network
will be unsecured if the password field
is left blank).
B. SSID – in AP mode, the network
created has this name. In network connection mode, it will try to connect to
a network with this name.
C. Password – in AP mode, if blank,
the network will be unsecured. Otherwise, you will need this password to
access it. In network connection mode,
it will use this password to connect to
the secured network specified.
siliconchip.com.au
Australia's electronics magazine
December 2026 57
D. Hostname – the Clock will identify
itself by this name. It will show up
in the router/access point connected devices list under this
name, and the web interface should be accessible
via https://<hostname>
(you may need to add
.local to the end).
E. Reconnect –
this will drop the
Wi-Fi connection
and then try to
connect again. If
you’ve changed
any of the above,
that will cause it
to use the new
settings.
F. Scan for networks – this will
cause a new window to open, and
after a few seconds, it
should populate with
a list of the Wi-Fi networks that were found,
along with their signal
strengths. Clicking one will
copy its SSID into the SSID field
on the main web interface and close
the new window.
G. Forget network – this blanks out
the SSID and password fields and
saves the new configuration to flash,
so the current network details will
be erased (not necessarily securely,
though).
H. Disable Wi-Fi – this will allow the
clock to keep running but switch Wi-Fi
off. You’ll likely need to reboot it to
restore Wi-Fi operation, unless you’ve
configured one of the buttons to toggle it. If set to use NTP time, this will
cause the time to eventually become
invalid, so it’s better to use GNSS time
if using this feature.
#3 Time settings
A. Time source. This can be:
↳ NTP – use NTP only. Typically
used when no GNSS module is connected.
↳ GNSS – use GNSS only. If the
GNSS module is not giving a valid
time, the display will switch to a waiting pattern.
↳ GNSS Then NTP – it will use the
time from the GNSS module as long
as it’s valid. If not, it will fall back on
NTP time.
↳ NTP then GNSS – it will use the
NTP time as long as it’s valid. If it
58
Silicon Chip
becomes invalid, it will fall back on
the GNSS time, if that is available.
↳ Internal – similar to NTP mode
but it will show the time regardless of
whether a valid NTP synchronisation
has been performed in the last four
hours. If selected after boot, the clock
will start at 12:00 noon.
B. Current time – this echoes the
time field in the status area
C. NTP server – allows you to
change which server is used for NTP
synchronisation. The default is pool.
ntp.org.
D. GNSS baud rate – the baud rate
used to receive data from the GNSS
module. Changing this has immediate
effect. Typical values are 4800, 9600,
14400, 19200, 38400, 57600, 115200,
230400, 460800 and 921600. If unsure,
check your module’s data sheet; some
modules can be configured for various baud rates, but there’s usually a
default.
E. Synchronise – if connected to
a Wi-Fi network, this will trigger an
immediate NTP synchronisation. If
you’re connected to the USB serial
console, you will see messages indicating when it’s triggered and when
Australia's electronics magazine
synchronisation is completed, usually a few seconds later.
F. GNSS Debug Window
– opens a separate window that streams serial
data from the GNSS
module with options
like pausing the display and changing the baud rate.
This can help you
decide if the unit
is getting valid
data from the
module without
having to use a
separate USB/
serial adaptor.
G. Set internal time – if the
time source is set
to Internal, you
can enter the UTC
time and date here
in the format YYYYMM-DDThh:mm and
click Set. The clock’s time
and date will be updated,
with the seconds value set to
zero. The date is important so
that daylight saving changes will
be made correctly in regions that
observe DST.
#4 Time zone
A. You can set two different time
zone offsets from UTC here. They
are set in terms of hours and minutes
(they are typically between -12:00 and
+12:00 but there are some outside that
range).
If your region doesn’t use daylight
saving, you can set them identically. If
it does, set the first one to the time zone
offset when DST is not in effect and the
second one to when it is. The defaults
suit NSW, Vic, Tas and the ACT: +10:00
(non-DST) and +11:00 (DST).
The time zone names can also be
set, although they don’t figure into the
time display calculations.
B. There are three DST modes:
↳ Automatic – use the rules below
to determine when to switch time
zones for DST.
↳ On – force the use of the second
(DST) time zone. This can be toggled
manually.
↳ Off – force the use of the first (nonDST) time zone. This can be toggled
manually.
C. DST starts – this specifies the conditions for the start of daylight saving
siliconchip.com.au
time. You need to specify the time,
week, day of the week and month.
Look up your local time zone/DST
rules. The default is 2am on the first
Sunday in October, which suits NSW,
Vic, Tas and the ACT.
D. DST ends – this specifies the conditions for the end of daylight saving
time. You need to specify the time,
week, day of the week and month.
Look up your local time zone/DST
rules. The default is 3am on the first
Sunday in April, which suits NSW,
Vic, Tas and the ACT.
#5 Clock face master control
This lets you switch all the LEDs on
and off (eg, if they are bothering you
at night) and set the master brightness, which always has effect. If you
set the master brightness to 50%, the
LEDs will never run at more than 50%
brightness.
#6 Hands
Here you can set the RGB colours
of the hour, minute, second and sub-
second hands using a colour picker. If
you set them to dimmer colours, that
will affect their relative brightness.
If you enable “Smooth transition”
for a hand, when it moves to the next
position, it will fade over a second
rather than jumping.
You can also enable and disable
the second and sub-second hands
here, determine whether the sub-
second hand chaser runs clockwise or
anti-clockwise, and set the sub-second
hand trail length in pixels (the trail
fades behind it as it moves). Setting
the trail value to 0 disables the trail.
Screen 1:
this shows
all the Clock
settings,
including
those for WiFi and OTA
updating. The
top section
is a status
display. It’s
implemented
as static
HTML
that uses
JavaScript
to fetch the
status and
settings from
endpoints
on the
device and
triggers other
endpoints to
make settings
changes or
trigger state
changes.
#7 AM/PM LEDs
You can enable and disable these
individually and set their colours
when on.
#8 Auto-dimming
When enabled, if there’s more light
detected by the LDR, the pixels will
run brighter. Otherwise, only the master brightness control has an effect.
You can set the maximum and minimum brightnesses applied to the LEDs
for the two extremes of the LDR sensing, and the LDR values that correspond to these.
By default, the LDR values are
sensed automatically; they are the lowest and highest values seen. This can
be calibrated by placing a black insulating object over the LDR to block all
siliconchip.com.au
Australia's electronics magazine
December 2026 59
light, then shining a bright light at it.
Until you do this, the clock display
may appear to flicker as the LDR hasn’t
been calibrated. If you want to save
this calibration permanently, simply
switch off the Auto-calibrate setting
after calibration and save the settings.
The detected values will be permanent
unless you re-enable Auto-calibrate
or change them manually by editing
the fields.
connected to your Wi-Fi network.
Once you’ve done that, it should only
take a few seconds for it to get the current time from the network, and the
clock face will come alive. Assuming you’ve set the time zone (also
described in the last article), the time
should be correct and you can hang
the clock on the wall.
There are all sorts of things you can
customise, but the defaults are usable,
and many people could be happy with
#9 Testing
them. One of the great things about the
Clicking either button initiates one Wi-Fi interface is that if you decide
of the LED test modes. The first, “Test you want to change something later,
all LEDs”, lights them up white one you don’t need to touch the Clock!
at a time, starting with 12 o’clock and Just open its web page and change the
working its way clockwise around the required settings.
clock face, then the AM/PM LEDs.
If you’re using a GNSS module, you
It returns to the normal clock mode will probably want to change the time
after that. The second, “Test colours”, source to “GNSS then NTP”. That way,
lights up all LEDs red for 10s, then it will use the satellite-based time
green for 10s, then blue for 10s, then as long as it’s certain it’s accurate. If
returns to the normal clock mode.
that goes away for some reason (eg,
marginal satellite signal reception),
#10 Button functions
it will fall back to NTP, then switch
These let you select what the two back to GNSS when the signal comes
buttons, S1 (A) and S2 (B), do when good again.
pressed. You can set separate functions
You may want to adjust the ‘hand’
for a short (<1s) or long (>1s) press of colours, especially if that helps you
each. The defaults are:
easily remember which shows the
• S1 short press shuffles the ‘hand’ hour, minute and second. You may
colours
also want to change or disable the
• S1 long press toggles auto- sub-second chaser if you don’t like it;
dimming
for example, adjusting its colour to be
• S2 short press cycles the second/ more dim or shortening the trail.
sub-second hand modes
It’s a good idea to use auto-dimming
• S2 long press triggers NTP syn- so that the display isn’t too bright at
chronisation.
night. Calibrate the LDR as per the
Other possible options include tog- instructions last month or above, then
gling the AM/PM LEDs on and off, tog- adjust the maximum brightness so it’s
gling DST on/off, changing the master comfortably readable during the day
brightness, saving the settings and and the minimum brightness so it’s a
switching Wi-Fi on and off.
suitable brightness at night.
#11 Save/reset
The options are: Save will store the
current configuration in flash memory
so these same settings are used after
subsequent power-on events; Reload
saved settings discards any recent
changes and goes back to the settings
saved in flash (similar to rebooting
the unit); Restore factory defaults
loads the default settings but doesn’t
change what’s in flash unless you subsequently press Save; Reboot reboots
the unit as if its power was cycled off
and on again.
Screen 2: the VSCode project view
showing all the files needed to build
the firmware.
60
Silicon Chip
Setup and use
The instructions last month
explained how to get the Clock
Australia's electronics magazine
Waiting display colours
If the Wi-Fi network drops out, the
clock will attempt to automatically
reconnect periodically. When it can’t
display the time, or while it’s booting
up, it will show a coloured ‘waiting’
pattern:
• Red ■ – trying to connect to Wi-Fi
• Mauve/purple ■ – connected to
Wi-Fi, waiting for NTP time synchronisation
• Green ■ – Wi-Fi access point is
active (switch to internal or GNSS time
for time display)
• Blue ■ – GNSS mode selected, no
valid GNSS signal yet
• Yellow ■ – Valid GNSS signal
detected, no valid GNSS time yet
siliconchip.com.au
Firmware design
The firmware compiles to about
512kiB, barely more than a quarter
of the Pico W’s flash space. Much of
that is related to Wi-Fi and web serving, as required for the setup, remote
configuration and network time (NTP)
systems.
The program was developed using
VSCode and mostly uses libraries provided for use with the Raspberry Pi
Pico W. We initially used the WS2812
library on GitHub (https://github.com/
ForsakenNGS/Pico_WS2812) for
driving the LEDs but found several
bugs and deficiencies that we had to
address. So the library in our code is
not the same as the one on GitHub.
While we have our own GNSS code,
to keep things simple and robust, we
used the open-source minmea NMEA
parser from https://github.com/kosma/
minmea
The program is split into these main
files (see Screen 2):
• Bootloader.cpp/BootMetadata.
cpp: the code that handles booting the
latest version of the code and switching between the two flash ‘slots’ to
allow over-the-air (OTA) updating.
• ClockFace.cpp: logic for driving
the clock LEDs
• Config.cpp: storing the configuration data in RAM and flash
• dhcpserver.c & dnsserver.c:
example files supplied with
the Pico W SDK needed
for Wi-Fi access point
mode
• EEPROM.cpp:
saving and loading
from flash memory (EEPROM
emulation)
• GNSS.cpp:
helpers to feed
serial data to
the parser and
update the
clock with the
GNSS-derived
time
• The http_
server.cpp file:
works with the
Pico W’s HTTP
server to deliver
the various web
pages, update the
data on them and
react to configuration
changes in the web interface
siliconchip.com.au
• LDR.cpp: using the ADC to get
the LDR brightness, perform LDR calibration etc
• Log.cpp: code for logging messages and implement the USB serial
port debugging console
• minmea.c: the NMEA string parsing library
• NTP.cpp: wrappers and helpers
for NTP time
• OTA.cpp: the over-the-air firmware update code for writing new
firmware to flash memory
• RGBClock.cpp: the main file that
does start-up initialisation, runs the
main loop, and handles S1/S2 button presses
• Timezone.cpp: our own timezone
implementation
• WiFi.cpp: helpers for setting up
a Wi-Fi Access Point, connecting to a
network etc
• WS2812.cpp: the aforementioned
LED driving library
We’ll describe what the code does in
a little more detail later, but first let’s
look at some of the challenges we ran
into during development.
WS2812 library problems
This library is designed specifically for the Pico series, using the
Australia's electronics magazine
PIO peripheral to transmit data in the
serial format required by the WS2812B
‘NeoPixel’ LEDs, so it seemed perfect. It also looked quite simple and
straightforward. Our initial tests
showed it worked.
However, when we put the code in
a tight loop, updating the clock face
continuously, strange things started
to happen. The clock face would only
update occasionally, sometimes jumping whole seconds and being jittery,
despite the LED update routine being
called hundreds of times per second,
as confirmed by debugging over the
USB serial console.
So we had a close look at the library
code and realised that it had a serious bug! The way you update a daisy-
chained string of WS2812B LEDs is
that you transmit 24 bits of data for
each LED in the chain, then you leave
the serial line low for at least 50μs.
This is the ‘reset period’ that tells all
the LEDs in the chain to latch the data
and update their state.
The WS2812 library we used has no
provision for a reset period. It simply
feeds the data to the LEDs and then
returns. So if you call it again within
50μs, as might be the case in a tight
loop, there is no reset period and the
LEDs never update. The fact that our
LEDs updated at all was a coincidence
due to the code occasionally taking
longer!
Also, we realised the library
is not interrupt-safe because
a reset could be triggered
by an interrupt that
stalls the transmission
code for at least 50μs.
We
therefore rewrote the
library to change
a few things.
One important
change was to
add a loop at
the end of transmission to wait
until the last bit
has been sent,
then sleep for
500μs to make
sure the LEDs
reset and update
their states each
time (500μs may
seem excessive, but we
wanted it to be safe, and
the clock face still updates
at over 50Hz).
December 2026 61
We actually ended up putting the
wait/sleep at the top of the function,
since it does the same job that way,
but it means most of the actual LED
data transmission can happen simultaneously with other main loop code,
which improves code efficiency.
We also enabled the PIO FIFO
(first-in first-out queue) and switched
from software control to feeding the
PIO from RAM using DMA (direct
memory access). This way, even if
the processor is interrupted by, say,
a Wi-Fi event, the LED data transmission will complete without interruption. Some of this new code is shown
in Listing 1.
Another change we made to the
WS2812 library was to add the ability
to accumulate LED data in the array.
As initially provided, you could only
set a pixel colour. But say the hour
hand is red, the minute hand is blue
and they are at the same position. We
want that to show up as a purple LED
(red + blue). So we modified the setPixelColor() function to produce an
addPixelColor() function as well.
Time zone difficulties
The next bit of ‘fun’ I had was with
time zones. The Pico W’s C library is
compatible with the standard Unix C
library that uses a “TZ” environment
variable to set the time zone.
Once set, functions like localtime_r() can be used to convert a Unix
timestamp to the local time, which
can be used to update the clock display. The specific code in the Pico W
example is:
struct timespec ts;
get_time_utc(&ts);
struct tm tm;
pico_localtime_r(&ts.tv_sec, &tm);
// DMA completion only means the last word has been
// written to the PIO FIFO. It may not have finished
// shifting out yet.
while (!pio_sm_is_tx_fifo_empty(pio, sm))
tight_loop_contents();
// Allow the final 24-bit word to finish shifting,
// followed by the WS2812 reset/latch interval.
sleep_us(500);
dma_channel_set_read_addr(dma_chan, data, false);
dma_channel_set_trans_count(dma_chan, length, true);
dma_channel_wait_for_finish_blocking(dma_chan);
Listing 1: the modified WS2812::show() function that uses DMA instead of a
blocking loop to transfer data to the LEDs.
case WIFI_CONNECTING:
if (link == CYW43_LINK_UP) {
console_logf(“wifi: link up”);
wifi_state = WIFI_CONNECTED;
wifi_state_since_us = now;
Wi-Fi conundrums
start_mdns(&cyw43_state.netif[CYW43_ITF_STA]);
if (!httpd_started) {
console_logf(“httpd start”);
cyw43_arch_lwip_begin();
httpd_init();
cyw43_arch_lwip_end();
httpd_started = true;
}
console_logf(“ntp start”);
start_ntp();
console_logf(“Wi-Fi connected, IP=%s”, ip4addr_ntoa
(netif_ip4_addr(&cyw43_state.netif[CYW43_ITF_STA])));
}
Listing 2: some of the Wi-Fi state machine logic for bringing the network
interface up and down cleanly.
62
Silicon Chip
The first two lines get the current
UTC time from the internal real-time
clock (which is synchronised to NTP
once per hour). The last two lines then
extract the local time (date, hours, minutes, seconds etc) from that data using
the current time zone per the “TZ”
environment variable.
This all works as long as you never
change the “TZ” variable! If you do, it
will just keep using the old time zone,
no matter what you do. This particularly baffled me (and my AI assistant)
as we checked the underlying code
and it seemed correct.
The Pico’s library code caches the
time zone data after extracting it from
the TZ variable, but it seemed to have
the code to correctly update the cache
if it detected a change in the contents
of the TZ variable.
But despite verifying we were definitely updating the TZ variable, there
was nothing we could do to get it to
use any time zone specified other than
the very first one.
Obviously, there is a caching
bug somewhere in the library, but I
couldn’t figure out a way around it.
This is a non-starter in our application, since the user can change the
time zone information at any time via
the web interface. So I had to write my
own timezone functions, converting
our time zone configuration and timestamp directly into local time, rather
than relying on the library functions.
This is probably a bit more efficient
too, but annoying that I had to ‘reinvent the wheel’.
You can find the resulting functions in the Timezone.cpp file, some
of which is shown in Listing 3.
Australia's electronics magazine
While it’s impressively easy to connect to a Wi-Fi network with the Pico
W (or to create an access point), there
are quite a few steps beyond simply connecting that you need for it
to work properly, and some of them
have pitfalls.
Annoyingly, if you do things in the
wrong order, the Pico W has a tendency to ‘panic’, print an error and
stop running code.
You can’t even use the normal
reflash procedure in that case; you
have to pull out the USB cable, hold
down the BOOTSEL button, and plug
it back in. That’s annoying when the
module is mounted on the back of a
PCB.
siliconchip.com.au
After connecting to a Wi-Fi network,
we need to start the mDNS and NTP
services. mDNS allows you to access
the Clock’s web page via its name in
the URL opened with a web browser,
instead of having to figure out its IP
address (which can change) and putting that into the URL instead.
Alternatively, if creating a Wi-Fi
access point (so that you can connect to the Clock’s Wi-Fi network,
rather than the other way around),
we instead have to start the DHCP
and DNS servers (implemented in the
files dhcpserver.c and dnsserver.c,
respectively).
The DHCP server is needed to
assign the client (your computer/tablet/phone) an IP address; otherwise,
it will just disconnect from the network. In this mode, the role of DNS
is similar to that of mDNS in Network mode, allowing you to look up
and connect to the web configuration
page by name.
The catch with these is in switching
modes, which you might do, for example, during set-up. The Clock will start
in access point mode and you connect
to its network. You can then scan for
Wi-Fi networks, select an SSID and
enter a password.
At that point, it needs to ‘tear
down’ its Access Point, including
shutting down DNS and DHCP. It
can then connect to the Wi-Fi network specified, after which it has to
start the mDNS and NTP services. If
these aren’t done in the correct order
and a service is left behind that was
connected to a Wi-Fi network that’s
gone, the Pico W stack can ‘panic’
and crash the whole thing.
We also have to set the device’s host
name at various points.
So there was quite a bit of work
in developing and testing the Wi-Fi
code so that it can switch modes
back and forth without crashing the
device. Some of the code is shown
in Listing 2.
Wi-Fi library bugs
To provide a Wi-Fi signal strength
(RSSI) indication on the status screen,
when the status was queried through
the HTTP server (httpd), we initially
called the cyw43_wifi_get_rssi() function and report the result.
This mostly worked, but periodically it would fault and print the
message “[CYW43] do_ioctl(0, 127,
4): timeout”.
siliconchip.com.au
Using AI to aid development
A large portion of the code for this project was written manually, but there
were certain areas of ‘drudgery’ where I enlisted an artificial intelligence system to help.
For example, the “http_server.cpp” file has a function that needs to parse
a JSON array from the web interface and extract one of several dozen possible configuration changes, then call the appropriate function to perform the
update. That is not hard code to write, but it is a very repetitive function with
many similar sections.
I drafted some specifications regarding the various settings and what they
did, explaining the context to the AI. It produced the whole function in under
a minute. It would have probably taken me an hour to write and another hour
to test and debug it. It worked first time.
The AI is very useful because it understands all the Pico W APIs, while I
have not programmed a Pico W before. It was able to tell me all sorts of esoteric configuration changes and code snippets for doing things like enabling
NTP and the HTTP server. I could have read the documentation and figured
it out myself —eventually. What took it minutes would have taken me hours.
I wrote a lot of the actual structure of the code and critical functions, like
clock face handling, myself. In other sections of the code, for example, managing the Wi-Fi state, my role was more like a manager. I explained what was
required, provided existing code and used the AI to write helper functions,
which I then integrated into the code and tested.
Naturally, the AI-generated code has to be tested, so I went through and
checked that every function worked as expected. I would have had to do the
same had I written all the code manually. The only code that had to be significantly changed after being written was the Wi-Fi state machine code, which
was fairly complex and required some debugging. The AI helped me refactor
that until it worked well. In total, this cut the development down from what I
estimate would be a couple of weeks to just a few days.
static bool daylight_saving_active(time_t utc) {
if (settings.dst_mode == RGBClockSettings::DST_OFF)
return false;
if (settings.dst_mode == RGBClockSettings::DST_ON)
return true;
struct tm utc_tm;
if (!gmtime_r(&utc, &utc_tm))
return false;
int year = utc_tm.tm_year + 1900;
time_t start = dst_transition_utc(year, true);
time_t end = dst_transition_utc(year, false);
if (start == (time_t)-1 || end == (time_t)-1)
return false;
if (start < end) {
// Northern hemisphere:
// e.g. March -> October.
return utc >= start && utc < end;
}
else {
// Southern hemisphere:
// e.g. October -> April.
return utc >= start || utc < end;
}
}
Listing 3: one of the functions needed for our ‘roll your own’ daylight savings
logic to replace the faulty logic in the Pico W’s library.
Australia's electronics magazine
December 2026 63
The occasional RSSI query failure
wouldn’t matter except that it caused
the whole clock to stop functioning for
several seconds each time, including
updating the clock face. That made it
very annoying.
The solution I came up with was to
move the cyw43_wifi_get_rssi() function call into the main loop twice a second and store the latest RSSI value in
a variable that the HTTP server function can access. That fixed it.
Apparently, calling the cyw43_
wifi_get_rssi() function from within
the HTTP context (possibly inside an
interrupt) is not allowed, although that
isn’t clearly mentioned in the documentation.
More severe Wi-Fi problems
The next problem was subtler and
took a lot of time to debug since it
sometimes wouldn’t occur for several
days and the cause wasn’t obvious.
The clock would connect to the Wi-Fi
network and get NTP time reliably. It
would then refresh NTP time hourly
and remain up to date. It would do this
for hours or days.
Eventually, it would drop off the
Wi-Fi network; it no longer appeared
in the Wi-Fi base station’s client list.
However, the Wi-Fi system on the Pico
W believed it was still connected!
I had built a Wi-Fi state machine that
would do things like automatically
reconnect if the connection dropped. I
figured that was necessary, given that
you might have to reboot the router, or
it might occasionally freeze. (This is
something I think the Pico W library
really should handle by itself, but it
doesn’t).
However, that won’t work if the Pico
W is reporting that the network is still
connected when it isn’t.
I had difficulty reproducing this
problem reliably enough to fully discover why it happens, but I came up
with a solution nonetheless. If Wi-Fi
appears to be working but the NTP
time hasn’t been synchronised in over
two hours, that means two synchronisation events were missed, suggesting a network problem. In that case,
we force reconnection to the Wi-Fi
network.
Assuming that’s successful, it
should trigger an immediate NTP
update, and the time will be refreshed.
We continue to display the time normally during this process, and it
should seamlessly update once the
Wi-Fi connection is working again.
Over-the-air updating
The Clock supports over-the-air
(OTA) firmware updating, so new firmware can be installed without connecting the Pico W to a computer. This is
particularly useful if the Pico W is surface-mounted (so a USB cable can’t be
int main()
{
BootMetadata metadata;
boot_metadata_get(&metadata);
BootSlot permanent = (BootSlot)metadata.permanent_slot;
BootSlot trial = (BootSlot)metadata.trial_slot;
BootSlot target = permanent;
if (trial != BOOT_SLOT_NONE) {
if (boot_metadata_set(permanent, BOOT_SLOT_NONE, false))
target = trial;
}
if (!slot_valid(target)) {
BootSlot alternative = target == BOOT_SLOT_A ?
BOOT_SLOT_B : BOOT_SLOT_A;
if (slot_valid(alternative))
target = alternative;
else
while (true)
tight_loop_contents();
}
boot_slot(target);
}
Listing 4: the main bootloader logic that starts the Clock code.
64
Silicon Chip
Australia's electronics magazine
easily plugged in) or otherwise difficult to access after the Clock has been
assembled or mounted.
We considered using the code
found at https://github.com/usedbytes/
picowota to provide over-the-air
updating for the Pico W. However, it
requires a special tool to upload new
firmware, and we thought it would be
more convenient to use the existing
web interface so that all you need to
perform an update is a web browser
on a Wi-Fi-enabled device.
The Pico W has 2MiB of flash memory, while the Clock firmware only
occupies about 512kiB. We take advantage of the remaining space by storing
two copies of the application firmware
in separate areas of flash, referred to
as slots A and B, each of which is just
under 1MiB. A small bootloader determines which slot should be run when
the Pico W starts.
The 2MiB of flash on the Pico W
board is occupied by the bootloader
(32kiB, although only a fraction of that
is currently used), two copies of the
bootloader metadata (two 4kiB pages),
the two program slots (972kiB each)
and the space reserved for saved settings (64kiB, again much larger than
currently required).
Firmware update files contain two
versions of the same program: one
linked to run from slot A and the
other linked to run from slot B (as
the addresses vary). A small 16-byte
header identifies the file, specifies the
length of each image and contains a
32-bit CRC checksum for each one.
The update is uploaded through
the Clock’s normal web interface.
Because the complete update file is
over 1MiB and is much too large to
hold in the Pico W’s RAM, the HTTP
server processes it as a stream. If the
Clock is currently running the firmware in slot A, it ignores the slot-A
image and writes the slot-B image to
flash. If it is running from slot B, it
does the reverse.
A CRC is calculated as the new
image is received and compared with
the value stored in the update-file
header. The programmed flash is also
read back and checked before the
update is accepted. This ensures that
a truncated or corrupted upload cannot be selected for execution.
Once a new image has been received
and verified, the bootloader metadata is changed so that the new slot
will be tried once on the next reboot.
siliconchip.com.au
Screen 3: the Clock’s log messages
at power-up, shown in the VSCode
Serial Monitor tab.
Importantly, the previously working
firmware remains untouched.
The bootloader then starts the new
firmware (see Listing 4). If it runs
successfully, after a few seconds, it
updates the bootloader metadata to
make itself the permanent version. If
instead the new firmware crashes or
fails before it can do this, the bootloader will return to the previous
known-good slot on the following
reboot.
This also makes interrupted updates
relatively harmless. If power is lost
while slot B is being updated, for
example, slot A is still intact and
remains the selected firmware. The
bootloader is only instructed to try
the new slot after the complete image
has been received and its checksum
has been verified.
The result is a simple form of automatic rollback: updating the Clock
does not overwrite the firmware that
is currently running, and a new version only becomes permanent after it
has successfully booted.
This system is pretty safe from corrupting the firmware. However, there’s
a remote possibility that new firmware
could run long enough to make itself
the default but still be broken. In that
case, you can hold switches S1 & S2 at
power-up and the bootloader will boot
the other, hopefully still good version
of the firmware.
When the code is compiled in
VSCode, the makefile and Python
scripts produce the following files:
• RGBClock-full.uf2: this is the file
to initially program onto the Pico W
over the USB connection. It includes
the bootloader and main program.
• RGBClock-update.bin: this is the
file to upload for OTA updates. It’s
siliconchip.com.au
similar to the UF2 file, but omits the
bootloader and uses a simplified file
structure.
• RGBClock.elf/bin: the file that
will be uploaded automatically if you
click the “Run” button in VSCode. It
will work, except for OTA updates,
because it doesn’t include the bootloader.
Logging and debugging
Because it was difficult tracking
down some of the inherent problems
with the Pico W and its libraries mentioned above, I implemented a logging
and debugging system over the USB
serial interface. If you plug the Clock
into a USB port and open its virtual
serial interface in a terminal program,
each time something important happens, you will see a fairly detailed
message.
These messages are timestamped
with both the time since the clock was
powered up in fractional seconds and
with the current time on the clock (if
available) – see Screen 3.
These messages are also recorded
in an internal buffer so that you can
connect to the USB serial interface
later and run the “dmesg” command
(it needs to be terminated with an LF).
That will show you the 128 most recent
messages with timestamps.
This is very useful if, for example,
the Wi-Fi interface stops working and
you want to see why. You can also
issue numerous other commands over
this interface, including some to display the current status, display the
settings, change all settings, and trigger actions like connecting to Wi-Fi
or rebooting. Issue the “help” command to see the full list of commands
available.
Australia's electronics magazine
Pico 2 W support?
We considered testing this code on
a Pico 2 W, but since the Pico W does
everything we need, it didn’t seem
necessary.
We think it’s likely the code would
compile and run correctly on a Pico 2
W except for the bootloader and OTA
updating code. The Pico 2 W has an
official OTA update solution, so you
could remove our bootloader/OTA
mechanism and just use that.
In short: we haven’t tried it, but if
you have any sort of programming
skills, you could likely get it working. We would be interested in hearing from any reader who attempts this.
Conclusion
The Clock is a straightforward hardware design that makes it very easy to
show accurate local time via NTP or
satellite signals. It’s easy to build and
set up, and the web interface over
Wi-Fi makes customising it simple at
any time. It hides some pretty complex
firmware inside the Pico W module to
make all that possible.
Readers with programming skills
may be interested in downloading the
VSCode project, including the source
code, and customising it. The OTA
update process can be used for simpler changes, or for any future firmware updates we may release. Programming it over a USB cable would
be better for those wishing to make
significant changes.
The general code structure could
actually be useful in a range of
Wi-Fi-enabled projects; features like
Wi-Fi setup using Access Point mode,
OTA firmware updates and web configuration could be adapted for a range
SC
of different projects.
December 2026 65
Working with
AVR & PIC Configuration Fuses
Part 1 by Tim Blythman
Configuration fuses can be a confusing and troublesome aspect of firmware development for
microcontrollers. Setting the fuses incorrectly can sometimes make a chip unusable! So it pays to
be prepared and know what to expect before working with them.
I
n this series, we will look at the ‘fuses’
in 8-bit PIC and AVR parts, including
the hardware and software that might
be needed to program them. We’ll also
have tips for preventing your chip from
become a ‘brick’! This advice may
also aid in understanding the configuration of other chips, such as 16-bit
and 32-bit PICs.
If your only exposure to microcontrollers is something like the Arduino,
you may not even be aware of configuration fuses and why they are necessary. Different manufacturers and
processor families use different terminology, but all the ‘configuration
fuses’ have a similar purpose, regardless of name.
Configuration fuses set the behaviour
of a microcontroller before it even
starts executing code. They are typically settings that can’t or shouldn’t
be changed while the chip is running,
such as the processor clock source or
the function of certain pins on the
chip. They are often programmed at
the same time as the program flash
memory that holds the binary application data.
The original AVR-based Arduino software environment used a
bootloader system to make it very
difficult to inadvertently change the
fuses or other troublesome settings.
That is a small but important part of
why that ecosystem has become so
popular. Still, the underlying microcontroller depends on configuration
fuses to operate.
The ‘fuse’ terminology comes from
the days when OTP (one-time programmable) memory was common.
Back then, once the fuse was set,
or ‘blown’, the setting could not be
reverted. Modern microcontrollers are
more flexible, using reprogrammable
flash memory or EEPROM. Despite
that, you might still find it difficult
to fix some fuse settings if they are
changed carelessly.
Atmel parts use the term ‘fuse bits’,
while PIC microcontrollers refer to
them as ‘configuration bits’. They do
much the same job. You might also see
terms such as ‘configuration registers’
or ‘option bytes’.
The important thing is that they are
non-volatile memory in a microcontroller that cannot be changed by the
running program; they are generally
only changed by an external programmer or debugger.
Some are often intended as security features to avoid deliberate or
accidental changes to the code. Other
features, like a watchdog timer, can
be enabled by the configuration fuses
and can restart the micro if the code
stops working correctly.
Some microcontrollers have multiple programming interfaces, and the
fuses may restrict access to some of
these. This may be for security reasons
or to enable features that are otherwise unavailable. All of this probably
sounds a bit academic, so we should
look at some real-life examples.
This will be quite a detailed exploration of configuration fuses and microcontroller programming, but we hope
that it will give you the background
and insight to avoid getting into a situation where you ‘brick’ a chip and it
doesn’t work anymore. If you are unfamiliar with some of the terminology,
we have provided a glossary.
For the most part, we’ll stick to 8-bit
processors such as Microchip’s PIC16
parts or Atmel’s AVR series of chips.
Microchip has owned Atmel since
2016, but the two architectures remain
distinct, and their respective device
configurations are quite different.
Table 1 – List of recent projects with AVR microcontrollers
Issue
Project
Link (not including https:// prefix)
Microcontroller
June-July 2019
RF Signal Generator
siliconchip.au/Series/336
ATmega328P
September 2020
Pocket Audio DDS Oscillator
siliconchip.au/Article/14563
ATtiny85V
November 2020
Two Large LED Christmas Stars
siliconchip.au/Article/14638
ATmega328P
July 2021
Digital Radio Receiver
siliconchip.au/Article/14926
ATmega328P
July 2022
0-110dB RF Attenuator
siliconchip.au/Article/15385
ATmega328P
June 2023
Basic RF Signal Generator
siliconchip.au/Article/15817
ATmega328P
October 2023
2m Test Signal Generator
siliconchip.au/Article/15971
ATtiny45
March 2026
Graphing Thermometer
siliconchip.au/Article/19833
ATtiny85
May 2026
Simple LC Meter
siliconchip.au/Article/20235
ATtiny85
June 2026
Simple USB Power Monitor
siliconchip.au/Article/20365
ATtiny85
August 2026
Transceiver Test Set
siliconchip.au/Article/20673
ATtiny85
66
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
We’ll use the 8-pin ATtiny85 AVR
family (which includes parts like the
ATtiny45 and ATtiny25) as examples, since they continue to be used
in applications where only a few I/O
pins are needed. We’ll also mention
some ATmega parts, including the
ATmega328P variants that are used in
many Arduino boards.
Watch out for the part naming on
some of the ATmega328 variants.
The ATmega328, ATmega328P and
ATmega328PB are not always interchangeable. It is the ATmega328P,
used on the Uno, that we will concentrate on.
Table 1 shows a list of recent projects
that include these various AVR parts.
The Pocket Audio DDS Oscillator article from September 2020 also includes
a design for a Programming Adaptor
Board that can be used to program an
eight-pin DIP ATtiny85 chip before it
is fitted to the PCB.
More recent AVR microcontrollers
have introduced different programming interfaces. So we will have a
look at these parts and the hardware
needed to work with them.
The PIC16F84
Let’s start with PICs from Microchip
Technology. My first experience with
a microcontroller was the PIC16F84. It
had several features that made it easy
to work with, including reprogrammable flash memory that allowed the
program code to be readily changed.
These early PIC parts had a simple
serial programming interface, referred
to as ICSP (in-circuit serial programming). It requires three signal wires
plus power and ground. We published
a DIY PIC programmer project as early
as the March 1999 issue (siliconchip.
au/Article/4600).
The advent of ICSP meant that it
was possible to quickly iterate software designs without having to
remove the chip from its circuit; only
three pins needed to be shared with
the application circuit. You can find
the ICSP programming specifications
online. The version used on newer
chips is slightly different, but the
overall operation is quite similar; see
siliconchip.au/link/accx
One wire (often labelled as VPP,
for programming voltage, and usually
shared with the MCLR reset pin) is
dedicated to signalling entry into programming mode. The other two wires,
ICSPCLK and ICSPDAT (for ICSP clock
siliconchip.com.au
AVR & PIC Glossary
Application
> The program that runs during normal operation of a microcontroller. Used to
contrast with “Bootloader”, which is a small program in the microcontroller that
can be used to load a new application program.
AVR microcontroller
> An 8-bit microcontroller architecture developed at the Norwegian Institute
of Technology in the 1990s. The first device, the AT90S8515, was a drop-in
replacement for the 8051. The ATmega series was used in the early Arduino boards.
AVRs are still being released with a similar architecture and instruction set.
Bootloader
> A small program in a microcontroller’s firmware that can update application
firmware via a user-defined interface like serial or USB. The serial port
uploader of an Arduino board like the Uno is one example of a bootloader. Some
chips have settings to prevent the bootloader from being overwritten.
Brownout Detector (BOD)
> A peripheral in a microcontroller that can reset the processor if the supply
voltage is too low to avoid improper circuit operation. Typically, different
thresholds can be set through the configuration fuses.
Firmware
> The machine code that a microcontroller executes. It is named this as it is more
permanent than the software on a device like a personal computer, but more easily
changed than hardware.
Flash memory
> A form of non-volatile memory that is typically used to store the firmware of a
microcontroller. Since it does not need to be changed often, it does not require
random write access. Instead, it is erased in bulk (set to binary ones), then
individual bits can be programmed to zero.
HEX file
> A term for a file used to hold the data needed to program the fuses, flash memory
and other non-volatile memories such as EEPROM on a microcontroller. The Intel
HEX format is often used, which contains a list of memory addresses and data
contents in hexadecimal format.
JTAG (Joint Test Action Group) interface
> Designed as an industry-standard testing and programming interface, the JTAG
interface is common among 32-bit ARM processors and other 32-bit parts such
as the Microchip PIC32MX series. Other types of chips, such as FPGAs (fieldprogrammable gate arrays) might support a JTAG interface.
OTP (one-time programmable) memory
> Before flash memory, PROM (programmable read-only memory) devices were used for
firmware storage in microcontrollers. These would only allow their bits to be
changed from 1 to 0, although UV-erasable EPROM (erasable PROM) parts could be
reused after a time-consuming erase process.
Peripherals
> A part of the hardware on a microcontroller that is dedicated to a task not
directly related to the processor. Examples include general-purpose I/O (input/
output) pins, timers and advanced I/O functions such as PWM (pulse-width
modulation) or communication hardware. Core-independent peripherals that allow
circuitry to interact without needing attention from the main processor.
PIC (Peripheral Interface Controller) microcontroller
> The first PIC chips from the 1970s were intended as programmable general-purpose
peripherals to the 16-bit General Instrument CP1600 microprocessor. In the
1990s, the release of parts like the PIC16F84 with erasable flash memory made
microcontroller programming accessible to hobbyists. Newer parts continue to be
released with a broadly similar architecture and instruction set, although the
instruction set has been updated to be better suited to running C programming
language code.
Programmer
> A device that can be used to read and modify the firmware of a microcontroller.
Many programmers, such as the PICkit series, can also be used to perform live
debugging, so they might also be referred to as debuggers.
RISC (Reduced Instruction Set Computer) architecture
> RISC architectures favour a set of simple instructions, usually resulting in
a less complex and cheaper processor. Both AVR and PIC microcontrollers are
considered RISC processors (see the panel on p15 for more about RISC).
Watchdog Timer
> A timer that is capable of resetting a microcontroller if it is not regularly
‘kicked’ by code during normal operation. Abnormal code behaviour, such as an
infinite loop, will cause the watchdog timer to time out and force a processor
reset. Some watchdogs use a timer window so that early kicks can also signify a
fault and trigger a reset.
Australia's electronics magazine
December 2026 67
The PICkit Basic (left), PCB for the PICkit 5
(middle) and Snap (above) programmers. All
of these can program or debug most PIC chips.
and data) are used to shift commands
and data into and out of the chip via
an internal state machine.
The PIC16F84 has a 14-bit configuration word. Ten of the bits are repeated
copies of the CP (code protection)
bit. When CP = 1, code protection is
disabled. When CP = 0, code protection is enabled and it is not possible
to read out the program memory via
the programming interface. Clearing
the CP bit would have been a typical
step in ensuring the security of a critical device.
Fig.1 shows what this looks like in
the PIC16F84 data sheet. The default
‘1’ value corresponds to the state of
flash memory after it has been erased.
So code protection can be disabled by
erasing the flash memory, with the
proviso that flash memory can only be
erased in bulk. Of course, the code is
still ‘secure’ because it has just been
erased.
This also leads us to an important
aspect of configuration fuses that often
has confusing terminology. The process of writing a ‘0’ to a flash memory
cell is often termed programming (eg,
programming the CP bit to change it
from 1 to 0). Counterintuitively, activating a feature sometimes means settings its configuration bit to ‘0’ when
it was previously ‘1’.
We’ll start with PICs because it’s
generally possible to perform a fullchip erase that will reset all the flash
memory bits to 1. In this state, the
configuration fuses are at their factory
defaults, and the part is effectively
as-new. The Atmel chips can be more
complex; we’ll get to them shortly.
Newer chips
Older parts like the PIC16F84 only
have one programming protocol, but
modern PICs have at least two, with
the advent of LVP (low-voltage programming). The older high-voltage
programming (HVP) required a specific voltage, typically 9V or higher,
to be applied to the VPP/MCLR pin
to enter programming mode.
If you’ve read our reviews of recent
PICkit programmers, you will note that
the economical variants completely
omit HVP capability. The PICkit Basic,
which we reviewed in September 2025
(siliconchip.au/Article/18849) is one
example.
The default (erased) state of these
chips is to allow LVP. To program the
LVP bit and disable low-voltage programming, the programmer must use
HVP. So there is a negligible chance of
putting the chip into an unrecoverable
state. At worst, it should be possible to
recover by means of a full erase.
One reason that LVP might be disabled is that it can allow use of the
VPP/MCLR pin as a general-purpose
I/O (usually input-only) pin. In this
case, HVP is still possible, since the
higher VPP voltage used for HVP is
distinct from the usual logic-level
signals.
In LVP, the roles of the ICSPCLK and
ICSPDAT pins are much the same as
in HVP, so the same five-pin interface
can be used for both LVP and HVP;
the programming connections (eg, to
a PICkit) are identical. Fig.2 shows the
minimum wiring for a modern 14-pin
PIC like the PIC16F18126. Other components, like bypass capacitors, are
not shown.
Many other PIC families use a very
similar interface. This includes 8-bit
PIC18 parts, 16-bit PIC24 parts and
32-bit PIC32MX parts.
Programmers
Fig.1: the configuration fuses on the
PIC16F84 provide settings that don’t
change after programming. The Code
Protection bit allows programmed
parts to be used in designs with less
chance of being copied or cloned.
68
Silicon Chip
Australia's electronics magazine
For 8-bit PIC chips (with very few
exceptions), any programmer that supports HVP can be used to perform a full
erase and thus recover any improper
fuse settings. This includes the PICkit
3, PICkit 4 and PICkit 5. Low-cost programmers like the Snap and PICkit
Basic do not support HVP.
Note that the PICkit 3 is not supported by the most recent versions of
the MPLAB X IDE (integrated development environment) software.
siliconchip.com.au
Fig.2: the ICSP programming interface is simple and only requires three I/O pins. While we often dedicate these pins to
ICSP to simplify development and debugging, it’s sometimes possible to use them in the application circuit and still allow
in-circuit programming.
In the past, we have seen and published hobbyist and DIY PIC programmers, but PICkits are cheap enough,
with support for hundreds of chips,
so there has been little need for DIY
designs lately. The official Microchip
programmers work well for us! The
required software programs (MPLAB
X IDE or MPLAB IPE [integrated programming environment]) are free
downloads.
We have just seen the release of the
PDID feature for some new 8-bit PIC
micros. PDID stands for Programming
and Debugging Interface Disable; it can
be used to permanently lock down
the programming interface, even preventing chip erasure. So this is a new
exception to the rule that PICs can
usually be recovered by a full erase.
We have not tested this out yet.
Configuration bits
Fig.3 shows the (many) configuration bits that are available on the
PIC16F18146, which we’ve used in
quite a few projects. This view is from
the MPLAB X IDE, which allows the
configuration to be edited and then
exported as code for incorporation
into the project. The detailed descriptions are handy for avoiding improper
settings.
The settings shown are a good example of what the configuration bits
might control, with the oscillator and
clock sources being first and foremost.
Software running on the PIC16F18146
can control its oscillators but the
CSWEN bit can disable this to avoid
accidental changes. The programmed
(zero) state is the more secure option,
with clock switching disabled.
The brownout voltage setting
(BORV) is another example, since it
needs to be correctly set to guarantee
proper operation under the expected
conditions. Settings like these are
intended to override software selections and prevent problems due to
runaway code.
The configuration word at 0x800A
is used to set up a bootloader. It
defines regions (for the bootloader
code) that can’t be overwritten. This
would include code that executes
during bootloader operation so that the
Fig.3: the PIC16F18146 has many more configuration bits than the PIC16F84. It supports a bootloader, numerous clock
and oscillator selections, plus a windowed watchdog timer.
siliconchip.com.au
Australia's electronics magazine
December 2026 69
Microchip’s
PIC16F84 uses
in-circuit serial
programming (ICSP) to
write to its internal flash memory
and EEPROM. It was popular with hobbyists
because it is easy to program, requiring only
three signal wires, power and ground.
The Arduino Uno uses an ATmega328P,
which is programmed via the in-system
programming (ISP) interface. Source: www.
flickr.com/photos/snazzyguy/16421989212
(CC-BY-SA 2.0)
bootloader is protected. This word also
includes the LVP fuse noted earlier.
The final word includes the same
CP bit as found on the PIC16F84. The
values and settings shown all correspond to a blank chip, with all bits
set to ‘1’. Generally, they are the least
secure but safest option.
Ultimately, those configuration bit
settings are exported as part of the HEX
file that is programmed onto the chip.
In other words, the complete chip programming state can be captured in a
single HEX file. This, and the ability to
recover a chip via a full erase (possibly requiring HVP), no matter the fuse
settings, are some of the reasons that
we prefer PIC chips over Atmel chips.
Atmel ISP interface
The Arduino ecosystem emerged
perhaps a decade after the rise of the
PIC16F84. It was initially based on
various Atmel AVR parts. The Arduino environment had many features
that made it very easy to start programming microcontrollers, including the
fact that a separate programmer was
not needed.
Good software support, including an
IDE (integrated development environment), intuitive library functions, and
a bootloader that simplified firmware
programming all contributed to Arduino’s popularity. Going beyond Arduino’s limitations typically requires
knowledge of a fundamental programming protocol such as ISP.
Parts like the ATmega328P (and
variants), as used in boards like the
70
Silicon Chip
Arduino Uno and Nano, as well as the
ATtiny85 mentioned earlier, use the
‘ISP’ (in-system programming) interface. This is a serial protocol very similar to SPI (serial peripheral interface).
In fact, the ISP pins on the
ATmega328P are the same pins that
are used for the hardware SPI peripheral. In total, the ISP interface uses
four signal wires. The SCK, MOSI and
MISO lines operate as expected for SPI,
while RESET is used to enter and exit
programming mode. Two other connections for power and ground are
also needed.
Like the PIC parts, it’s possible to
disable the RESET pin function to
allow its use as a general-purpose
input. This can be done without any
checks, meaning that it’s possible to
inadvertently disable the ISP interface
from the ISP interface.
Another important distinction is
that ISP programming depends on
the chip having a suitable processor
clock that is at least four times faster
than the ISP clock. This is important
in that it is possible to switch from the
internal oscillator to an external oscillator (such as a crystal), which can be
a problem if such a clock source is not
present when you are trying to program or debug the chip.
So it’s also possible to get into a
state where ISP programming is disabled due to the lack of a suitable
processor clock. This is not terrible,
as external clock modes will usually
accept an external clock signal, and
some programmers provide a clock
Table 2: ATtiny85 family fuse bits (default values shown)
Extended Fuse
Unused
1
1
1
SELFPRGEN
1
1
1
1
1
High Fuse
RSTDISBL
DWEN
SPIEN
1
1
0
WDTON EESAVE
1
BODLEVEL
1
1
0
0
1
1
Low Fuse
CKDIV8
CKOUT
0
1
SUT
1
CLKSEL
0
1
0
Lock Bits
Unused
LB
1
1
1
1
1
1
1
0
Bit 7
Bit 6
Bit 5
Bit 4
Bit 3
Bit 2
Bit 1
Bit 0
Australia's electronics magazine
siliconchip.com.au
signal for this purpose. Still, it can be
inconvenient.
There is also a configuration fuse
that can be used to completely disable
ISP programming. Naturally, if you
manage to set this, you will need to use
a different interface to reset this fuse.
Unlike the PIC parts, a standard chip
erase does not affect the fuse bits. Also,
the default fuse values are not all ‘1’.
So, to return an AVR chip to factory
condition, you must determine the
default fuse values and program those.
Many programing apps keep a database of default values for this purpose.
Finally, there are ‘lock bits’ that
can be used to prevent reading and
programming the flash memory and
EEPROM via any external programmer interface. Like the CP bits on PIC
parts, they can only be cleared with a
full chip erase, ensuring the security
of the memory contents.
So there are a few ways in which ISP
programming can be used to indirectly
disable ISP programming and appear
to ‘brick’ an AVR chip.
Similar to PICs, the AVR HVSP
(high-voltage serial programming)
mode uses a higher voltage on RESET
to force a different mode. This works
differently to ISP and does not require
a processor clock to function. The
catch is that most of the cheaper ISP
programmers do not support HVSP
and so cannot be used to recover an
AVR chip with ISP disabled.
Some AVR devices also support
HVPP (high-voltage parallel programming), which uses a parallel interface
instead of a serial interface. These
devices still need a high-voltage signal, so HVPP is also unsupported by
many common ISP programmers. So
it is clear that AVR fuse programming
is a bit more complex than for a PIC.
It’s entirely possible to create a HEX
file for an AVR chip that contains the
flash memory data and fuse settings.
Many programmers will correctly program the flash memory before applying
the fuses. But to avoid problems, the
fuses are often provided separately to
force the user to double-check the fuse
values and avoid a problem occurring.
In that case, programming is a multistep process, making it considerably
more complex than simply loading a
HEX file into the chip.
ATtiny85 fuses
We’ve noted the ATtiny85 as our
part of interest, but the same data sheet
siliconchip.com.au
Table 3: ATtiny85 family fuse fields
Fuse
Meaning (when set to zero for binary options)
SELFPRGEN
Self-programming (of flash memory by the user program) is
possible.
RSTDISBL
The RESET pin is disabled (the pin can be used for alternative
features such as I/O).
DWEN
The debugWIRE debugging interface is enabled.
SPIEN
The ISP programming interface is enabled.
WDTON
The Watchdog Timer is forced to run.
EESAVE
EEPROM contents are preserved during a chip erase.
BODLEVEL
111: Brownout detector disabled.
110: Brownout enabled with a 1.8V nominal threshold.
101: Brownout enabled with a 2.7V nominal threshold.
100: Brownout enabled with a 4.3V nominal threshold.
Other values are not defined.
CKDIV8
The system clock is divided by 8.
CKOUT
The system clock is fed to the CLKOUT pin.
SUT
Values 00-10 give different startup delays that depend on
CLKSEL. The default 10 value gives the longest delay in all
cases. A value of 11 is not defined.
CLKSEL
These four bits choose the system clock source:
0000: External clock source.
0001: High-frequency internal PLL (16MHz nominal).
0010: 8MHz internal RC oscillator.
0011: 6.4MHz internal RC oscillator.
0100: internal 128kHz oscillator.
0110: low-frequency external crystal oscillator.
1000-1111: various high-frequency external crystal oscillator
options
Other values are not defined.
LB
11: No memory locking
10: Programming of flash memory and EEPROM is disabled.
00: Programming and reading of flash memory and EEPROM is
disabled.
applies to the ATtiny25 and ATtiny45
variants. The main differences are the
amounts of flash memory, SRAM and
EEPROM that each part features. There
are also -V suffixed variants that can
operate over a slightly wider voltage
range. Tables 2 & 3 show the fuse bit
mapping.
These are spread over three bytes,
with a fourth byte dedicated to the
‘lock bits’. Not all of the bits map to features; this can sometimes lead to confusion when fuse values are reported,
since the value of unimplemented bits
is unimportant and different fuse values can mean the same thing.
The recommended behaviour,
which is consistent with the values
reported in the data sheets, is to mark
all unused bits as ‘1’.
Importantly, the fuse bits are the
same for all the ATtiny parts noted.
Unlike the PIC16F parts, the default
Australia's electronics magazine
values are not all ‘1’. Table 2 shows
how these are organised and their purpose. Features are generally enabled
when the bits are programmed and
set to ‘0’, which can be unintuitive.
Some fields use multiple bits to select
between more than two options. Some
fields depend on others, too.
To use the lock bits, the flash memory and EEPROM should be written
first, followed by the fuses as needed.
Finally, the lock bits are set. It’s necessary to do this last; otherwise, changes
to the other parts of memory will not
be possible.
The ATtiny85 has a
high-performance
8-bit core in an
8-pin package.
December 2026 71
The following fuses can be a problem if set incorrectly:
• If RSTDISBL is set to zero, the
RESET pin is not available for ISP
programming, so it becomes impossible. This situation requires HVSP
to recover the RSTDISBL fuse to its
default state.
• If DWEN is programmed, the
debugWIRE interface can be used for
debugging an ATtiny85 chip, which
uses the RESET pin for bidirectional
communication with a debugWIRE
debugger. This too will disable ISP.
To recover from this state requires a
HVSP programmer or a debugWIRE
debugger to temporarily disable the
debugWIRE interface and allow ISP
again.
• SPIEN is used to control ISP, so
if the SPIEN fuse is programmed, ISP
is explicitly disabled and an HVSP
programmer is needed to restore the
SPIEN fuse to its unprogrammed state.
Most other fuses will not cause
problems if they are programmed
incorrectly, although the CLKSEL fuse
will cause a chip to become unresponsive if the chip’s expected clock source
is absent. A typical scenario might be
a circuit that expects the chip to run
from an internal source, but an external source, such as a crystal oscillator,
is selected.
In this case, you should try applying a clock waveform to the CLKI pin;
this is pin 2 on the SOIC and DIP variants of the ATtiny85. Anything up to
4MHz should work for this family,
but remember that if the CKDIV8 bit is
programmed, the system clock will be
divided by a factor of eight. You may
need to choose a very low ISP clock
speed (100kHz is often suggested) to
allow communication in this case.
Using CLKI will only be possible if
this pin is available; that may not be
the case if it is shared with the application circuit. It’s also possible that
certain BODLEVEL settings might
cause problems for circuits that run
at lower voltages.
Table 4: ATmega328P fuse bits (default values shown)
Extended Fuse
Unused
1
1
BODLEVEL
1
1
1
1
1
1
High Fuse
RSTDISBL
DWEN
SPIEN
WDTON
EESAVE
1
1
0
1
1
BOOTSZ
0
BOOTRST
0
1
Low Fuse
CKDIV8
CKOUT
0
1
SUT
CLKSEL
1
0
0
0
BLB12
BLB11
BLB02
BLB01
1
0
Lock Bits
Unused
LB
1
1
1
1
1
1
1
1
Bit 7
Bit 6
Bit 5
Bit 4
Bit 3
Bit 2
Bit 1
Bit 0
Table 5: ATmega328P fuse fields (only those that differ from the ATtiny85 are listed)
Fuse
Meaning (when set to zero for binary options)
BOOTSZ
The bootloader size is:
00: 2048 words.
01: 1024 words.
10: 512 words.
11: 256 words.
BOOTRST
The reset vector is the start of the bootloader (ie, it runs
immediately after a reset).
BLB12
Code in the application section can’t be read by the bootloader.
BLB11
Writes are not allowed in the bootloader section.
BLB02
Code in the bootloader section can’t be read by the application.
BLB01
Writes are not allowed in the application section.
72
Silicon Chip
Australia's electronics magazine
The above scenarios should cover
most of the problematic fuse scenarios that might occur with an ATtiny85.
You can see that there are a few ways
to ‘dig yourself into a hole’ that is difficult (but not impossible) to get out
of with these chips.
ATmega328P fuses
The ATMega328P has a few more
fuses and lock bits than the ATtiny85.
They are a superset of those of the
ATtiny85 but note that they might be
in slightly different locations. Thus,
the binary values of the fuse bytes will
probably differ between the two parts
even for the same configuration – see
Tables 4 & 5.
The big difference is that the
ATmega328P allows a protected bootloader to be programmed. The protection comes from the extra lock bits.
It’s possible to provide a bootloader
on the ATtiny85; the micronucleus
bootloader on boards like the Digi
Spark is such an example. The catch
is that there is no way to prevent the
application program from overwriting
the bootloader, since there are no protections in place.
The protected bootloader is part
of the reason that boards like the
Uno are so robust against ‘bricking’.
The bootloader’s serial programming
interface effectively hides it and the
fuses from being corrupted by the
user program.
The programming environment for
the ATmega328P is much the same as
the Uno, and none of the bootloader
fuses will interfere with the ISP interface; at worst, you may need to perform a full erase, but that should still
be possible without needing to resort
to a different programmer.
If you need to apply an external
clock signal to an ATmega328P, then
the XTAL1 pin should be used. This
is pin 9 on the DIP part.
End of theory
So far, we have described the theory of configuration fuses and how
they can interfere with part programming. In the second and final part of
this series next month, we will take a
hands-on look at the hardware needed
to perform ISP programming of 8-bit
AVR parts, including fuse programming.
We will also investigate some of the
control software that can be used with
SC
programmers.
siliconchip.com.au
CIRCUIT NOTEBOOK
Interesting circuit ideas which we have checked but not built and tested. Contributions will be paid for at
standard rates. All submissions should include full name, address & phone number.
Hybrid SMPS charge pump power supply
I was interested to see the circuit from
Matthew Curlis in Circuit Notebook
(July 2026; siliconchip.au/Article/
20459). It is similar to a circuit I have
used many times in the past. The
difference is that I used to wind my
own inductors and I would invariably power the circuit from a 3.6V
Li-ion cell.
By winding my own inductor, I
could increase the turns on the negative side, thus compensating for the
supply voltage ‘boost’ on the positive
side, giving better tracking between
the output voltages and better efficiency.
I stopped using the double-winding
technique in favour of the simpler
configuration shown here because I
wanted something smaller and simpler.
This circuit uses a combination of
switch-mode and charge pump techniques. It is also based on the very
common MC34063, although almost
any switch-mode IC can be used. The
negative output will be of approximately the same absolute magnitude
as the positive output.
The HW-107 single-cell battery
management system (BMS) provides
charging via a USB connector, charge
regulation and over-discharge protection by disconnecting the load when
the cell drops to about 2.5V.
Even though the MC34063A is only
rated to operate with a minimum
siliconchip.com.au
supply of 3V, I have had no problems
using it down to 2.5V.
The IC is wired in a standard boost
configuration with the 150pF capacitor setting the switching frequency
to 100kHz.
When the internal transistor is on,
pin 1 is pulled to ground, causing
increasing current to flow through L1,
charging its magnetic field.
After a period, it switches off and
the inductor’s magnetic field starts
to collapse, maintaining the (now
decreasing) current flowing out of the
inductor. At the same time, the voltage at the anode of D1 increases, producing a higher voltage at the output
than the input.
By adding a few extra components,
we can also generate a negative output. When REG1’s internal switch is
off, the 10μF capacitor connected to
pin 1 charges up to about 0.6V lower
than the output voltage, via diode D3.
When pin 1 is again pulled to ground,
the capacitor’s positive terminal is
grounded.
So the voltage at the lower end of
the capacitor becomes negative and
causes current to flow through D4, producing the negative output V−. This
repeats continuously, topping up the
capacitors.
REG1 only regulates the positive
output voltage, which it does by
maintaining 1.25V at its feedback pin
(pin 5). If the positive output voltage
Australia's electronics magazine
drops, the voltage at pin 5 will also
drop via the resistive divider. REG1
compensates by increasing the duty
(on time) of its switch, so the output
voltage rises.
There is some regulation of the negative output as well due to the fact
that if the capacitors discharge more
on each cycle, they will draw more
current from the switching node.
However, the negative output voltage
won’t necessarily track the positive
output exactly.
Diode D2 has been added to provide
better balancing of the two output voltages. The negative output experiences
two diode voltage drops, due to D3
and D4. So by incorporating D2, the
positive output also experiences two
diode voltage drops.
Note that current must be drawn
from the positive output to make the
IC switch. If the IC is not switching
enough, the negative output voltage
will drop. A power indicator LED/
resistor can be added from the positive output to ground (as shown) to
increase current draw, thus increasing switching activity and maintaining the negative rail.
The maximum current the IC can
handle is 1.5A. To be on the conservative side, the current limiting is set
to 0.6A by the 0.5W resistor.
Other output voltages can be
obtained by changing the ratio of
the 11kW and 2.2kW resistors. The
December 2026 73
Testing the power supply with a cell voltage of 3.8V and 50mA on both outputs,
it had an efficiency around 60%.
maximum positive output voltage
that can be reliably obtained from
this circuit is about 15V due to the
maximum duty cycle of the regulator
and the lowest expected input voltage of 2.5V. The minimum positive
output voltage is about 1V above the
cell voltage.
The scope grab is from a test with a
cell voltage of 3.8V, with 50mA loads
on both outputs. The efficiency is
about 60%. Switching is somewhat
erratic, which is typical for this IC.
Sometimes switching behaviour can
be improved by putting a 1nF capacitor across the upper feedback divider
resistor, but sometimes it makes it
worse.
The yellow trace is the voltage at
the anode of D3 (10V/div), cyan is the
voltage across the 0.5W current monitoring resistor (200mA/div), magenta
is the voltage at pin 1 of REG1 (10V/
div) and blue is the ripple voltage on
the negative output (20mV/div). The
ripple on the positive supply is lower
than on the negative output but is not
shown here.
M. Housley,
Waiuku, New Zealand. ($100)
Accurate 1Hz oscillator
This simple circuit generates an
accurate 1Hz square wave from a
standard watch crystal. The 32768Hz
(32.768kHz) crystal is driven by the
74HC4060 14-stage binary counter,
which has an internal crystal oscillator driving circuit. The only required
external components are two load
capacitors and one 10MW resistor.
The 18pF capacitor values may
need to be changed to suit the needs
of the particular crystal, but these
values should suit many watch crystals. IC1 divides the 32768Hz signal
74
Silicon Chip
by 214 to produce a 2Hz signal at
its O13 output. That’s then fed to a
74HC74 dual flip-flop, which divides
it by two, resulting in a 1Hz output at
its pins 5 & 9.
The pin 5 output can be fed to external circuitry, while pin 8 (Q2) is used
to drive LED1, which flashes at 1Hz
with a 50% duty cycle to confirm the
circuit is working properly.
Since a typical watch crystal has a
tolerance of 20ppm, the output should
be within the range of 0.99998Hz to
1.00002Hz.
Australia's electronics magazine
This circuit should operate from 2V
to 6V, meaning a common 5V supply
(eg, from USB) can be used.
Instead of a 74HC4060 (IC1) and
74HC74 (IC2), you could use a 4060B
(IC1) and 4013B (IC2). In that case, the
supply voltage range is 3-15V. Note,
though, that while the 4013B is functionally identical to the 74HC74, they
have different pinouts, so you’ll need
to change some of the connections
to the IC2 package. The mapping is
shown in the table below.
Silicon Chip
Function
74HC74
4013B
RD1
1
4
D1
2
5
CP1
3
3
SD1
4
6
Q1
5
1
Q1
6
2
GND
7
7
Q2
8
12
Q2
9
13
SD2
10
8
CP2
11
11
D2
12
9
RD2
13
10
VCC
14
14
siliconchip.com.au
Simulating crystals in LTspice
I started using LTspice because I
am trying to simulate complex radio-
frequency (RF) designs, including VLF
for a world radio project. It’s a lot easier to use than older SPICE simulators.
My designs require crystals to control the frequency of oscillation for
mixers, beat-frequency oscillators
(BFOs) and RF amplifiers. The software has features where a voltage
generator is easily placed and programmed to generate tones (frequencies). However, my design target was
to simulate a crystal oscillator generating the RF and using the harmonics.
The simulator can generate harmonics, but it is impossible to accurately
control the frequency of the crystal
model. Nothing from the internet produced an answer to simulate a crystal
accurately controlling a frequency in
the simulation, so I made one.
A crystal is normally modelled as
an RLC series network with a parallel capacitor, as shown in Fig.1 (for
3MHz). Accurately controlling its resonant frequency in the simulation is
not possible. External loading of this
model drags the resonant point around
by a huge amount, unlike a real crystal.
The model is missing the electromechanical momentum structure of a real
crystal. Thinking along the path of how
to give the model a mechanical force
at the desired frequency, I developed
the model shown on the right-side of
Fig.1. The model is easily placed as a
component in the test circuit.
The 3MHz crystal model has an
internal signal source to apply a
‘mechanical force’ from within the RLC
model that activates exactly on the target frequency. That internal force is a
very low-level 10mV peak sinewave,
but that is enough to force the model
exactly onto the required frequency
with a very high Q. External loading
has little ability to greatly pull the crystal frequency, just as in real life.
The coupling (and turns ratio) of L1
and L2 (simulated coils on an air-cored
transformer) is determined by the K
Spice directive shown on the model’s
schematic. The turns ratio of L1 & L2
is determined by the respective inductance values, while the inductive coupling factor is determined by the final
value in the K directive (1).
You can download the files that
include the model from siliconchip.
au/Shop/6/3622
The file contents may be viewed
with a text editor like Notepad or
Notepad++.
These files also contain a solution
for another problem in the simulation
software where the starting state of a
schematic is very stable and has no
noise. This idle stable condition will
not start oscillators in the simulation.
It is a well-known problem with some
cumbersome solutions suggested.
The simple solution I have uncovered is to keep the software busy
by adding a voltage source that is
not wired to and does not alter the
oscillator circuit; it just generates a
high-frequency signal. The extra voltage source is not required when using
the active crystal model.
Michael Harvey,
Albury, NSW. ($75)
That time of year is nearly here...
CHRISTMAS
Spice up your festive season
with eight LED decorations!
Tiny LED Xmas Tree
54 x 41mm PCB
SC5181 – $2.50
Tiny LED Cap
55 x 57mm PCB
SC5687 – $3.00
Tiny LED Stocking
41 x 83mm PCB
SC5688 – $3.00
Tiny LED Reindeer
91 x 98mm PCB
SC5689 – $3.00
Tiny LED Bauble
52.5 x 45.5mm
SC5690 – $3.00
Tiny LED Sleigh
80 x 92mm PCB
SC5691 – $3.00
Tiny LED Star
57 x 54mm PCB
SC5692 – $3.00
Tiny LED Cane
84 x 60mm PCB
SC5693 – $3.00
Fig.1: a SPICE model designed to simulate a 3MHz crystal via an RLC network
(left) and providing a ‘mechanical force’ (right).
siliconchip.com.au
Australia's electronics magazine
We also sell a kit containing all
required components for just
$15 per board ➟ SC5579
December 2026 75
SILICON
CHIP
.com.au/shop
ONLINESHOP
HOW TO ORDER
INTERNET (24/7)
PAYPAL (24/7)
eMAIL (24/7)
MAIL (24/7)
PHONE – (9-5:00 AET, Mon-Fri)
siliconchip.com.au/Shop
silicon<at>siliconchip.com.au
silicon<at>siliconchip.com.au
PO Box 194, MATRAVILLE, NSW 2036
(02) 9939 3295, +612 for international
You can also pay by cheque/money order (Orders by mail only) or bank transfer. Make cheques payable to Silicon Chip.
12/26
YES! You can also order or renew your Silicon Chip subscription via any of these methods as well!
The best benefit, apart from the magazine? Subscribers get a 10% discount on all orders for parts.
PRE-PROGRAMMED MICROS
For a complete list, go to siliconchip.com.au/Shop/9
$10 MICROS
$15 MICROS
ATtiny85-20PU
Graphing Thermometer (Mar26), Simple LC Meter (May26)
Simple USB Power Monitor (Jun26), Transceiver Test Set (Aug26)
ATmega328PB-AU Low-Power FM Transmitter (Sep26)
PIC12F617-I/P
Active Mains Soft Starter (Feb23), Model Railway Uncoupler (Jul23)
Battery-Powered Model Railway Transmitter (Jan25)
PIC16F1455-I/P
Battery-Powered Model Railway TH Receiver (Jan25)
Dual Train Controller (Transmitter / TH Receiver, Oct25)
PIC16F1455-I/SL Battery-Powered Model Railway SMD Receiver (Jan25)
USB Programmable Frequency Divider (Feb25)
Dual Train Controller (SMD Receiver, Oct25)
PIC16LF1455-I/P New GPS-Synchronised Analog Clock (Sep22)
PIC16F1459-I/P
Mains Power-Up Sequencer (Feb24 | repurposed firmware Jul24)
8CH Learning IR Remote (Oct24), Heat Transfer Controller (Aug25)
Vacuum Controller (Oct25), Adjustable Ultrasonic Cleaner (Jul26)
PIC16F15214-I/SN Silicon Chirp Cricket (Apr23), Mic The Mouse (Aug25)
PIC16F15214-I/P Filament Dryer (Oct24), Tool Safety Timer (May25)
PIC16F15224-I/SL Multi-Channel Volume Control (OLED Module; Dec23)
NFC IR Keyfob Transmitter (Feb25), Rotating Light (Apr25)
PIC16F18115-I/SN Model Railway Destination Display (Aug26)
Mighty USB-C Bench PSU (Oct26)
PIC16F18126-I/SL RGB LED Star (Dec25), DCC/DC Stepper Motor Driver (Apr26)
μDCC Decoder (May26; bell [G] or whistle [W])
PIC16F18146-I/SO Versatile Battery Checker (May25), RGB LED ‘Analog’ Clock (May25)
USB-C Power Monitor (Aug25), DCC Remote Controller (Feb26)
DCC Booster & Reverse Loop Controller (Mar26)
DCC Accessory Decoder (Snap / Servo-type, Jul26)
Mighty USB-C Bench PSU (Oct26)
STM32G030K6T6 Variable Speed Drive Mk2 (Nov24)
PIC16F1847-I/P
PIC16F18877-I/PT
Digital Capacitance Meter (Jan25)
Dual-Channel Breadboard PSU Display Adaptor (Dec22)
Wideband Fuel Mixture Display (WFMD; Apr23)
PIC16F88-I/P
Battery Charge Controller (Jun22), Railway Semaphore (Apr22)
PIC24FJ256GA702-I/SS
Ohmmeter (Aug22), Advanced SMD Test Tweezers (Feb23)
ESR Test Tweezers (Jun24), Human Comfort Indicator (Jun26)
PIC32MX170F256D-501P/T 44-pin Micromite Mk2 (Aug14), 4DoF Simulation Seat (Sep19)
PIC32MX170F256B-50I/SP Micromite LCD BackPack V1-V3 (Feb16 / May17 / Aug19)
Advanced GPS Computer (Jun21), Touchscreen Digital Preamp (Sep21)
PIC32MX170F256B-I/SO
Battery Multi Logger (Feb21), Battery Manager BackPack (Aug21)
PIC32MX270F256B-50I/SP ASCII Video Terminal (Jul14), USB M&K Adaptor (Feb19)
STM32L031F6P6
SmartProbe (Jul25)
$20 MICROS
ATmega32U4
ATmega644PA-AU
PIC32MK0128MCA048
PIC32MX270F256D-50I/PT
Wii Nunchuk RGB Light Driver (Mar24)
AM-FM DDS Signal Generator (May22)
Power LCR Meter (Mar25)
Digital Preamplifier (Oct25)
$25 MICROS
PIC32MX170F256B-50I/SO + PIC16F1455-I/SL
Micromite Explore-40 (SC5157, Oct24)
PIC32MX470F512H-120/PT Micromite Explore 64 (Aug 16), Micromite Plus (Nov16)
PIC32MX470F512L-120/PT Micromite Explore 100 (Sep16)
$30 MICROS
PIC32MX695F512H-80I/PT Touchscreen Audio Recorder (Jun14)
PIC32MZ2048EFH064-I/PT DSP Crossover/Equaliser (May19), Low-Distortion DDS (Feb20)
DIY Reflow Oven Controller (Apr20), Dual Hybrid Supply (Feb22)
KITS, SPECIALISED COMPONENTS ETC
NEOPIXEL RGB LED ANALOG CLOCK (SC7752)
(NOV 26)
Kit: includes everything in the parts list, with the PCB mostly assembled;
optional GNSS module available separately (see p33, Nov26)
- NEO-7M GNSS module (SC6737; $20) and external antenna (SC6738; $10)
- VK2828U7G5LF GNSS module (SC3362)
- BZ-121 GNSS module (SC7414)
PORTABLE EMF PROBE (SC7715)
(NOV 26)
HIGH POWER IDEAL BRIDGE RECTIFIER (SC7718)
(NOV 26)
AUDIO SPOT FREQUENCY TEST GENERATOR
(OCT 26)
$85.00
$25.00
$30.00
Kit: includes all parts except the case, cell and insulating tape (see p55, Nov26) $30.00
Kit: includes all the parts needed to build the TO-220 version (mostly SMD components)
without the conformal coating (see p73, Nov26)
$65.00
- 0.96in OLED display module, white (SC6936) or cyan (SC6176)
- PCM5102 DAC module (SC7749)
- NJM5532D ($3.50, SC7695) or NJM5532DD ($5.00, SC7694) dual op amp
MIGHTY USB-C BENCH PSU (SC7739)
(OCT 26)
Kit: an almost complete kit which includes the case, PCBs, USB-PPS module
and all onboard parts (see p33, Oct26)
- pre-assembled USB-C PPS control module (SC7740)
- 0.91in OLED screen (SC7484)
SEMICONDUCTOR ANALYSER (SC7725)
$10.00
$10.00
$95.00
$25.00
$7.50
(SEP 26)
Kit: includes an assembled PCB with the top-side components already fitted, plus
all other non-optional parts except for the case, battery & label (see p32, Sep26) $95.00
- Hammond 1593XBK plastic case (SC7732)
$17.50
LOW-POWER FM TRANSMITTER
- Elechouse FM transmitter module (SC7712)
- 0.96in OLED display module, white (SC6936) or cyan (SC6176)
- ND0205MA 3V-to-5V DC step-up converter module (SC7713)
(SEP 26)
BATTERY BACKPACK KIT (SC7707)
(SEP 26)
MODEL RAILWAY DESTINATION DISPLAY (SC7697)
(AUG 26)
Kit: includes all the parts except for a Li-ion cell (see p82, Sep26)
Kit: includes all parts, except for the OLED screen (see p54, Aug26)
$10.00
$10.00
$5.00
$25.00
$22.50
siliconchip.com.au/Shop/
- 0.32in white OLED screen (SC7698)
- 0.50in white OLED screen (SC7699)
$5.00
$6.50
DCC ACCESSORY DECODERS
(JUL 26)
I2C CONTROLLER COMPLETE KIT (SC7690)
(JUL 26)
HUMAN COMFORT INDICATOR (SC7646)
(JUN 26)
PINBALL MACHINE KITS
(JUN 26)
SIMPLE USB POWER MONITOR (SC7683)
(JUN 26)
SIMPLE LC METER COMPLETE KIT (SC7657)
(MAY 26)
μDCC DECODER KIT (SC7617)
(MAY 26)
POWER AMPLIFIER CLIPPING INDICATOR (SC7649)
(MAY 26)
STEPPER MOTOR DRIVER KIT (SC7601)
(APR 26)
Snap-type (SC7685): includes the PCB and all non-optional onboard parts
Servo-type (SC7686): includes the PCB and all non-optional onboard parts
Includes the PCB and all onboard parts (see p83, Jul26)
Kit: includes all parts, except the case and battery (see p49, Jun26)
- white 3D-printed case: portrait (SC7453) or landscape (SC7684) version
- 3.3V GY-BME280 module (SC5482)
Control Board (SC7659): includes the PCB and all non-optional onboard parts
Power Supply (SC7680): includes the PCB and all onboard parts
Cable & Connector Set (SC7681): includes 17 10-pin box headers,
34 10-pin IDC connectors, 10m of 10-way ribbon cable,
30 2-way pluggable terminal blocks and 20 2-way polarised headers
Includes the PCB and all onboard parts (see p63, Jun26)
- 0.96in OLED display module, white (SC6936) or cyan (SC6176)
Includes all the parts and the 3D-printed enclosure (see p67, May26)
Includes all the parts and the optional piezo (wire not included). Specify if
you want a bell or whistle sound for the microcontroller (see p88, May26)
$40.00
$40.00
$30.00
$60.00
$12.50
$10.00
$150.00
$50.00
$65.00
$50.00
$10.00
$45.00
$25.00
Short-form kit: includes the PCB and all onboard parts, the case and power supply
are not included (see p35, May26)
$95.00
- pair of red & white PCB-mounting RCA sockets (SC2615)
$4.00
Includes all required parts for DCC or DC mode (see p55, Apr26)
*Prices valid for month of magazine issue only. All prices in Australian dollars and include GST where applicable. # Overseas? Place an order on our website for a quote.
$35.00
PRINTED CIRCUIT BOARDS
PRINTED CIRCUIT BOARD TO SUIT PROJECT
PICO GAMER
ESP32-CAM BACKPACK
WIFI DDS FUNCTION GENERATOR
10MHz to 1MHz / 1Hz FREQUENCY DIVIDER (BLUE)
FAN SPEED CONTROLLER MK2
ESR TEST TWEEZERS (SET OF FOUR, WHITE)
DC SUPPLY PROTECTOR (ADJUSTABLE SMD)
↳ ADJUSTABLE THROUGH-HOLE
↳ FIXED THROUGH-HOLE
USB-C SERIAL ADAPTOR (BLACK)
AUTOMATIC LQ METER MAIN
AUTOMATIC LQ METER FRONT PANEL (BLACK)
180-230V DC MOTOR SPEED CONTROLLER
STYLOCLONE (CASE VERSION)
↳ STANDALONE VERSION
DUAL MINI LED DICE (THROUGH-HOLE LEDs)
↳ SMD LEDs
GUITAR PICKGUARD (FENDER JAZZ BASS)
↳ J&D T-STYLE BASS
↳ MUSIC MAN STINGRAY BASS
↳ FENDER TELECASTER
COMPACT OLED CLOCK & TIMER
USB MIXED-SIGNAL LOGIC ANALYSER (PicoMSA)
DISCRETE IDEAL BRIDGE RECTIFIER (TH)
↳ SMD VERSION
MICROMITE EXPLORE-40 (BLUE)
PICO BACKPACK AUDIO BREAKOUT (with conns.)
8-CHANNEL LEARNING IR REMOTE (BLUE)
3D PRINTER FILAMENT DRYER
DUAL-RAIL LOAD PROTECTOR
VARIABLE SPEED DRIVE Mk2 (BLACK)
FLEXIDICE (RED, PAIR OF PCBs)
SURF SOUND SIMULATOR (BLUE)
COMPACT HIFI HEADPHONE AMP (BLUE)
CAPACITOR DISCHARGER
PICO COMPUTER
↳ FRONT PANEL (BLACK)
↳ PWM AUDIO MODULE
DIGITAL CAPACITANCE METER
5MHZ 40A CURRENT PROBE (BLACK)
BATTERY MODEL RAILWAY TRANSMITTER
↳ THROUGH-HOLE (TH) RECEIVER
↳ SMD RECEIVER
↳ CHARGER
USB PROGRAMMABLE FREQUENCY DIVIDER
HIGH-BANDWIDTH DIFFERENTIAL PROBE
NFC IR KEYFOB TRANSMITTER
POWER LCR METER
WAVEFORM GENERATOR
PICO 2 AUDIO ANALYSER (BLACK)
PICO/2/COMPUTER
↳ FRONT & REAR PANELS (BLACK)
ROTATING LIGHT (BLACK)
433MHZ TRANSMITTER
VERSATILE BATTERY CHECKER
↳ FRONT PANEL (BLACK, 0.8mm)
TOOL SAFETY TIMER
RGB LED ANALOG CLOCK (BLACK)
USB POWER ADAPTOR (BLACK, 1mm)
HWS SOLAR DIVERTER PCB & INSULATING PANELS
SSB SHORTWAVE RECEIVER PCB SET
↳ FRONT PANEL (BLACK)
433MHz RECEIVER
SMARTPROBE
↳ SWD PROGRAMMING ADAPTOR
DUCTED HEAT TRANSFER CONTROLLER
↳ TEMPERATURE SENSOR ADAPTOR
↳ CONTROL PANEL
MIC THE MOUSE (PCB SET, WHITE)
USB-C POWER MONITOR (PCB SET, INCLUDES FFC)
HOME AUTOMATION SATELLITE
PICKIT BASIC POWER BREAKOUT
DUAL TRAIN CONTROLLER TRANSMITTER
DATE
APR24
APR24
MAY24
MAY24
MAY24
JUN24
JUN24
JUN24
JUN24
JUN24
JUL24
JUL24
JUL24
AUG24
AUG24
AUG24
AUG24
SEP24
SEP24
SEP24
SEP24
SEP24
SEP24
SEP24
SEP24
OCT24
OCT24
OCT24
OCT24
OCT24
NOV24
NOV24
NOV24
DEC24
DEC24
DEC24
DEC24
DEC24
JAN25
JAN25
JAN25
JAN25
JAN25
JAN25
FEB25
FEB25
FEB25
MAR25
MAR25
MAR25
APR25
APR25
APR25
APR25
MAY25
MAY25
MAY25
MAY25
MAY25
JUN25
JUN25
JUN25
JUN25
JUL25
JUL25
AUG25
AUG25
AUG25
AUG25
AUG25
SEP25
SEP25
OCT25
For a complete list, go to siliconchip.com.au/Shop/8
PCB CODE
Price
08104241
$10.00
07102241
$5.00
04104241
$10.00
04112231
$2.50
10104241
$5.00
SC6963
$10.00
08106241
$2.50
08106242
$2.50
08106243
$2.50
24106241
$2.50
CSE240203A $5.00
CSE240204A $5.00
11104241
$15.00
23106241
$10.00
23106242
$12.50
08103241
$2.50
08103242
$2.50
23109241
$10.00
23109242
$10.00
23109243
$10.00
23109244
$5.00
19101231
$5.00
04109241
$7.50
18108241
$5.00
18108242
$2.50
07106241
$2.50
07101222
$2.50
15108241
$7.50
28110241
$7.50
18109241
$5.00
11111241
$15.00
08107241/2 $5.00
01111241
$10.00
01103241
$7.50
9047-01
$5.00
07112234
$5.00
07112235
$2.50
07112238
$2.50
04111241
$5.00
9049-01
$5.00
09110241
$2.50
09110242
$2.50
09110243
$2.50
09110244
$2.50
04108241
$5.00
9015-D
$5.00
15109231
$2.50
04103251
$10.00
04104251
$5.00
04107231
$5.00
07104251
$5.00
07104252/3 $10.00
09101251
$2.50
15103251
$2.50
11104251
$5.00
11104252
$7.50
10104251
$5.00
19101251
$15.00
18101251
$2.50
18110241
$20.00
CSE250202-3 $15.00
CSE250204 $7.50
15103252
$2.50
P9054-04
$5.00
P9045-A
$2.50
17101251
$10.00
17101252
$2.50
17101253
$2.50
SC7528
$7.50
SC7527
$7.50
15104251
$3.50
18106251
$2.00
09110245
$3.00
PRINTED CIRCUIT BOARD TO SUIT PROJECT
DIGITAL PREAMPLIFIER MAIN PCB (4 LAYERS)
↳ FRONT PANEL CONTROL
↳ POWER SUPPLY
VACUUM CONTROLLER MAIN PCB
↳ BLAST GATE ADAPTOR
POWER RAIL PROBE
RGB LED STAR
EARTH RADIO
DCC DECODER
DCC BASE STATION MAIN PCB
↳ FRONT PANEL
REMOTE SPEAKER SWITCH
↳ CONTROL PANEL
DCC REMOTE CONTROLLER
MAINS HUM NOTCH FILTER
MAINS LED INDICATOR
DCC BOOSTER / REVERSE LOOP CONTROLLER
↳ FRONT PANEL
SOLAR PANEL PROTECTOR (WHITE)
GRAPHING THERMOMETER
PICOSDR CONTROL PCB
↳ RF PCB
↳ FRONT PANEL (BLACK)
DCC/DC STEPPER MOTOR DRIVER
CALLIOPE AMPLIFIER
MICROMITE AUDIO PLAYER ADD-ON
↳ ALL-IN-ONE
μDCC DECODER
SIMPLE LC METER
WIFI ALARM MONITOR
POWER AMPLIFIER CLIPPING INDICATOR
PINBALL MACHINE CONTROL BOARD
↳ POWER SUPPLY
↳ PLAYER LED BOARD
↳ SCORE LED BOARD
↳ LED OUTPUT BOARD
↳ BUMPER LED BOARD
↳ CASCADE LED BOARD
↳ SWITCH INPUT BOARD
↳ GENERAL INPUT BOARD
↳ HIGH-CURRENT INTERFACE
↳ ROLLOVER INTERFACE
↳ BUMPER DRIVER
SSB TRANSMITTER (MikeOne/Two/Three)
SIMPLE USB POWER MONITOR
HUMAN COMFORT INDICATOR
ADJUSTABLE ULTRASONIC CLEANER MAIN PCB
↳ FRONT PANEL CONTROL PCB
SNAP-TYPE DCC ACCESSORY DECODER
↳ SERVO-TYPE
I2C CONTROLLER
TRANSCEIVER TEST SET RF/AUDIO PCB
↳ CONTROL PCB
MODEL RAILWAY DESTINATION DISPLAY
↳ FLEX ANTENNA PCB
SEMICONDUCTOR ANALYSER
LOW-POWER FM TRANSMITTER
↳ LID (BLACK, 0.8mm)
BATTERY BACKPACK
MIGHTY USB-C BENCH PSU MAIN PCB
↳ FRONT PANEL
AUDIO SPOT FREQUENCY TEST GENERATOR
NEOPIXEL RGB LED ‘ANALOG’ CLOCK (WHITE)
PORTABLE EMF PROBE
IDEAL BRIDGE RECTIFIER
↳ TO-246 VERSION
↳ SMD TO-263 VERSION
↳ SMD TO-220 VERSION
DATE
OCT25
OCT25
OCT25
OCT25
OCT25
NOV25
DEC25
DEC25
DEC25
JAN26
JAN26
JAN26
JAN26
FEB26
FEB26
FEB26
MAR26
MAR26
MAR26
MAR26
APR26
APR26
APR26
APR26
APR26
APR26
APR26
MAY26
MAY26
MAY26
MAY26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUN26
JUL26
JUL26
JUL26
JUL26
JUL26
AUG26
AUG26
AUG26
AUG26
SEP26
SEP26
SEP26
SEP26
OCT26
OCT26
OCT26
NOV26
NOV26
NOV26
NOV26
NOV26
NOV26
PCB CODE
01107251
01107252
01107253
10109251
10109252
P9058-1-C
16112251
06110251
09111241
09111243
09111244
01106251
01106252
09111245
01003261
10111251
09111248
09111249
17112251
04102261
CSE251101
CSE251102
CSE251103
09111242
01111212
01110251
01110252
09111247
04103261
01304261
01104261
08107261
08107262
08107263
08107264
08107265
08107266
08107267
08107268
08107269
08107260
08117261
08117262
06103261
04104261
21105261
04105261
04105262
09111254
09111255
09111256
06104261
06104262
09111252
06101233
P9062-1-C
CSE260501C
CSE260502
11105261
04107261
04107264
04111261
19101252
04109261
18109261
18109262
18109263
18109264
Price
$30.00
$2.50
$7.50
$10.00
$2.50
$5.00
$12.50
$5.00
$2.50
$5.00
$5.00
$5.00
$2.50
$5.00
$7.50
$2.50
$5.00
$5.00
$7.50
$3.00
$5.00
$5.00
$7.50
$2.00
$5.00
$2.50
$5.00
$1.50
$2.50
$2.50
$15.00
$25.00
$7.50
$2.50
$5.00
$2.50
$5.00
$5.00
$2.50
$2.50
$2.50
$2.50
$5.00
$2.50
$5.00
$5.00
$7.50
$5.00
$3.00
$3.00
$3.00
$5.00
$5.00
$2.00
$2.00
$7.50
$5.00
$5.00
$3.00
$5.00
$5.00
$5.00
$15.00
$5.00
$5.00
$5.00
$5.00
$5.00
INRUSH CURRENT LIMITER
ADVANCED CABLE TESTER
↳ LID (BLUE)
MAGIC CANDLE
DEC26
DEC26
DEC26
DEC26
10110261
04101261
04101262
05112251
$7.50
$5.00
$10.00
$2.50
NEW PCBs
We also sell the Silicon Chip PDFs on USB, RTV&H USB, Vintage Radio USB and more at siliconchip.com.au/Shop/3
Dr Hugo Holden’s
Magic
Candle
Impress your friends and family
with an electric lamp that you
light like a candle – with a
flame from a match or a lighter
– and blow out with your breath!
A
fter the invention of the
transistor at Bell Labs in
1948, many portable battery-
powered projects became possible that
were previously impractical using
valves. Various projects appeared in
hobby magazines in the 1950s to 1970s
using transistors in different roles.
One such project used a small light
bulb to simulate a candle. You could
activate it with a burning match or cigarette lighter held near the lamp, then
blow it out (switch it off) with your
breath, just like blowing out a candle! I attempted to make one of these
in the early 1970s, but the design was
not ideal and it didn’t work properly.
I thought it would be fun to revisit
this idea. Rather than trying to make
it look exactly like a candle, it could
instead be a lamp used to provide
a romantic light source at a dinner
engagement. The colour temperature
and intensity of the lamp would provide similar illumination and colour
temperature to a single candle. Importantly, it could be lit using a match or
lighter, then later blown out, just like
a candle.
This would mean it wouldn’t need
any external on/off switch, although I
78
Silicon Chip
wanted to have a way to switch it on
or off without using a flame if necessary (more on that later). You can see
a video of the Magic Candle in operation at www.youtube.com/shorts/
yPeLhmbzurw
After using it a few times, my wife
suggested it could be improved by
placing a diffuser (lampshade) over
the bulb. So I added a collar on the
upper part of the structure, 30mm
from the top, to support a removable
light diffuser made from some translucent acrylic tube that was 80mm tall.
This worked very well, as shown in
Photos 1 & 2.
The tube needs to be slipped off to
light the candle with the match, but
with the tube present, it is just as easy
to blow the candle out as before.
Design basics
I decided on negative temperature-
coefficient resistors (NTC thermistors)
for temperature sensing, rather than a
thermocouple or IC-type sensor using
a P-N junction. NTC thermistors are
inexpensive and very small. They have
a large change in output voltage per °C
with a simple series resistor; their lack
of linearity doesn’t matter in this case.
Australia's electronics magazine
I decided on three C cells as the
power source to keep the size reasonable while still providing a long battery life. AA cells are too low in capacity for the task, while D cells are too
large. Keep in mind that we’re using
an incandescent lamp, rather than a
more modern LED, because its light
is more pleasing and relaxing, similar to a candle.
This is because, in essence, a candle
is incandescent. They both emit continuous light spectra. A candle burns
yellow due to the incandescence of
burning carbon particles (soot) in the
wax.
A wax candle generally has a light
output equivalent to a 5-7W incandescent bulb, so I elected to use a 6V, 5W
automotive filament lamp, running
from an average voltage of around 4V.
This gives a 2.2W lamp with a similar colour temperature to a candle; I
found that was plenty of light output
for the task.
The light emitted is a little whiter
than a candle with fresh cells, but
still (just) usable with a more orange
colour down to 1V per cell. Using good
alkaline cells, the lamp will run for
around 10 hours total before needing
to be replaced.
The device is 25cm tall with a 33mm
diameter above the base. C cells have
close to a one-inch (25.4mm) diameter,
so they fit easily into the 30mm inner
diameter tube. Conveniently, Bunnings carries a suitable white plastic
pressure pipe, made by Holman and
said to be a “25mm nominal pressure
pipe” (PVP2512-1). The pipe I bought
was just a little ovoid, but it was still a
good press fit into 33mm holes I cut in
the base rings with a standard 33mm
hole cutter tool, which cuts the hole
a little oversized.
So that you can leave the cells in
the lamp between uses without them
going flat, the Magic Candle’s power
consumption needs to be near zero
when not in use. This is a little tricky
given that I didn’t want it to have a
mechanical on/off switch. The measured current consumption when not
in use is a mere 4nA, so the cells will
last for their shelf life.
I also had to consider the supply
voltage; it would start over 4.5V with
fresh cells, but the lamp is still usable
down to about 2V. So the control electronics would need to work over that
range. This created some interesting
design challenges; I managed to make
siliconchip.com.au
Photos 1 & 2: the shade is optional. It helps to diffuse the light, but you have to
take it off to ‘light’ the Magic Candle.
it work down to around 2.2V, which I
thought was adequate.
The thermal sensor
In this design, the thermistor temperature sensors are inside a hollow
brass tube that is close to 3.175mm
in outer diameter, 34mm long, with a
wall thickness of about 0.35mm – see
Photo 3. The brass tube dominates
the thermodynamic properties of the
sensor because the mass of the brass,
its heat capacity, and its surface area.
So, if the tube is heated 10°C above
ambient temperature, how long might
it take to cool back down? Objects gain
or lose heat by radiation, conduction
and convection. For small objects
that are not extremely hot, convection tends to dominate. Radiation is a
weak mechanism of heat loss unless
the object is very hot or very big.
Newton’s law of cooling says that an
object cools in an exponential decay,
ie, dT/dt = -k(T – Tenv), where dT/dt is
the change in temperature with time,
T is the current temperature and Tenv
is the ambient temperature.
The decay constant, k, can be found
by experimentation or estimated as k =
hA ÷ mc, where A is the surface area,
m the mass, c the specific heat capacity
of the material and h the ‘convective
heat transfer coefficient’. h depends
on the geometry of the object, its orientation, air or fluid convection and
surface air speed.
In free air, h typically varies between
5W/m2K and 25/m2K. The cooling
time constant, TC, to decay to by 63%
is 1/k. So after 3TC, it would have
cooled down 95% of the way toward
ambient temperature.
siliconchip.com.au
For a small tube cooling in air, we
can assume an h value of around 10.
For the brass tube used here, the constants are m = 0.000914kg, c = 380J/
kgK and A = 0.000355m2, making k =
0.0102. So the ‘cooling time constant’
(somewhat analogous to the time constant for a capacitor to discharge to
37% of its initial value) is 98 seconds
(1 ÷ 0.0102). Thus, in 3 × 98 seconds
or about five minutes, after heating by
10°C, it will be within about 0.5°C of
ambient temperature.
So there is moderate thermal inertia in the small brass tube, much more
than the thermistors, which have a
thermal time constant of about 7-10
seconds. But you cannot just place
the fragile thermistors in the space by
the lamp; they must have some form
of mechanical protection and support.
Clearly, a differential temperature
sensor is required. It must detect a
small change in temperature and still
work over a wide range of ambient temperatures. So, if we can detect a temperature rise at the exposed end of the
tube compared to the lower portion,
Fig.1: by placing the two NTC
thermistors in a resistor bridge,
the comparator (IC2d) responds
to the difference in temperature
between them. The different-value
fixed resistors add a fixed offset
so that the comparator doesn’t
trigger until one thermistor is a
few degrees hotter than the other.
we should be able to detect if a match
or lighter is lit near it.
The thermistors are in a differential bridge arrangement to decrease
the response time in detecting heating, and to shorten the recovery time
after the sensor tube has been significantly heated.
Thermistors NTC1 and NTC2 are
wired into a bridge circuit, feeding
an op amp that’s used as a comparator
with a small amount of hysteresis (see
Fig.1). When the comparator output
goes positive, the lamp is activated and
is held on by an S-R flip-flop.
The 5kW thermistors used (RS 151221) have a sensitivity of around -4%
per °C at 25°C, indicating a ‘B value’
of about 3640. So a change of +1°C
would drop the 5kW resistance by
around 200W near room temperature.
A plot of resistance versus temperature for an NTC thermistor has a familiar exponential decay format, and the
tangent to the graph, representing a
change in resistance with a change in
temperature, decreases with increasing temperature – see Fig.2 overleaf.
Photo 3: the two NTC
thermistors are placed
25mm apart in this brass
tube.
Australia's electronics magazine
December 2026 79
Fig.2: this shows how the resistance
of an NTC thermistor varies with its
temperature. The relationship is welldefined but nonlinear.
Fig.3: forming a voltage divider with
an NTC thermistor and a fixed resistor
gives this sort of voltage output.
When a resistor is added to make
a divider, the result is an S-shaped
curve, as shown in Fig.3.
The output of the bridge is related to
the difference between the two thermistor’s temperatures, and the types
of thermistors I used are designed to
have matched curves, which helps to
balance the bridge. The bridge circuit
works effectively over the full voltage
range (2.2-4.8V) and a wide temperature range. It does not require that the
voltage or current supply to the bridge
is stabilised or regulated.
I found it was not necessary to compensate for the non-linearity, or the
effects of lowering battery voltage.
The offset voltage due to the 1% tolerances of the 4.3kW and 4.7kW resistors is much higher than the op amp’s
input voltage offset.
80
Silicon Chip
Another advantage of the bridge
arrangement is the tiny amount of
self-heating of the thermistors due to
their operating currents largely cancelling out.
The different values of the 4.3kW
and 4.7kW resistors in the voltage
dividers set up a 180mV offset with
new cells (about half that with discharged cells), which corresponds
to an approximate 2°C offset at 25°C.
This offset keeps the comparator output negative unless there is a transient temperature gradient, with NTC2
being hotter than NTC1. That causes
the comparator output to go positive.
The offset of 2°C is high enough to
eliminate noise and ambient temperature change effects, but low enough
that the response time is fast once a
flame is held in proximity to the brass
tube holding the thermistors.
After the flame is removed, which
is the natural response after the lamp
lights, the brass tube and sensor temperatures settle to a fairly uniform
temperature along the tube length.
The comparator output then returns
negative. The tube cools back down to
near ambient temperature over some
minutes, but due to the offset voltage, the comparator output is low the
whole time.
Even though the brass tube takes
around 98 seconds to lose 63% of the
acquired temperature increase, the
sensor system becomes re-operational
much more quickly because of the
bridge arrangement. On examining
the comparator output after the match
is removed, the heat in the brass tube
reaches equilibrium in 8-10 seconds
(similar to the thermal delay time
between the two sensors). Both the
sensors reach similar resistance values, and the output of the comparator
goes low. At that point, another match
can be used to light the lamp.
In the circuit diagram (Fig.4), the
output from the heat sensor comparator is AC-coupled, with a one-second
time constant, to the S-R flip-flop that
controls the lamp’s on/off function.
The flip-flop is thus driven with a
pulse when the comparator output
goes positive.
Otherwise, the eight-to-ten-second
sensor thermal recovery time could
inhibit the S-R flip-flop and prevent
the user from blowing the lamp out
in a short time frame, immediately
after they lit it with the match, if they
chose to do that.
Australia's electronics magazine
Heat from the lamp
When the lamp is running, it results
in a temperature gradient along the
brass tube, heating thermistor NTC2
more than NTC1. This reduces the
offset voltage. To correct for this and
make sure the sensor system does not
become too sensitive, the lamp’s current flows through R26, an 11-turn coil
of 32 AWG (0.2mm diameter) enamelled copper wire.
This heats the NTC1 end of the brass
tube to the tune of around 30mW, balancing out the thermal offset due to
heat from the lamp.
Blowing it out
Blowing onto the sensor, the breath
temperature is around 29°C, and it’s
possible the ambient temperature
is around that figure. Therefore, it’s
unlikely we can detect blowing on
the tube simply using a temperature
change. On the other hand, the heat
from a match or lighter is always significantly higher than ambient temperature, so it’s easy to detect.
Therefore, the task of sensing when
the lamp is ‘blown out’ was assigned to
an electret microphone (MIC1).
The output from the microphone is
AC-coupled to an amplifier made from
another op amp. A blow of air near the
Magic Candle’s lamp/microphone port
will reset the S-R flip-flop and switch
off the lamp, even if the thermal sensor
comparator output is still high.
A blow of air from the mouth onto
the area near the lamp from any direction is enough to switch it off. The
microphone port does not have to
be blown into directly; just blowing
toward the lamp from any angle will
extinguish it. The arrangement is resistant to background noise and, within
limits, is quite stable. It’s a little easier
to blow out than a real candle.
The microphone arrangement is surprisingly resistant to moderate ambient noise. For example, holding the
activated Magic Candle right at the output vent of a window air conditioner,
running on full fan speed, won’t extinguish the lamp, nor will fairly loud
voices or noises nearby, yet a blow
on it from the mouth will easily do it.
Standby mode & alternative
switching methods
When the Magic Candle has not
been in use for a few minutes (the
lamp is off), it goes into a super-lowpower shutdown mode with just 4nA
siliconchip.com.au
of current consumption. If the Magic
Candle is inverted for a second or two
and returned upright, gravity activates
a mercury switch and the control circuitry returns to operating mode for
three to four minutes.
In this mode, it consumes around
20mA, waiting for the lamp to be
lit by a match. If no match is available, inverting it again and holding it
there for four seconds, then putting
it back upright, provides an alternative method of lighting the lamp. The
same technique can be used to switch
it off if you don’t want to blow on it
for some reason.
More circuit details
The circuit is more elaborate than I
imagined it would be. For the lamp to
receive power from the battery, PNP
transistors Q1 & Q3 must be biased on.
They are driven by N-channel Mosfets
Q2 & Q4, respectively.
The 470μF capacitor provides significant charge storage in the gate circuit of 2N7000 Mosfet Q2. When the
Magic Candle is inverted, the mercury switch closes, and this capacitor is rapidly charged via diode D1
and the 120W current-limiting resistor to the battery voltage (less the forward drop of D1). This causes Q2 to
conduct, bringing PNP transistor Q1
into saturation.
At the same time, the S-R flip-flop is
reset via diode D2, ensuring the lamp
starts in the off state. The Magic Candle
is now ‘armed’ and ready to be lit. If
it is not lit after three to four minutes,
the charge stored in the 470μF capacitor is depleted via the 750kW resistor and the lamp’s filament, and the
circuit goes back into the low-current
standby mode.
If the lamp is on, the current flow
is reversed in the 750kW resistor, preventing the 470μF capacitor from discharging. So the lamp stays on unless
the S-R flip-flop is toggled from the
output of the microphone amplifier
or the mercury switch/timer.
One might wonder why the emitter
of Q3 was not fed directly from the
battery supply, which would gain an
extra 60mV for the lamp.
When the unit is in the process of
powering down, after the lamp has
not been on for some minutes, the
OP495’s total power rail voltage falls
below 2.6V. This can cause the S-R
latch output, pin 7 of the OP495, to
go positive and start recharging the
470μF timing capacitor.
With the arrangement settled on,
the current powering the lamp helps
to pull down the op amp’s supply rail
quickly, preventing the re-activation of
the lamp during switch-off.
I used PNP transistors instead of
Mosfets for Q1 & Q3 because they provide reverse battery protection since
their base-emitter reverse breakdown
voltage is higher than the battery supply of 4.5V. The intrinsic body diodes
in Mosfets conduct with just a couple
of volts in reverse across the channel.
Regarding the lamp drive voltage,
around 50mV is dropped by R26; the
two saturation voltage drops of the
ZTX951 transistors are only about
60-70mV each along with the internal
resistance of the C cells, the result is an
operating current of around 0.5-0.56A
for the 6V 5W lamp.
To allow the circuitry to operate
down to low total battery voltages of
around 2.2V, an LMC7660 (IC1) has
been deployed to generate a negative
voltage rail. This results in a near-
doubled total voltage supply for the
OP495 op amp. The LMC7660 operates down to 1.5V.
A more modern version of this
device, which is pin compatible, is
the MAX1044. Of note, the original
IC(L)7660 parts had a latch-up problem and required an external diode
to prevent it. The LMC7660 and the
Fig.4: the Magic Candle circuit is based around quad op amp IC2. Its V+ supply comes from the battery, while its Vsupply is a negative rail generated by IC1 from the battery supply. The roles of the four op amps within IC2 are shown by
the dark blue text. PNP transistor Q1, driven by Mosfet Q2, is used to save power, while identical transistor Q3, driven by
Mosfet Q4, switches the lamp.
siliconchip.com.au
Australia's electronics magazine
December 2026 81
Fig.5: this simple circuit, with
a variable power supply and
voltmeter, lets you determine the
threshold voltage of a 2N7000
Mosfet.
Photo 4 (top): you insert the three C
cells through this hole in the base.
A metal plate attaches via the two
exposed screw threads to hold them in
and make electrical contact.
Fig.6: assemble and wire up the PCB as shown here. Most axial components,
like resistors and diodes, are mounted vertically on the PCB to save space.
Watch the orientations of the diodes, shown by their cathode stripes here.
Maxim part do not have the same concern, as far as I am aware.
With the LMC7660, the minimum
total voltage the op amp receives is
in the order of 4V even with batteries
discharged down to 2.2V (0.73V per
cell). While there are very low voltage
op amps available, I found it better to
simply use a standard op amp with
the boosted supply via the LMC7660,
which has the additional benefit of
creating a split supply.
Component variances
This design relies on properties of
the 2N7000 Mosfets (Q2 & Q4) that are
not guaranteed for every sample of the
device. The data sheet for the 2N7000
suggested that the drain-source leakage current (with a zero gate-source
voltage) could be as high as 1μA, but
for the ones I tested, the leakage in this
condition is around 4nA.
Even if it were as high as 1μA, that
would still take over 900 years to discharge the C cells!
Ideally, the 2N7000 Mosfets should
be selected for a low gate-source
threshold voltage as this lowers the
82
Silicon Chip
minimum operating voltage. If the
2N7000s are not selected parts, it will
work, but it may stop working with the
battery discharged to 2.5-3.0V rather
than the 2.2V design target.
I had seven 2N7000s in my junk box,
and all but one had a gate threshold
voltage close to 1.55V. The seventh
was quite different, at 2.5V. The data
sheet indicates it can range from 0.8V
to 3V. A maximum of 1.6V is ideal for
the project.
Testing is simple, as shown in Fig.5.
This uses a similar drain current to the
operating circuit. As the power supply
voltage in this setup is reduced, the
drain voltage will fairly abruptly rise
at some point. Above that gate-source
(power supply) voltage, the drain voltage will stay low.
I tested several 470μF 6.3V capacitors and their leakage resistance was
typically at least 10MW. This does not
present any difficulties here given the
750kW charge/discharge resistor.
Construction
Start by assembling the 25.5 ×
68.5mm PCB, using the overlay
Australia's electronics magazine
Photo 5 (bottom): here’s the
metal plate, held in place by two
thumbscrews.
diagram shown in Fig.6 as a guide. All
the PCB pad holes are 0.7mm in diameter, except for the ones on the four
corners and the two for the mercury
switch strap. To keep it as compact as
possible, many of the 1/8W resistors and
diodes are mounted upright.
0.9mm PCB pins are used on the
board corners. They provide plug-in
connections for the two battery wires
using their mating sockets. The pins
near the lamp are soldered directly to
the lamp socket tags. This attaches the
PCB to the top disc.
Overall assembly
Fig.7 shows the basic dimensions.
The size I made the base rings (Fig.8)
was determined by the hole saws I had
in my tool kit. They can be modified as
desired. I used a white material called
Bramite, but 8mm-thick white acrylic,
HDPE or PEHD will also be suitable.
I found that a 33mm hole cutter
made a hole that was a little over
33mm and created an interference
fit with the PVC tube, so no glue was
required, but glue can be used. Photos
4 & 5 show the base assembly.
siliconchip.com.au
Fig.7: the overall
configuration of the Magic
Candle inside its PVC tube.
Also shown on the left
are the details of the
metal strap that
connects the
battery positive
terminal to the
circuit board.
Fig.8: this
shows how
I made the
Candle base
out of three
rings with
staggered sizes.
You don’t have
to make them
exactly the same
size as mine.
December 2026 83
Thermal balance coil
(R26):
11 turns
of 32
AWG wire
Fig.9: the details of the top cap and
battery retaining plate.
Fig.10: here is how to assemble the
temperature-sensing brass tube with
the two NTC thermistors inside it.
Cable tie
For the particular tube I used (purchased from Bunnings), making
the diameter of the top disc around
29.4mm allowed it to be a tight press
fit. Glue can be added if required. Making the outer diameter around 33.5mm
closely matched the outer diameter
of the pipe. A taper is a good idea
on the top disc because then there is
less shading of the light around the
lamp’s base.
The top disc (Fig.9) can be made of
various white materials; FR4 fibreglass
is fine, and it can be painted white.
The light bulb socket has a flare on
it where the lamp screws in, so the
hole in the top disc needs to be made
10.5mm rather than 10mm. I wrapped
some Scotch 27 tape around that to
make sure the socket was a good central fit in the hole.
There is a layer of clear heatshrink
tubing on the wire side of the sensor
tube. The enamelled 32 AWG wire is
wound directly onto this. The top part
of the sensor tube is painted with red
VHT paint to make it obvious that the
match should be held near it, and not
near the microphone port.
Making the sensor tube
Photo 6: the PCB mounts vertically,
with the thermal tube on one side and
the microphone tube on the other.
The construction of the sensor tube
is shown in Fig.10. The two thermistors are at each end of a 34mm-long
hollow 1/8in (3.175mm) diameter brass
tube, spaced about 25mm apart. This
tube is filled with non-conductive
thermal paste. One end of the tube is
a blind end (sealed with a small soldered brass disc). This part of the tube
projects 11.5mm above the top surface
of the Magic Candle.
2mm thick, 10mm OD
O-ring
Heatshrink sleeve
Photo 7: a view of the control board from the opposite side compared to Photo 6, so you can see the microphone tube and
insert. It’s a good idea to add an O-ring to acoustically isolate the microphone capsule from the Candle body.
84
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
A 14.5mm length of the tube projects inside the Candle body; 3mm of it
there has an external metal sleeve. This
helps it retain the heatshrink sleeve.
The four wires from the two thermistors exit this end of the sensor tube.
The wire end of the tube is sealed with
two pieces of heatshrink sleeving, to
retain the thermal paste.
I found that the 1/8in brass tube was
a good press fit into a 1/8in hole drilled
in the top disc, because the particular
tube measured 3.2mm even though
1/8in is officially 3.175mm.
Model/hobby shops and eBay sellers stock this standard brass tube made
by K&S Metals. All the 14-thou-thick
(~0.3556mm) wall brass tubes they
make are designed to slide inside each
other. Sometimes they can be a firm
fit and press together well. In the case
that they are a little loose, they are easily soldered together, or glued if heat
transfer is not required (such as with
the microphone mount).
Given that the internal diameter of
the brass tube is 2.46mm, it could have
been a little tight to pass the RS 151221 thermistor wires through the tube.
However, the parts I received were a
little smaller than their 2.4mm stated
maximum diameter (just under 2mm)
and were slightly ovoid.
They come with bare wires. I used
insulation taken from some thin insulated Kynar wire-wrap wire to insulate their bare wires. You can use any
insulation tubing, but it must be very
slim or you won’t get everything into
the brass tube.
The alternative RS thermistors specified are only 1.3mm in diameter,
leaving more room for the lead wires.
It would not be wise to increase the
diameter of the brass tube for a larger
thermistor, because that will increase
the thermal inertia and delay the overall response to heating and cooling.
When assembling the sensor tube,
the thermal paste is introduced with
a syringe and blunt needle. Insert the
needle until it reaches the blind end
and withdraw it as you slowly push the
paste into the tube. Since each thermistor has the same length leads, it is easy
to tell when they are ~25mm apart.
The 11-turn coil that forms R26 can
then be wound over that heatshrink
sleeve (see Photo 6). Once the coil is
wound, a small amount of epoxy resin
can be used to keep the turns in place,
or another piece of heatshrink sleeve
could be placed over it.
1 double-sided PCB coded 05112251, 68.5 × 25.5mm
1 6V 5W MES base automotive lamp (LAMP1)
1 MES lampholder (LAMP1) [Jaycar SL2659]
1 mercury tilt switch (S1)
1 electret microphone insert with pins (MIC1) [Jaycar AM4011]
2 5kW NTC thermistors (NTC1, NTC2) [RS 151-221 or 484-0133]
1 200mm length of 32 AWG (0.2mm diameter) enamelled copper wire (R26) OR
1 300mm length of 30 AWG (0.25mm diameter) enamelled copper wire (17 turns)
4 0.9mm PCB pins
3 alkaline C cells
Hardware & wire
1 241mm length of 33mm outer diameter, 27-30mm inner diameter (“25mm”) PVC
pressure pipe [Holman PVP2512-1, Bunnings I/N 4770100]
1 25mm press-fit PVC pipe end cap
1 ~91mm diameter, 8mm-thick plastic disc (eg, white acrylic)
1 ~81mm diameter, 8mm-thick plastic disc (eg, white acrylic)
1 ~66mm diameter, 8mm-thick plastic disc (eg, white acrylic)
1 34mm length of 1/8in (3.2mm) diameter brass tube,
14 thou (0.35mm) wall thickness, 2.46mm inner diameter [K&S Metals #8127]
1 3mm length of 5/32in (4mm) diameter brass tube,
14 thou (0.35mm) wall thickness, ~3.2mm inner diameter [K&S Metals #8128]
1 18mm length of 7/32in (5.5mm) diameter brass tube,
14 thou (0.35mm) wall thickness [K&S Metals #8130]
1 4mm length of 1/4in (6.35mm) diameter brass tube,
14 thou (0.35mm) wall thickness [K&S Metals #8131]
1 4mm length of 9/32in (7.15mm) diameter brass tube,
14 thou (0.35mm) wall thickness [K&S Metals #8132]
1 4mm length of 5/16in (7.9mm) diameter brass tube,
14 thou (0.35mm) wall thickness [K&S Metals #8133]
1 brass end plug, 2.46mm diameter
1 50 × 50mm piece of 16 thou (0.4mm) thick brass sheet
1 225mm-long, 1/4in (6.35mm) wide, 25 thou (0.65mm) thick brass sheet
1 small tube of non-conductive thermal paste (eg, MX-4)
1 10mm outer diameter, 2mm-thick O-ring
1 small spring to retain C cells (eg, taken from a C cell holder)
various lengths and colours of medium-duty hookup wire
various fasteners (machine screws, nuts etc) and a small cable tie
heatshrink tubing
Semiconductors
1 LMC7660 or MAX1044 switched-capacitor voltage inverter, DIP-8 (IC1)
1 OP495 quad rail-to-rail op amp, DIP-14 (IC2)
2 ZTX951 60V 4A PNP transistors, TO-92 (Q1, Q3)
2 2N7000 N-channel Mosfets, TO-92 (Q2, Q4)
7 BAT46 small-signal schottky diodes, DO-35 (D1-D7)
Capacitors
1 470μF 6.3V/10V electrolytic
3 47μF 16V tag tantalum
1 10μF 16V tag tantalum
1 1μF 63V/100V MKT
1 100nF 63V/100V MKT
1 100nF 63V/100V MLCC
Resistors (all ⅛W ±5% axial unless noted)
1 10MW
1 10kW
1 4.7MW
1 8.2kW
1 1MW
1 4.7kW 1%
2 750kW
1 4.3kW 1%
6 100kW
2 270W
1 51kW
1 120W
2 47kW
1 0W
2 22kW
siliconchip.com.au
Australia's electronics magazine
Parts List – Magic Candle
December 2026 85
Photo 9: the completed top cap
without the PCB attached.
Photo 10: here you can
see how the lamp base
attaches to the top cap.
Photo 11: the
finished assembly, ready
to insert into the Candle tube.
Photo 12: I got the battery spring
from a C cell holder.
The thermal conductivity along the
tube is dominated by the brass tube,
not the paste within it. The paste
helps with the thermal coupling of the
thermistor bodies to the inner wall of
the brass tube. An air gap here would
not be helpful because air has only a
fraction (0.3%) of the thermal conductivity of thermal paste.
Microphone tube
The microphone mounting tube
(Fig.11) is inserted into a 5.5mm hole
in the top disc until 2mm of it projects
above the disc surface. It is a press fit
because the outer diameter of the brass
tube is 5.56mm (7/32in).
Glue can be used if required. The
three additional short sections of tubing can be glued to it or soldered if they
are both brass.
This creates a way to mount the
microphone with a short section of
heatshrink tubing. Since the diameter of heatshrink tubing doesn’t quite
get small enough to get a good grip on
the brass tubing, I added a cable tie. I
also added a rubber O-ring (see Photo
7). This way, the microphone is not
too sensitive to a knock on the candle’s housing.
Photos 9 & 10 show the top plate
with the lamp, thermal tube and
microphone tube fitted to it. Photo
11 shows the same assembly with the
PCB attached.
Battery connections
For the battery’s negative terminal,
I made a ring of PVC taken from a tap
fitting from Bunnings (Fig.12). However, there are other suitable materials.
Photo 13: a view of the top battery
contact from the Candle tube’s top.
Fig.11: the microphone tube is
somewhat simpler. The O-ring helps
isolate it acoustically from the body
of the Candle.
86
Australia's electronics magazine
Silicon Chip
I made it a firm press fit. When it was
in position, I drilled a 0.9mm hole
through the side wall of the tube and
pressed in a 1 × 10mm-long pin to
retain it.
Since it was PVC, it could have
been PVC-glued into position in the
PVC pipe, but that would be permanent and could be messy. Still,
the battery contact plate, spring and
screws could be undone if any repairs
were needed even with the PVC ring
glued in position. Photo 12 shows the
battery spring in place, taken from a
C-cell holder.
Photo 13 shows the negative battery
terminal assembly fitted to the tube.
The retaining pin can be seen projecting into the central hole a little, just
below and to the right of the negative
(black) wire as it passes up through
the hole. The strap that connects the
positive end of the battery to the PCB
was shown earlier, in Fig.7.
Once the Magic Candle is complete,
remember to invert it briefly before
trying to light it. Holding it inverted
for a few seconds should cause it to
SC
switch on.
Fig.12: how I made the battery
negative terminal assembly.
siliconchip.com.au
Silicon Chip
Binders
REAL
VALUE AT
$21.50*
PLUS
P&P
Are your copies of
Silicon Chip getting
damaged or dog-eared
just lying around in a
cupboard or on a shelf?
Can you quickly find a
particular issue that you
need to refer to?
Keep your copies
safe, secure and
always available with
these handy binders
These binders will protect your
copies of S ilicon C hip . They
feature heavy-board covers, hold
12 issues & will look great on your
bookshelf.
H 80mm internal width
H Silicon Chip logo printed in goldcoloured lettering on spine &
cover
Silicon Chip Publications
PO Box 194
Matraville NSW 2036
The Magic Candle shown just a little smaller than actual size. It provides a soft
light suitable for setting the mood.
siliconchip.com.au
Australia's electronics magazine
Order online from www.
siliconchip.com.au/Shop/4
or call (02) 9939 3295
and quote your credit card
number.
*see website for delivery prices.
December 2026 87
SERVICEMAN’S LOG
Working with switch-mode power supplies
O
ften I work on cancer treatment machines that use a
couple of 16-port managed network switches. These
switches usually function perfectly for over a decade, but
occasionally one fails.
A colleague came up with the idea of removing the faulty
switch-mode PSU (SMPS) and replacing it with one from
an older, unused machine if available. This avoids the hassle of waiting for a new switch to be ordered, delivered
and reconfigured.
Replacing the PSU is a quick and easy fix. It involves
removing four screws from the cover and the four securing the PSU.
I’ve always been curious about how these ubiquitous
devices work. I’ve replaced many SMPSs over the years
because they’re considered field-replaceable and never
repaired. Recently, another SMPS failed and was replaced.
This time, I decided to try repairing one myself.
I did some internet searches on SMPS theory to refresh
my understanding. A few years earlier, I’d delved deep into
power electronics while fixing an old split system air conditioner that no one else could repair. I found the recent
series of articles in Silicon Chip on the subject (Power Electronics; siliconchip.au/Series/452) particularly interesting.
It was a shame they hadn’t been published five years ago!
I eventually repaired the air conditioner, which only
finally gave up the ghost a few months ago due to a gas leak.
The next challenge was finding a circuit diagram for the
SMPS. It turns out that this particular SMPS is used in a
wide range of devices. I managed to get a circuit diagram
that, while not exactly the same as my PSU, was close
enough to help me figure out what was happening.
The SMPS occasionally powered on OK, but not very
often. I began by checking the input-side components. The
bridge rectifier output was around 300V DC, which was
tapped off using a couple of high-value resistors to power
the controller chip. Apparently, once the device is working correctly, the chip receives power from a rectified secondary winding in the transformer.
I wasn’t sure why this was the case, but since the device
operates at some kHz, the secondary supply might produce
a smoother voltage.
Editor’s note: it’s likely to improve the power factor and
efficiency of the device.
I then checked the controller chip’s output pin and
expected to see a square wave at some multi-kHz frequency. However, the output frequency was surprisingly
low, which seemed odd.
Further reading up on SMPS theory revealed that when
the PSU had no load, it would enter power-saving mode. A
minimum load current is required to prevent it from going
into standby. I soldered an appropriate value resistor to the
output pads to draw current, but that didn’t help.
I removed and tested the Mosfet, rectifier diodes, snubber diode and a few other usual suspects. I had recently
bought a tweezer-style in-circuit tester but, being a bit oldschool, I wasn’t entirely convinced of the results.
For example, the output rectifier diodes were measuring
about 0.9V, which seemed a little strange. They were part
of a dual pack and, after checking the spec sheet, I discovered that they were fast recovery diodes with a forward
voltage in the range I was reading. Another mystery solved.
I also learned something interesting about Mosfets during
this process. You can switch them on out of circuit using
the diode test mode on your DVM. Apparently, you can
charge the gate with the DVM and it stays charged for a
Items Covered This Month
• Fixing the power supply in cancer treatment
machines
• Two HP laptop repairs
• Repairing a clock for preschool children
• A simple fix for flickering
• Dyson bladeless fan repair
Cartoonist – Louis Decrevel
Website: loueee.com
88
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
while, which gives you curious results when checking the
drain-source resistance. Sometimes the drain source was
over-range, sometimes short circuit.
Coincidentally, I stumbled across a YouTube video at
about the same time. It featured an electronics technician
explaining that you couldn’t trust component testers with
Mosfets. This memory came in handy when I was getting
strange results testing my Mosfet. After testing it correctly,
I realised it was fine and soldered it back into the PCB.
As you might know, these PSUs have a feedback loop to
control the output voltage. Feedback systems can be tricky
to fault-find. I was aware that there are various techniques
to diagnose feedback systems, but hadn’t delved into them
much before.
One approach I considered was to break the loop and
feed an external control voltage into the control chip, then
fault-find the system linearly afterwards.
Since the PSU only operated correctly on occasion, I
couldn’t get readings for normal voltages in the circuit. I
wanted to break the feedback loop at the opto-coupler and
feed an appropriate voltage forward from there. However,
not knowing an appropriate voltage, I decided to just wing
it and removed the opto-coupler.
I switched the PSU back on, and this time it was working. I was instantly reminded of one formula: P = I2R. The
load resistor had gone up in smoke! I’d used a 1W resistor, but its value was low enough to result in a dissipation
much higher than that. I had some low-value 5W resistors
lying around, so I soldered in a replacement.
It occurred to me that the unregulated output voltage
might not be a small percentage higher than the nominal voltage, as I’d expected. It could be quite a lot higher,
explaining why the original output resistor had fried. The
5W resistor might not have survived the excess voltage
either, but I might have had enough time to switch off the
PSU before it fried as well.
Before continuing, I decided to solder a flying lead to the
transformer pin connected to the Mosfet drain so I could
attach an oscilloscope probe while checking other circuit
parts with a DVM. However, I noticed that the solder on
the pad didn’t flow as expected. It remained grey, not turning silver when melted as anticipated.
This made me wonder about dry joints. I then desoldered
all the transformer pins and the solder on each behaved
identically.
I was confident I’d nailed it. I resoldered all the joints
with fresh solder and powered the PSU back on. A steady
5V output appeared, and now it powered back on reliably
every time I switched it off and on again.
The switch drew a lot less than the 6A maximum rating current of the PSU, and the case was ventilated, so I
doubted the transformer was getting too hot. However, it
may still have been hot enough to compromise the solder
joints over many heat cycles. Who knows?
The rest of the components’ solder joints that I looked at
during my troubleshooting all flowed as expected, including the power Mosfet driving the transformer.
I suspect every PSU of this type would fail in the same
way. As with many pieces of equipment, electrical, electronic, mechanical etc, there are common failure modes.
Once known, repairs become straightforward. Whether it’s
practical to repair is another question.
Since the repaired switch won’t be going back into a
siliconchip.com.au
machine, it’ll likely end up with my son, a programmer
with a few PCs he’s about to network together.
Overall, it’s always satisfying to fix something yourself
and learn something along the way.
Malcolm Batchelor, Parramatta, NSW.
HP dv5 laptop repair
My son picked up this HP dv5 laptop at the tip shop a
few weeks ago. I could see it needed a lot of work but the
potential was there. When I first tested it, I found that it
worked but the keyboard had three missing keys. I obtained
a replacement keyboard via eBay, as I didn’t have one in
my salvaged keyboard collection.
It came with an Intel P8400 processor at 2.26GHz, but
I could upgrade it to an Intel P8600 processor at 2.4GHz.
The CR2032 cell that keeps the CMOS data and the time
and date current was under a cover on the back of the laptop, so it was easily accessible. More often than not, I need
to fully dismantle the laptop to replace the cell, which I
have done many times.
I would be dismantling this laptop anyway, to clean the
CPU heatsink and fan, and upgrade the CPU. With the laptop apart, I removed the motherboard, then detached the
heatsink and fan as an assembly. This laptop has a dual
heatsink for the CPU and GPU.
The fan is attached to the heatsink with three tiny screws
and a piece of tape. With it removed, I could see that both
heatsinks were about 20% clogged, which isn’t too bad.
Of greater concern was that the heatsink compound had
dried out completely, which is also common with these old
laptops (it’s 18 years old). The motherboard will support
an Intel P9600 CPU at 2.53GHz, but I did not have one, so
I dropped in the P8600, which isn’t far behind.
I reassembled the laptop and then fitted the replacement
keyboard. This laptop came with 4GB of PC2 RAM and a
320GB hard drive, which I will reuse. I don’t know what’s
on the hard drive; I’ll just wipe it when I install Linux.
Installing Linux, I realised there was a problem. When
Australia's electronics magazine
December 2026 89
The HP dv5 laptop (shown above) that Bruce Pierson picked up from a tip shop. The photos opposite, at lower right, show
another HP-branded laptop that had a broken case.
I was setting the time and checking other settings, I could
see that there was a blemish in the screen that I had not
noticed when I’d done a quick test before dismantling the
laptop. It looked like something was inside the screen, even
though the full picture was there. Time to take the laptop
apart again to replace the screen.
It’s easy to remove the lid as it just entails removing the
panel above the keyboard, the keyboard and the wiring.
Then the hinges are easily accessible without having to
remove the top shell, so it’s not much of a job to get the lid
free. With the lid off, I removed the two screws, pulled off
the screen surround, removed the other screws and then
disconnected the screen.
It is different to most laptop screens as it has a separate
plug for the backlight. I looked through all the screens I
had in my electronics workshop and I came up empty. I
had some screens in two boxes down in the back shed from
laptops that I’d wrecked years ago, so I went down to look
through the boxes, hoping to find a compatible screen.
90
Silicon Chip
Nothing in the first box, and I was almost to the bottom
of the second box when I found two screens that looked
like they were from the same family as the faulty screen.
It happened that one was identical, with the same brand
and the same part number as the faulty screen.
The other screen was a different brand, but it looked
identical also, with the same 30-pin connector for the
screen and the separate plug for the backlight. I tested both
screens and they were both good. I selected the screen that
was a different brand as it was slightly brighter than the
identical screen.
The next day, I started installing the new screen and I
found that the screws I had neatly arranged on the workbench when I dismantled the lid were mostly not there
now! I looked on the floor and found several of them, but
some were missing. It looks like I had an overnight visitor
who messed up my system.
I suspect that it was a green tree frog, as I previously saw
one on the floor inside my electronics workshop. It must
have jumped up on the workbench and sent the screws
flying. As I was missing several screws, I had to search
through a container of screws to find replacement screws
of the correct size. Finally, I could reassemble the lid, refit
it and finish assembling the laptop again.
I decided to try Big Linux. It has a KDE Plasma desktop
that is highly customisable. The ISO installer file is 5.2GB,
so it needs to be burned to a double-layer DVD.
I booted up the laptop and now there was another problem. I could hear the DVD drive making grinding noises
partway through loading the files. I powered the laptop
off, removed the optical drive and looked through my box
of salvaged optical drives. I found an identical one with
the same faceplate.
However, when I tried to fit it, it would not engage the
connector, so I got another drive, changed the faceplate
and it fitted correctly. This time it loaded. Unfortunately,
after I installed it, I could only get a resolution of 640 × 480
pixels, even after updating the video drivers, so I’ll have
to try it again on a newer laptop later.
The best Linux distros for older hardware are Lubuntu,
Kubuntu and Linux Mint Cinnamon, which also have the
advantage of being more like Windows XP with their applications menu (start menu). I recently set my wife up with
Australia's electronics magazine
siliconchip.com.au
Linux Mint Cinnamon on a Sony Vaio laptop with an i7
processor and 8GB of RAM. She said it’s much faster than
her old Windows 10 laptop and she is very happy with it.
With so many older and even more recent laptops and
computers not supporting Windows 11, Linux can save
these laptops and computers. While Linux may not be
suitable for anyone with specific software needs, it’s fine
for use as an operating system for anyone who just does
email, web browsing, YouTube and online banking, which
is what my wife does.
Linux distros that are based on Ubuntu have a massive
library of free applications that will cover the needs of
most users.
So with a bit of work and some replacement parts that I
either had on hand or could get via eBay, this old HP dv5
laptop now has a new lease on life and that’s another one
saved from the junk pile.
I’ve found that there are many sellers on eBay, and likely
other online stores, who stock parts for these old laptops.
It is sometimes necessary to order parts from overseas but
a bit of searching usually turns up any required parts.
Bruce Pierson, Dundathu, Qld.
HP 15 Laptop repair
Continuing with another laptop; when I got this HP 15
laptop with an i7-4500U 1.8GHz processor, I tested it and
found that it worked, but the back right-hand corner of
the bottom shell was broken. It just happened that I had
a nearly identical laptop with an AMD processor that did
not work, so I was hoping that I could use the bottom shell
from it to repair this laptop.
However, when I dismantled the donor laptop, I found
that the bottom shell was also broken, with the nuts for the
hinges ripped out and some other damage, so I couldn’t
use it. However, the part that was broken on the otherwise
good laptop was still OK on the broken one, so it was just
a matter of removing it from the bottom shell.
siliconchip.com.au
Servicing Stories Wanted
Do you have any good servicing stories that you would like
to share in The Serviceman column in SILICON CHIP? If so,
why not send those stories in to us? It doesn’t matter what
the story is about as long as it’s in some way related to the
electronics or electrical industries, to computers or even to
cars and similar.
We pay for all contributions published but please note that
your material must be original. Send your contribution by
email to: editor<at>siliconchip.com.au
Please be sure to include your full name and address details.
I found that it was clipped on, but it had some small tabs
that had been melted over as well. I started prying it off the
bottom shell and it popped off without too much trouble,
with the melted-over tabs giving way easily.
I went through the process of dismantling the good laptop, and once I had everything out of the bottom shell, I
prised off the broken part so I could replace it. I gave everything a quick clean and then proceeded to attach the new
part, which clipped on easily and sat in place nicely. Now
it was time to reassemble the laptop and get it ready for an
operating system.
This laptop does not have a CR2032 cell to power the
CMOS, but it instead relies on the main battery to keep the
time and date correct. Fortunately, the battery in this laptop
is still good, unlike another I have that uses a similar scheme.
In that case, the time and date need to be reset whenever that
other laptop is powered up, which is rather inconvenient.
This HP 15 laptop is an entry-level laptop, with the i7
CPU soldered to the motherboard instead of being in a
socket. Also, the hard drive is inside the laptop with no
bottom access panel.
With the laptop put back together, it was time to install
Linux on it. I had previously installed Debian on a laptop
Australia's electronics magazine
December 2026 91
with an Intel i3 processor and found that it was a bit on the
slow side, so I would now install it on this laptop with the
i7 processor. The installation went smoothly, so another
laptop was saved, having a new life with Debian Linux.
Bruce Pierson, Dundathu, Qld.
Child’s clock repair
This clock is designed for preschool children who have
not yet learned to tell the time. It is shaped like a traditional wind-up alarm clock but sports a backlit digital display, which shows the time using a 7-segment display, and
large LCD icons showing a sun or a moon.
The backlight changes colour, illuminating yellow for
daytime and blue for nighttime. This means that even little kids can tell at a glance whether it is time to sleep, or
to get up and hassle mum and dad.
The clock has an internal 500mAh lithium-polymer cell
that is charged via a micro-USB socket, using the ubiquitous TP4056 charge controller IC. The battery typically
lasts for a week or two between charges.
This particular clock has been in service for almost three
years, but was no longer charging reliably. It would only
charge if the micro-USB cable was held at a very specific
angle. Several different USB cables were tried, but the
problem remained.
Unlike many modern consumer devices, this one was
easy to open, with a handful of Phillips screws securing
the front and back mouldings of the clock in a clamshell
design. The main PCB and battery were mounted on the
rear moulding, with the large LCD screen and backlight on
the front, joined by a conductive rubber strip.
The charging socket was all alone on a tiny PCB ‘daughterboard’ soldered at right-angles to the main PCB. Close
inspection showed that the power pins of the micro-USB
socket had fatigued loose from the daughterboard.
I desoldered the daughterboard, then used two strips of
desoldering braid as bridges between the socket’s power
pins and the daughterboard. I also built up a thicker fillet
of solder around the metal body of the socket, to try to stop
it from flexing in the future. I then re-attached the daughterboard to the main PCB with more generous fillets of solder. Now the clock charges reliably again.
Nicholas West, Canberra, ACT.
A simple LED light bulb repair
Some time ago, I purchased two 4W LED globes with
Edison screw style bases. They were very cheap as I bought
The clock with the front part of the enclosure and display
removed. The micro-USB daughterboard can be seen at the
bottom of the main PCB.
The USB daughterboard; you can see where the socket has
flexed and lifted the PCB tracks. The power tracks were
repaired using strips of desoldering braid.
92
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
them from an op shop for only one dollar each. Though
technically second-hand, they looked new, as they were
still in their original packaging. Having glass envelopes,
they were well-packed in bubble wrap and cardboard.
I unpacked one of them and installed it in my lounge. I
tested it and all seemed good. That night, the globe started
flickering; oh dear, that’s why they were ‘thrown out’. For
the remainder of the night, I put the old globe back in. This
confirmed it was the new LED globe and not the wiring or
switch at fault.
The next day, I decided to investigate further. I could
see the nipple of solder at the base of the globe was quite
corroded. This is meant to connect to the Active terminal
in the socket. I thought surely it wouldn’t be that simple.
I very carefully filed the corrosion off the solder and bent
the brass contact up a bit. I also decided to add a very small
amount of WD40 to the rivets and other contacts that make
this globe operate.
On testing, all was well, with no flickering. It wasn’t the
fact that I got the globe to work properly that I was happy
about – it’s that it won’t have to be discarded for a lot longer now. The other globe also had some corrosion on it;
maybe they were very old stock, or got damp in storage.
Peter Eichhorn, Heathcote, Vic.
Dyson AM02 bladeless fan repair
This fan was originally released in 2010 at a price of
US$450. When I received it from Ewaste Connection in
Brisbane, it showed no sign of life. After taking off the
base by twisting it and releasing a catch, you gain access
to the circuit board.
Mains power is connected directly to the board, and the
AC power goes via a line filter and fusible link resistor to
a full-wave rectifier, then a large capacitor. The voltage on
the capacitor is about 340V DC, ie, the peak mains voltage.
Connected to the capacitor is an LNK340DN switching
regulator IC in a small surface-mount package (SO-8C) to
provide a 15V DC supply. This IC has its own internal
high-voltage FET. The 15V is used by a 24-pin SMA6863
brushless motor drive IC for its logic circuitry. The
SMA6863 has three Mosfet-based half-bridges to drive the
fan motor, all powered by the 340V DC supply.
The large 6-pin IC on the wiring side is a MOC3082, an
optically isolated Triac to drive the rotational motor.
A 5V three-terminal regulator, fed from the 15V supply,
supplies an ATtiny681 20-pin surface-mounted microprocessor. This IC controls the fan by reading the fan input
controls or a remote, and uses its outputs to drive the
SMA6863.
Measurements showed that the 340V DC supply was
OK, but there were no 15V DC or 5V DC rails. By using an
external floating 15V, with its negative to the 340V supply
capacitor, I found that the 5V regulator is working, and
the fan is otherwise functional. I had to disconnect the
LNK304’s inductor to avoid feeding the 15V DC back to it.
I just have to get a replacement LNK340DN to complete
the repair.
During this diagnosis, I had trouble reading the IC part
numbers due to the conformal coating on the board. I had
to scrape this off the ICs to read the part numbers. A solvent capable of removing the coating would be useful, as
long as it does not remove the part number as well.
SC
Roger Sanderson, Figtree Pocket, Qld.
siliconchip.com.au
Australia's electronics magazine
December 2026 93
Vintage Radio
STC Bantam model A141 mantel set
by Ian Batty
Thanks to its
quirky design,
this radio’s
original name
(implying a
small chook)
gave way to
one of the
mobile homes
that began
appearing in the
holidays in the early 1950s: “the caravan”.
I
t’s smaller than AWA’s similar
500M that I reviewed in September
2023 (siliconchip.au/Article/15947),
but more compact and with a larger,
more easily read dial. It’s also a bit different electrically, with an improved
converter and a high-gain output
valve.
Like the AWA, it only uses three signal valves (converter, intermediate frequency [IF]/demodulator and output),
so it’ll be interesting to find out how
it performs in comparison.
Rather than the 500’s 6A8 pentagrid
converter, it uses the triode-hexode
6K8G. It also substitutes the 6AG6 output pentode for the AWA’s 6V6.
Pentagrid converters were widely
adopted by American engineers. While
they could not perform well at higher
shortwave frequencies, their simplicity of manufacture saw them dominate the All American Five broadcast
receiver design.
Triode-hexodes & triode-heptodes,
more widely used in English and particularly European designs, offered
superior performance on shortwave
bands.
94
Silicon Chip
The 6K8 is a triode-hexode (see
Fig.1). Most triode-hexodes apply the
signal to grid 1 and the LO (local oscillator) to grid 3: this is outer-grid injection. Unusually, the signal is injected
into the 6K8’s grid 3. The 6K8’s separate triode (local oscillator) section is
unaffected by AGC action and is more
stable. It is also optimised for oscillation, making the 6K8 more reliable at
shortwave frequencies.
Fig.1: the 6K8 triode-hexode and
its ‘unusual’ construction.
A converter’s potential gain (Av)
depends on the product of its conversion transconductance (gc) and its
plate resistance (rp), with the 6A8’s rp
at 350kW and the 6K8 at 600kW.
Such high values for rp might appear
irrelevant in untuned circuit designs,
with load resistances in the low hundreds of kilohms, but IF transformer
tuned circuit impedances can reach or
exceed half a megohm, so a low converter plate resistance would potentially limit the overall stage gain.
By using the formula Av = gc × rp,
we can calculate the potential stage
gain. In other words, the gain is the
product of the transconductance and
plate resistance.
Such calculations give the 6A8 a
maximum possible stage gain of 198,
or 210 for 6K8. Be aware that this
implies a first IF transformer with a circuit impedance of well over 50MW, so
we’ll have to wait for the test results to
see whether the 6K8 really gives better
performance than the 6A8.
The 6AG6 is a European EL3-series
high-gain output valve, first released
by Philips in 1937.
siliconchip.com.au
We can divide output tetrodes/pentodes into three classes. The 6F6/42
designs offered a mutual conductance
(gm) of around 2mA/V, needing a grid
bias of about -20V. The 6L6/6V6 class,
with mutual conductances around
4-5mA/V, operated with around -12V
bias.
The EL3/6M5 class drew more
heater current than their predecessors;
as much as 1.2A for the initial EL3.
They gave a mutual conductance of
around 10mA/V with a bias voltage of
about -6V. This high mutual conductance allowed designers to remove the
audio preamplifier, feeding the signal
directly from the demodulator to the
output stage grid.
This class was developed into
duo-diode/output pentode designs
such as the locally produced 6BV7.
With a gm double that of the 6V6, we
might expect a set that’s overall more
sensitive than the AWA; we shall see.
Standard Telephones and
Cables (STC)
Standard Telephones and Cables
was founded in 1883 in London by
the US Western Electric Company
and incorporated as a British legal
entity in 1910. In 1925, Western Electric divested itself of all foreign operations, leaving STC as a wholly British
corporation.
As a radio manufacturer, STC joined
a consortium that set up the British
Broadcasting Company (BBC) in 1922.
In 1926, the company opened its
first Australian factory in Myrtle
Street, Chippendale (near the Sydney CBD), initially producing STC
telephones and radios. STC rapidly
The SILLIAC contained 2800
vacuum tubes, weighed nearly five tonnes
and was estimated to cost around £75,000 in the 1950s.
expanded, manufacturing radio broadcast transmitters, telephone repeater
coils and other types of telephone
equipment.
By 1935, the Chippendale factory
was unable to cope with the company’s rapidly expanding business.
Land was acquired in nearby Alexandria, and a new 2790m2 factory was
established. By 1939, they employed
500 people and opened a new factory
in Villawood, 27km west of Sydney.
STC’s work in multi-channel and
microwave transmission (among other
things) saw it become a leader in military electronics during World War II.
By 1945, the company had expanded
its manufacturing to cover just about
everything related to radio: train telephone equipment, multi-channel
telephone carrier equipment, PA systems, electronic telephone repeater
valves, high-power radio transmitting
valves, high-power rectifiers, UHF
aircraft beacons and a variety of telecom cables.
They also played an important part
in developing electronic communications in New Zealand.
SILLIAC (Sydney version of the Illinois Automatic Computer) was the
first computer built in an Australian
university. The size of a double-decker
bus, it contained 2800 vacuum tubes,
was programmed with paper tape and
drew 25kW in operation.
A version of the STC Bantam A141 of questionable provenance, and the rear of cabinet of the set shown in the lead photo.
siliconchip.com.au
Australia's electronics magazine
December 2026 95
This is not a safe mains cord retention method – don’t do it! Nor is it safe with
the twin core mains wire, lacking a mains Earth connection to chassis. A three
core mains cord should be used instead with mains Earth connected to chassis.
operation? Like the 500M, the Bantam
uses the padder capacitor to create a
standing bias due to the oscillator’s
grid current. Neat!
The converter feeds first IF transformer IFT1. The tuning capacitors are
marked as 50pF – this is a lower value
than usual, and given that Q can be
given by Q = √L ÷ C, it’s possible that
the IF channel will show a higher gain
than that of the 500M, which uses the
more common 70pF values.
IFT1 secondary feeds the 6G8 IF
amplifier. This duo-diode pentode,
with a remote cutoff characteristic,
can be used either as a gain-controlled
RF/IF amplifier or a fixed-gain audio
amplifier.
With its cathode returned directly
to ground and sharing screen voltage
with the converter, it’s economical
and a good performer. The 6G8 pentode feeds the primary of the second
IF transformer, IFT2.
One diode acts as the AGC rectifier, fed from the 6G8’s anode via
100pF capacitor C10. R6 (1MW) feeds
the resulting voltage to the AGC line,
which also acts as a low-pass filter in
conjunction with 50nF capacitor C3.
The AGC line gets standing (DC) bias
via divider R10/R5 from the back-bias
developed across the electrodynamic
speaker’s field coil. Audio filtering is
done by 250nF capacitor C3.
The second diode is fed by the secondary of IFT2, and rectified audio is
filtered by 350pF capacitor C11, and
developed across 500kW volume control potentiometer, R7.
SILLIAC was designed by Professor Harry Messel of Sydney University’s Physics Department, based on
the ILLIAC and ORDVAC computers
developed at the University of Illinois.
In July 1954, STC was contracted to
build it, with testing and installation
performed by technicians within the
School of Physics. This marked the
start of computer science and IT in
Australia.
It was finally switched off after 12
years of operation, in 1968. Parts of
it are preserved in the Powerhouse
Museum in Sydney.
In 1987, STC was purchased by
Alcatel NV, and a significant chapter
of our technological and scientific history ended.
Where pentagrids must use (at least)
grid 1 as the oscillator grid, hexode/
heptode designs may use the innermost grid 1 for signal, then grid 2 as a
screen, with grid 3 as the local oscillator (LO) injection grid. Alternatively,
grid 1 can be used for LO injection and
grid 3 for the signal. For more details,
see the Radiotron Designer’s Handbook, F. Langford-Smith, 4th edition,
1953, pages 973-986.
The 6K8 triode uses the LO transformer, L3/L4, to provide feedback
and tuning. As tuning gang C4/C5 uses
identical sections, tracking is provided
by fixed padder capacitor C7 (348pF).
There’s no coupling capacitor in the
circuit, so how does the 6K8 oscillator
get the bias need to keep it in Class-C
Circuit description
A good view of the chassis, with a new cordgrip grommet in place. Later on I
refitted this with a three-core mains lead to make sure the chassis was Earthed.
Like the 500M, the Bantam eliminated individual cathode biasing for
the converter and IF amp, relying on
a standing bias derived from the backbias circuit. Also like the 500M, the
converter and IF amplifier stages share
a common screen supply (see Fig.2).
Antenna transformer L1’s primary is
shunted by 100pF capacitor C1, presumably to improve gain at the lower
end of the band. It also uses top coupling via C2, which also serves as a
high-end adjustment for antenna-circuit alignment.
Unusually, it’s marked as 0~25pF.
You’d expect to see a minimum value
of a few picofarads but, given that it’s
described as a wire trimmer, maybe
you can adjust it to 0pF of capacitance.
The tuning gang’s specification is also
unusual at 10~346pF.
96
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
The audio from the wiper of R7 feeds
via 10nF coupling capacitor C12 to the
6AG6 output valve grid, with stopper
resistor R9 (50kW) inserted to prevent
parasitic oscillations in the high-gain
6AG6.
Output transformer T1 is kept away
from resonance by 10nF capacitor C14.
Be aware that this should be connected
across T1, so that if it goes short circuit, it doesn’t have the whole HT+
supply across it. Restorers have told
me of finding burned-out transformers
due to this very problem.
Editor’s note: most of the component labels
have been renumbered compared to the
original circuit.
Power supply
The power supply is what you’d
expect: a mains transformer with two
input tappings for low or normal Australian voltages, a full-wave rectifier
and a pi filter made using two electrolytic capacitors (C16/C13) and the
speaker’s field coil as the inductance.
Unlike the 500M, but like the
Mickey Oz, the field coil is in the supply negative, offering a convenient
negative bias voltage for delayed AGC
and removing the HT voltage from the
field coil, avoiding the possibility of
insulation breakdown.
A problem with my set
The Bantam, like its contemporaries, uses a metal chassis, moulded
Bakelite sockets and point-to-point/
tag-strip wiring. It’s about as easy to
work on as most other sets of the era.
My Bantam had been previously
fixed to a “competent” level of workmanship. I put competent in quotes
because the mains cord had been fed
through a grommet and tied in a knot
– see the photo at upper left. This is an
unsafe practice, as the cord can readily
twist in the grommet and break free.
Once it was fully working, I did fit
a three-core mains lead, which supplied an Earth connection and made
the set safe – see my article article on
the Astor Mickey Oz (January 2022;
siliconchip.au/Article/15179) for the
complete discussion.
Model variations
Radiomuseum (www.radiomuseum.
org) lists four- and five-valve versions
running from 1946 (141) to 1961 (415).
The original “caravan” was a distinctive design and the end-of-line version
is pretty eye-catching too.
There is also the super-rare
“reflected” version. A mate of mine
who lives in Glenelg, Adelaide, sent
Fig.2: the model A141 Bantam circuit uses a 6K8 triodehexode (oscillator & mixer), a 6G8 duo-diode-triode (IF
amplifier, AGC, demodulator), a 6AG6G pentode (audio
output) and a 6X5GT dual rectifier valve.
siliconchip.com.au
Australia's electronics magazine
December 2026 97
STC Bantam variants
Model
Converter
Audio preamp
Audio output
Biasing
141
6K8
6G8
6V6
Back
A-141
6K8
6G8
6AG6
Cathode
B-141
6K8
6G8 (reflexed)
6V6
Back
C-141
6A8
6G8 (reflexed)
6V6
Back
D-141
6A8
GG8 (reflexed)
6V6
Back
me the photo on page 95. At first, I
thought he was pulling my leg, but he
told me to look at the dial.
You’ll have to make your own mind
up, but I reckon the dial on Neven’s
radio is printed in the opposite direction, with the highest frequencies
(3AK/3XY/5AD/5AU) fully clockwise!
It isn’t just a mirror image, as that
would make the STC logo and all the
station lettering reversed.
How good is it?
It’s better than I expected. Although
electrically near-identical to AWA’s
500M, its performance is outstanding
for a set with only three active stages.
Lacking an audio amplifier or reflexing, the 500M came in with sensitivities (for 50mW out) of around 1mV at
the antenna terminal.
The Bantam gave the standard
50mW for 400Hz at 30% modulation
as little as 80μV input, making it more
than ten times as sensitive as the 500M.
The 6AG6’s higher gm improved audio
sensitivity by about 50% but that still
leaves a lot to make up. So let’s look
at the converter.
The 500M managed a best gain, from
antenna to the grid of the IF amp, of
around 60 times; remember that the
500M’s passive antenna circuit alone
gives a gain close to four times. The
Bantam’s entire antenna/converter
circuit shows a best figure of 400+,
remembering that the passive antenna
circuit (comparable to the 500M) gives
a gain of around 4.3.
So it looks like the 6K8’s characteristics really do give much better conversion gain (around 100 times) compared to that available from a 6A8 pentagrid, as long as the circuit is suitably
optimised for it.
Calculations, using the reactance of
the IFT’s 50pF capacitor at 455kHz,
indicated a reactance of some 7kW.
With a circuit Q of at least 200, implying a primary impedance of around
1.4MW. Considering the 6K8’s higher
plate resistance compared to the 6A8,
combined with better circuit design,
a conversion gain of 100 times seems
easy to achieve.
The set’s sensitivity for 50mW output is 80μV at 600kHz and 90μV at
1400kHz, with a signal-plus-noise-tonoise ratio (S+N:N) of about 18dB in
both cases. To achieve S+N:N figures
of 20dB, I increased the signal levels to
105μV and 120μV, respectively.
Its RF bandwidth is ±2kHz for -3dB
down or ±35kHz for -60dB. AGC permits a 40dB input range for an output
change of 6dB. It would not overload
even with a 1V (1000mV!) input.
The maximum audio output is
2.5W at 10% total harmonic distortion (THD). At 50mW, THD is 1.5%,
The unusual
knobs mean
the grub
screw must
clamp
against the
rounded
part of the
shaft, not the
flat section.
and at 10mW, it’s 2%. The frequency
response from the antenna to the
speaker runs 155Hz to 1.9kHz; from
the volume control to the speaker, it’s
from 150Hz to over 3.2kHz.
In conclusion, this radio is hardly
inferior in performance to many fivevalve sets with the usual converter/
IF/demodulator/audio preamp/audio
output lineup.
Special Handling
It’s a heavy set for its size due to the
Bakelite cabinet. The 6AG6G is a tall
valve and only just fits into the cabinet. I advise careful inspection to pick
up any deterioration caused by heating. It’s also best to resist just slipping
fingers under the top and lifting from
the centre of the cabinet.
The control knobs were a puzzle.
They came off easily but would not go
back on at all. A knob’s fixing screw
usually presents to the flat on the control shaft. The Bantam’s knobs include a
brass insert that drives against the shaft.
Opposite the insert, where the round
portion of the shaft would normally
mate to, the knob has an internal flat.
So, for this fitting, the flat in the
knob must mate to the flat on the shaft,
meaning that the screw presses the
brass insert against the rounded side
of the shaft – as shown in the photo
SC
above.
The top and front view of the STC Bantam’s chassis. See the table above for the slight differences
in valve lineup between the various model 141s.
98
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
DCC Model Railway Kits
DCC is a great way to control model trains, and we have a series of projects that can be used in a
model train layout from a DCC Decoder to a Destination Display (siliconchip.au/Series/455).
DCC Base Station
SC7539 Kit ($90 + P&P) – includes all parts except
the case, power supply and optional headers
DCC Remote Controller
SC7552 Kit ($35 + P&P) – includes all parts except the case and cabling
DCC Accessory Decoders
Snap-type Kit (SC7685, $40 + P&P)
Servo-type Kit (SC7686, $40 + P&P)
I2C Controller
SC7690 Kit ($30 + P&P) –
for DCC Accessory Decoders
DCC Booster and
Reverse Loop Controller
Kit (SC7579, $45 + P&P) – includes all parts, except the
case, OLED screen, power supply and front panel
Stepper Motor Driver
SC7601 Kit ($30 + P&P)
DCC Decoder
SC7524 Kit ($25 + P&P)
Destination Display
SC7697 ($22.50 + P&P) –
does not include the screen
(available separately)
siliconchip.com.au
microDCC Decoder
SC7617 Kit ($25 + P&P) –
choice of a bell or whistle sound for the software
Australia's electronics magazine
December 2026 99
Subscribe to
The VERY BEST DIY Projects
!
Built-in Wi-Fi (NTP) suppor
t»
62 addressable RGB LEDs »
Can be configured remotely
»
NeoPixel
RGB LED
Analog Clock
Portable
EMF Probe
Ideal Bridge
Australia’s top electronics magazine
Silicon Chip is one of the best DIY electronics magazines in the
world. Each month is filled with a variety of projects that you can
build yourself, along with features on a wide range of topics from
in-depth electronics articles to general tech overviews.
If you have an active subscription you receive 10% OFF
orders from our Online Shop (siliconchip.com.au/Shop/)*
Rest of
World
New
Zealand
Australia
* does not include the cost of postage
Length
Print
Combined
Online
6 months
$77.50
$87.50
$55
1 year
$145
$165
$105
2 years
$270
$305
$200
6 months
$95
$105
1 year
$180
$200
2 years
$335
$370
6 months
$115
$125
1 year
$220
$240
All prices are
in Australian
dollars (AUD).
Combined
subscriptions
include both
the printed
magazine and
online access.
2 years
$410
$445
Prices are valid
for the month
of issue.
Try our Online Subscription – now with PDF downloads!
RectifIers
for Power Ampl
ifiers
NOVEMBER 2026
ISSN 1030-2662
11
9 771030 266001
$15 00* NZ $15 90
INC GST
INC GST
Published in
Silicon Chip
Portable
EMF
Probe;
November
2026
RGB LED ‘Analog’
Clock; NovemberDecember 2026
Mighty USB-C Power
Supply; OctoberNovember 2026
An online issue is perfect for those who don’t want too much clutter around the house and
is the same price worldwide. Issues can be viewed online, or downloaded as a PDF.
To start your subscription go to
siliconchip.com.au/Shop/Subscribe
ASK SILICON CHIP
Got a technical problem? Can’t understand a piece of jargon or some technical principle? Drop us a line
and we’ll answer your question. Send your email to silicon<at>siliconchip.com.au
PicoSDR’s optional
screen is the 2.8in type
I recently completed the PicoSDR
project (April 2026; siliconchip.au/
Article/20079) and I would like to share
my experiences and thoughts on the
receiver. Firstly, I would like to congratulate Charles Kosina & Jon Dawson on
a successful job of designing a receiver
that performs exceptionally well.
There are two minor problems I
ran into. First, the LM386 oscillated,
producing a very loud tone that only
stopped with the volume turned right
down. I tried different lower-wattage
speakers without it making any difference. However, when using my 18W
headphones, a beautiful, rich sound
emerged. This pointed to the amplifier
chip being driven too hard.
Rather than replacing the speaker,
I opted to place a 10W ½W resistor in
series with the speaker. That stopped
the oscillation, and I still have plenty
of volume for comfortable listening.
The other problem was with the
optional 3.5-inch TFT screen. It illuminates with a white screen; I tried every
setting possible, but it wouldn’t initialise. I instead purchased a 2.8-inch
TFT LCD screen that has an ILI9341
driver, and it works perfectly, as the
attached photo shows.
Besides that, the PicoSDR has been
a fun project, and I highly recommend
it! (R. F., Holsworthy, NSW)
siliconchip.com.au
● It appears that the 2.8-inch screen
is the correct one to use. The model
number wasn’t specified in the parts
list supplied (in fact, it wasn’t even
mentioned in the list we received).
It appears we have incorrectly determined it was the 480×320-pixel 3.5inch screen with the ILI9488 controller
rather than the 320×240-pixel 2.8-inch
screen with the ILI9341 controller.
Since the resolutions and control
schemes differ, they are not interchangeable. You need to use the 2.8inch screen, and we’ve modified the
suggestions on our website shop to
account for that. The 3.5-inch screen
you purchased likely works fine if it’s
driven correctly.
Sound in DCC model
railway systems
I built the DCC Base Station (January
2026; siliconchip.au/Article/19558) in
February this year. It has been fantastic; it was so easy to program my five
locos and it has worked perfectly for
the past seven months. Last month, I
built the microDCC decoder for one of
my older Dapol locos, but I now have
bought a Dapol Prairie, which has a
Next18 socket.
Have you thought about a decoder
board with sound (eg, whistle) that
would slide in through the smokebox door in place of the expensive
Dapol Decoders? I commend you on
Australia's electronics magazine
the design of this DCC system. It was
easy to build, worked the first time and
has given great service. (A. L., Morayfield, Qld)
● You’re right that there is a wide
range of possibilities for soundequipped decoders. It’s not always
easy to find suitable sound samples
(eg, without background noise) and
even then, they are not reproduced
well through a piezo transducer.
It looks like breakout boards are
available for the Next18 interface (eBay
295391545334 & eBay 295391543322).
These might allow you to connect our
Decoders to the Prairie.
Our Super Sound Effects module
project from the August and September 2018 issues (siliconchip.au/
Series/325) is designed so that you
can place your own sound effects files
onto a microSD card and it can interface with the open-drain outputs of a
DCC decoder for control. The design
includes a 1.2W amplifier chip to
allow it to directly drive a speaker, so it
should be possible (space permitting)
to use it for custom sounds.
More accurate DCC
Base Station calibration
I have a question regarding the current calibration of the DCC Base Station (January 2026; siliconchip.au/
Article/19558).
I have built two base stations that
both seem to pass the current offset
adjustment procedure. One reads
1.6A with an 8W 10W resistor across
the tracks as the sole load, while the
other reads 2.1A with the same resistor as the sole load. They should both
be reading closer to 1.9A, ie, 15.7V
÷ 8W.
Even taking inductance or capacitance of the resistor into account, the
readings should be either high or low
for both Base Stations, not one high
and the other low. Why do they differ?
My guess is that the Is outputs are not
very linear, or as a colleague has suggested, the Pico’s analog-to-digital converter is not very good and other steps
December 2026 101
need to be taken in the Pico for any
sort of accuracy. (I. E., Higgins, ACT)
● We didn’t delve into it too deeply
in the article, but the I1x and I2x parameters for the BTN8962TA half-bridge
drivers are effectively the slope adjustments for the main track current and
programming track current. Like the
offsets you already calibrated, the
required slope adjustments for the
BTN8962 can vary quite a bit. This is
the diagram from the data sheet shown
to the right.
The figure they give (sense current
ratio; 5.4.6) effectively has a ±28%
tolerance from nominal (see the adjacent table).
The prototype was close to the typical value, but your values are still
within tolerance. If your units still
have their default values of I1x =
0.0081, we suggest trying the following (after ensuring the offset values
are correct first, as it seems you have
already done).
For the unit that read 1.6A with the
8W resistive load, change I1x to 0.0099.
For the unit that read 2.1A with the 8W
resistive load, change I1x to 0.0076.
This is a simple scaling ratio. So if, for
example, the current reading is 10%
high, reduce I1x by 10%.
The I2x channel should be fine, since
the shunt resistor has a much better tolerance than 28%, but it can be adjusted
in the same fashion if required.
A GPO can’t deliver
6kW+ of power
I would like to run a three-phase
7.5 horsepower air compressor from
a single-phase supply. I purchased a
VSD that was supposed to be suitable
for a 7.5hp motor, but it didn’t work.
(J. F., via email)
www.infineon.com/assets/row/public/documents/10/49/infineon-btn8962ta-datasheet-en.pdf
● Andrew Levido responds: The
converter we published (1.5kW Variable Speed Drive, November and
December 2024; siliconchip.au/
Series/430) allows small three-phase
motors to operate from a single-phase
mains supply. This is capable of driving three-phase motors rated at up to
1.5kW as long as their windings can
be configured for 230V/240V phaseto-phase operation.
Small three-phase motors (2.2kW
and under) are often built for
230V/240V operation with the windings in delta, and 415V with the windings in star. This way they can operate
from single-phase to three-phase converters like ours that can only supply
230-240V between phases, as well as
the normal 415V three-phase supply.
The nameplate on your motor indicates it is unsuitable for use with our
controller for two reasons. Firstly,
it is rated at 5.5kW, way beyond the
1.5kW rating of our controller. It is also
way beyond what a normal domestic
single-phase circuit can supply.
Secondly, according to the rating
plate, your motor is wound for 415V
in delta, so it requires a variable speed
drive capable of supplying 415V
phase-to-phase.
We are not aware of any 230V single-
phase input variable speed drives that
can do this. Motors this size are usually used with 415V three-phase input
variable speed drives.
A 25A outlet can’t supply enough
power to run a 5.5kW motor, and 230V
is not high enough to run your motor
WARNING!
Silicon Chip magazine regularly describes projects which employ a mains power supply or produce high voltage. All such projects
should be considered dangerous or even lethal if not used safely. Readers are warned that high voltage wiring should be carried
out according to the instructions in the articles.
When working on these projects use extreme care to ensure that you do not accidentally come into contact with mains AC
voltages or high voltage DC. If you are not confident about working with projects employing mains voltages or other high voltages,
you are advised not to attempt work on them. Silicon Chip Publications Pty Ltd disclaims any liability for damages should anyone
be killed or injured while working on a project or circuit described in any issue of Silicon Chip magazine.
Devices or circuits described in Silicon Chip may be covered by patents. Silicon Chip disclaims any liability for the infringement of such patents by the manufacturing or selling of any such equipment. Silicon Chip also disclaims any liability for projects
which are used in such a way as to infringe relevant government regulations and by-laws.
Advertisers are warned that they are responsible for the content of all advertisements and that they must conform to the Competition & Consumer Act 2010 or as subsequently amended and to any governmental regulations which are applicable.
102
Silicon Chip
Australia's electronics magazine
siliconchip.com.au
MARKET CENTRE
Advertise your product or services here in Silicon Chip
FOR SALE
FOR SALE
PCB PRODUCTION
LEDsales
Tronixlabs
PCB MANUFACTURE: single to multilayer. Bare board tested. One-offs to
any quantity. 48 hour service. Artwork
design. Excellent prices. Check out our
specials: www.ldelectronics.com.au
LEDS, BRAND NAME AND GENERIC
LEDs, filament LEDs, LED drivers,
heatsinks, power supplies, kits and
modules, components, breadboards,
hardware, magnets. Please visit www.
ledsales.com.au
WE’RE BACK – AND WE’VE
BROUGHT THE KITS!
Discover something new at the
Tronixlabs store. Explore our growing
range of Australian electronics kits,
learning resources and more for curious
beginners and seasoned enthusiasts
alike. Visit tronixlabs.com to find
your next project and sign up for our
informative newsletter.
Or for more information by post, write
to PO Box 87, Laidley QLD 4341.
KIT ASSEMBLY & REPAIR
KEITH RIPPON KIT ASSEMBLY &
REPAIR:
* Australia & New Zealand;
* Small production runs.
Phone Keith: 0409 662 794
keith.rippon<at>gmail.com
Advertising in Market Centre
PMD WAY offers (almost) everything
for the electronics enthusiast – with
full warranty, technical support and free
delivery worldwide. Visit pmdway.com
to get started.
anyway. Perhaps the 7.5hp rating of
the VSD you bought was ‘overly generous’.
Is the BF469 really
discontinued?
I’ve been a Silicon Chip reader for
many years and have particularly
enjoyed the audio projects, completing many since building your LM1875based amplifier in the 1990s.
One project I especially enjoyed was
the Hummingbird amplifier (December
2021, siliconchip.au/Article/15126).
When I first read that the BF469 transistor was becoming difficult to obtain,
I ordered some from a local supplier
that still had Philips stock.
After reading the recent Calliope
article (April 2026; siliconchip.au/
Article/20084) and its discussion of
the difficulty in sourcing a suitable
VAS transistor, I was surprised to find
that the BF469 was still available. I
then checked my Hummingbird modules more closely and discovered that
the BF469s I had fitted were not Philips
devices at all, but CDIL parts.
siliconchip.com.au
Classified Ad Rates: $32.00 for up to 20 words (punctuation not charged) plus $1.20 for
each additional word. Display ads in Market Centre start at $82.50 per column centimetre
per insertion. All prices include GST.
Closing date: 5 weeks prior to month of sale. To book, email the text to silicon<at>siliconchip.
com.au and include your name, address & credit card details, or phone (02) 9939 3295.
CDIL still appears to manufacture
the BF469, and its datasheet closely
resembles the Philips specifications.
Other manufacturers have also produced versions of the device.
There is plenty of discussion online
suggesting that CDIL BF469s are inferior to the Philips parts, but I have been
unable to find any useful side-by-side
measurements. The specifications for
both devices are remarkably similar,
and the CDIL part is listed as a drop-in
replacement.
The question seems increasingly
relevant as older semiconductor types
disappear or are quietly revised. I
recently encountered a similar situation with NE5532s, as you described
in the August 2026 issue. I rejected a
number of devices that measured on
par with Chinese counterfeits, only
to discover that they were genuine
TI parts!
I’m curious about how much importance should be attached to the manufacturer when a transistor carries the
same type number and the published
datasheets appear to specify essentially the same device.
Australia's electronics magazine
Substituting devices is nothing new,
and the Calliope article itself demonstrates how successfully this can be
done, but it leaves those of us without a distortion analyser or the expertise wondering whether an apparently
suitable substitute (or revision for that
matter!) still performs as it should.
More generally, how should a hobbyist approach this sort of problem?
Alas, can a CDIL BF469 therefore
reasonably be regarded as equivalent
to the original Philips device, or are
there manufacturer-specific differences significant enough to matter in
circuits such as the Hummingbird? (W.
J., Adelaide, SA)
● CDIL is a legitimate Indian semiconductor fabrication company that
has been in business since the 1960s.
However, there are some reasons that
we wouldn’t continue to specify the
BF469 if they are the sole supplier.
The short version is that often,
parts made by other manufacturers
are close enough to the originals to
be substituted in their stead, but it
isn’t always the case, so you have to
be a bit careful.
December 2026 103
Advertising Index
Altronics.................................41-44
Blackmagic Design................. 7, 11
DigiKey Electronics..................OBC
Emona Instruments.................. IBC
Hare & Forbes............................. 13
Jaycar............................. IFC, 26-29
Keith Rippon Kit Assembly....... 103
LD Electronics........................... 103
LEDsales................................... 103
Microchip Technology.................. 5
Mouser Electronics....................... 3
PCBWay......................................... 9
PE Back Issues............................. 4
PMD Way................................... 103
SC Versatile Battery Checker..... 12
Silicon Chip Binders.................. 87
Silicon Chip Christmas PCBs... 75
Silicon Chip DCC Kits................ 99
Silicon Chip Shop.................76-77
Silicon Chip Subscriptions...... 100
The Loudspeaker Kit.com.......... 93
Tronixlabs.................................. 103
Wagner Electronics....................... 8
Errata and on-sale date
PicoSDR, April 2026: the optional
external TFT LCD screen should
be the 2.8-inch type with the
ILI9341 controller (Silicon Chip
Cat SC3410), not the 3.5-inch type
with an ILI9488 controller, as was
specified in the parts list originally.
Next Issue: the January 2027 issue
is due on sale in newsagents by
Tuesday, December 29th. Expect
postal delivery of subscription
copies in Australia between
December 28th and January 13th.
104
Silicon Chip
It can also be hard to find a distributor for these parts. None of our usual
parts sources, including element14,
RS, DigiKey and Mouser, seem to stock
these CDIL BF469s. We can find some
on eBay, but that is not a recommended
place to source critical transistors for
audio amplifiers! It’s rife with counterfeits.
Without having some to test, we
couldn’t say if they are any good. It’s
possible they are fine, but we can’t
recommend them just because they’re
supposedly equivalent.
We’re also concerned if there is
only a single manufacturer making a
specific part. They could discontinue
production tomorrow, like many of the
previous BF469 manufacturers did.
Thus, we think the SOT-223 transistors Phil Prosser specified in the Calliope article, like the BF722, are the best
alternatives available. They’re inexpensive, tested and known to work
well. Their supply seems to be good,
and they are easy to buy.
One thing we should note is that
while looking for a supplier for the
CDIL BF469s, we discovered that element14 sells “Multicomp Pro” (their
house brand) BF469s (Cat 4343315).
Oddly, they list the Vceo rating as 50V,
when a proper BF469 has a 250V rating. Another case of a part claiming
to be something it isn’t? On the other
hand, they claim an ft rating of 1.5GHz
– impressive if true!
Sourcing parts for an
Ultra-LD Mk.4 Amp
I’m looking at putting together a
few of the Ultra-LD Mk4 Amplifiers (July-October 2015; siliconchip.
au/Series/289). The HN3A51F and
HN3C51F dual transistors used are
no longer available. Is there an equivalent?
Also, the NJL3281 and NJL1302DG
Thermaltrak transistors, although not
(yet) obsolete, seem to be somewhat
out of stock in many places. Jaycar
has them in stock at some stores, but
they’re listed as “Clearance” on their
website.
On an unrelated subject, I’ve ordered
two ADAU1467 Core boards through
AliExpress to build the Digital Preamplifier and Crossover (October-December
2025; siliconchip.au/Series/449). In
both cases, I got ADAU1466 core
boards. Talking to both AliExpress
and to the shop has been an exercise
Australia's electronics magazine
in frustration. (K. J., Cleveland, Qld)
● The IMX8-7-F is a reasonable
substitute for the HN3C51F dual NPN
transistor used in the Ultra-LD Mk.4
amplifier, but unfortunately, there
are no good alternatives to the PNP
input pair (HN3A51F) with the same
performance specifications and footprint. Happily, though, we still have
some and can supply a small set of
parts that includes both (siliconchip.
au/Shop/7/3400).
While the Thermaltrak transistors
are becoming harder to buy, they work
extremely well, and we don’t think
it’s worth designing a new amplifier
for a different transistor series until
they are simply unavailable. If Jaycar
runs out of stock, we suggest ordering
from DigiKey, which still has reasonable stock levels for now:
www.digikey.com.au/en/products/
detail/onsemi/1484399
www.digikey.com.au/en/products/
detail/onsemi/1484396
Regarding the ADAU1467 core
boards, Phil Prosser responds: that’s
unfortunate, they aren’t cheap! There
are some significant differences in
I/O between the ADAU1466 and
ADAU1467. Without doing a lot of
work, I can not say if you can make
it work. You would definitely need to
reprogram the ADAU chip using Sigma
Studio, and I am worried that they
won’t have sufficient I/O for the job.
When ordering from AliExpress,
you need to be really careful as sellers
often combine multiple different items
into the same listings, and you have
to choose the right “colour” (which
actually selects the model). Sometimes there’s a single item for sale, and
we can provide a link, then the seller
changes the listing to include multiple
items, which is a real trap.
Using ETI327 Hazard
Flasher project at 6V
Do I need to modify ETI project 327,
“Turn and hazard indicator for your
vehicle”, for use on a vehicle with a 6V
electrical system? (K. W., Manly, Qld)
● The 555 ICs used for timing
should operate at 6V without needing
any changes to that part of the circuit.
We suggest changing the base resistors
for Q2 and Q3 from 2.2kW to 1kW to
maintain a similar base current at 6V
compared to the original design at 12V.
Otherwise, we expect it will operate
SC
correctly with a 6V supply.
siliconchip.com.au
“Rigol Offer Australia’s Best
Value Test Instruments”
New
2026
Products
Oscilloscopes
New
12Bit
Scopes
New
Up To
13GHz
RIGOL DS-1000Z/E - FREE OPTIONS
RIGOL DHO/MHO Series
RIGOL DSO-8000/A Series
450MHz to 200MHz, 2/4 Ch
41GS/s Real Time Sampling
424Mpts Standard Memory Depth
470MHz to 800MHz, 2/4 Ch
412Bit Vertical Resolution
4Ultra Low Noise Floor
4600MHz to 13GHz, 4Ch
410GS/s to 40GS/s Real Time Sampling
4Up to 4Gpts Memory Depth
FROM $
649
FROM $
ex GST
684
FROM $
ex GST
13,191
Multimeters
Function/Arbitrary Function Generators
RIGOL DG-800/900 Pro Series
RIGOL DG-1000Z Series
RIGOL DM-858/E
425MHz to 200MHz, 1/2 Ch
416Bit, Up to 1.25GS/s
47” Colour Touch Screen
425MHz, 30MHz & 60MHz
42 Output Channels
4160 In-Built Waveforms
45 1/2 Digits
47” Colour Touch Screen
4USB & LAN
FROM $
713
FROM $
ex GST
Power Supplies
ex GST
604
FROM $
ex GST
Spectrum Analysers
632
ex GST
Real-Time Analysers
New
Up To
26.5GHz
RIGOL DP-932E
RIGOL DSA Series
RIGOL RSA Series
4Triple Output 2 x 32V/3A & 6V/3A
43 Electrically Isolated Channels
4Internal Series/Parallel Operation
4500MHz to 7.5GHz
4RBW settable down to 10 Hz
4Optional Tracking Generator
41.5GHz to 26.5Hz
4Modes: Real Time, Swept, VSA, EMI, VNA
4Optional Tracking Generator
ONLY $
799
FROM $
ex GST
1,321
FROM $
ex GST
3,210
ex GST
Buy on-line at www.emona.com.au/rigol
Sydney
Tel 02 9519 3933
Fax 02 9550 1378
Melbourne
Tel 03 9889 0427
Fax 03 9889 0715
email testinst<at>emona.com.au
Brisbane
Tel 07 3392 7170
Fax 07 3848 9046
Adelaide
Tel 08 8363 5733
Fax 08 83635799
Perth
Tel 08 9361 4200
Fax 08 9361 4300
web www.emona.com.au
EMONA
|