This is only a preview of the July 2025 issue of Practical Electronics. You can view 0 of the 80 pages in the full issue. Articles in this series:
Articles in this series:
Articles in this series:
Articles in this series:
Items relevant to "180-230V DC Motor Speed Controller":
Articles in this series:
Items relevant to "Repurposing the Mains Power-Up Sequencer":
Articles in this series:
Items relevant to "Intelligent Dual Hybrid Power Supply,.Part 2":
|
AUDIO
OUT
AUDIO OUT
L
R
By Jake Rothman
A complementary cascode amplifier
I
n the analog audio world, amplifier
circuits mostly all look the same. That
is because it’s a mature technology
and most engineers have settled on
recycling circuits from Douglas Self’s
books, simply because they’re proven to
give excellent results at a reasonable cost.
Unfortunately, this gives little scope
for creativity. I enjoy designing slightly
obscure discrete circuits, especially for
musical instruments, where they still
come into their own.
For big Hi-Fi audio amplifiers, it’s
difficult to beat Self’s circuits without a
further increase in complexity, such as
Bob Cordell’s DH-220C Mosfet power
amp. However, there is still a creative
+V
TR2
Input 2
Non-inverting
Re
TR1
Input 1
Inverting
Output
RL
0V
Fig.1: a complementary (NPN & PNP)
cascode. It can be considered similar
to a standard long-tailed pair.
While low levels of even harmonic
distortion are not a subjective problem
in audio, it will likely make the total
harmonic distortion (THD) higher on
the specification sheet.
window available between applicationnote design, such as the TDA2050 chip,
and Self’s “blameless” architecture.
This is where the complementary
cascode circuit comes in (Fig.1), along
with other early transistor circuits from
the 1960s. It offers the AC performance of
the long-tailed pair (LTP) with only two
transistors rather than five, if a current
mirror and sink is included.
This complementary cascode
configuration could be considered a
type of complementary LTP. The catch
is it exhibits a DC offset of 1.2V, which
is why it is rarely used. This is because
the two base-emitter (VBE) drops add
up rather than cancel. Having said that,
fixed offsets don’t matter if the amplifier
is configured as a single-rail AC-coupled
design with an adjustable half-rail bias.
The complementary LTP needs no
constant-current circuit since there is
no tail current to consider, because one
emitter directly feeds the other. Also,
only one emitter resistor (Re) is needed.
With the values shown in Fig.3, the
gain is reduced to around 2.2× from
100×, with a commensurate reduction
in distortion and an increase in signal
handling capability.
The even-order distortion cancellation
is not as good as with the LTP,
though, since the pair matching of
complementary transistors is never going
to be as good as same polarity devices.
Complementary cascode buffer
Before designing a power amplifier
using a new circuit topology, it makes
sense to start with a simpler practical
circuit, like a buffer. As with the LTP,
most cascodes have two inputs, with
one input often connected to a fixed
bias voltage.
With the complementary cascode,
one input is inverting and the other
non-inverting. As in op amp circuits, the
inverting input is often used for negative
feedback. Note that in this circuit, the
voltage amplifier stage (VAS), a common
emitter stage, is inverting. So the inputs
to the cascode have to be flipped.
To reduce the distortion of tantalum
and electrolytic output coupling
capacitors, negative feedback (NFB) can
be used, as shown in the op amp circuit
in Fig.2. The 22nF capacitor has to be
a low-distortion film type. The circuit
also needs to be loaded to avoid a very
low-frequency hump in the response.
This technique can also be applied
to the complementary cascode buffer
shown in Fig.3. A similar technique is
used for the power amplifier speaker
coupling capacitor.
½ NE5532
+
TR2
BC549
Output
–
100kΩ
R2
68kΩ
0V
½ NE5532
10µF
25V
+
Output
+
Input
C1
100nF
Input
–
100kΩ
100kΩ
RLoad
≈2.2kΩ
0V
0V
22nF
Fig.2: one of my favourite negativefeedback tricks I developed to avoid
distortion from output electrolytic caps.
Practical Electronics | July | 2025
R3
47kΩ
0V
R8
10kΩ
TR1
R1
BC559
470Ω 9.8V
C2 +
6.8µF
16V
C6
100nF
11.6V
R5
330Ω
R4
100kΩ
+24V
C5
22nF
C3
22pF
12.6V
12V
R6
330Ω
0.7V
R12
820Ω
R10
220kΩ
TR4
BC337
R7
100Ω
TR3
BC549
C4
10µF
25V
+
10µF
25V
+
Input
R11
47Ω
Output
12mA
R9
1kΩ
0V
Fig.3: a buffer amplifier using a complementary cascode as a complementary
long-tailed pair. I breadboarded this to prove the technique works.
57
Here’s a funny aside. I knew one
engineer who inadvertently relied
on the parasitic inductance of the big
output electrolytic (C10) to avoid using
an inductor on the output. He got his
comeuppance when the capacitors were
“improved” with a lower equivalent
series resistance (ESR) and the amps
oscillated and blew up.
An interesting characteristic of the
complementary cascode that I noticed
is that it has less switch-on/off thumps
compared to the LTP, which tends
to ‘snap on’. Power on/off transient
generation is an area of analog circuitry
that has not been fully researched yet
(perhaps this would be a good topic for
someone’s PhD).
As a designer of active loudspeaker
systems with up to eight power
amplifiers, I will do anything to get
rid of thumps, thereby avoiding the
unreliability of speaker protection
relays.
blocking capacitor is used on the output
so that the half-rail bias does not appear
as DC across the speaker.
I’ve found these capacitors can cause
problems when used with amplifiers
above 40W RMS because they can
generate significant distortion from
the high current passing through them.
With single-rail designs, the capacitor
is polarised at half rail, so the distortion
from polarity reversal does not occur.
The output capacitor also gives
reliable DC fault protection, and the
output transistors are also protected from
output short circuits at low volumes.
This is because the negative feedback
loop at DC isn’t broken. Of course, if the
amplifier is playing full pelt when a short
occurs, the capacitor offers no protection.
One major problem with single-rail
power amplifiers is that the signal
ground 0V can become contaminated
with the half-wave rectified currents
from the lower output transistor. This
means that for the same circuit, the THD
is always higher. Greater care has to be
taken with the wiring layout.
If gain is required, the NFB can be
reduced using a potential divider.
The amount of gain for low distortion
is limited in this circuit, because the
open-loop gain is low compared to an op
amp. However, the gain can be increased
by a bootstrapped or constant-current
collector load for resistor R9.
The power amplifier
The first commercial power amplifier
using the complementary LTP I had was
a Bi-Pre-Pak module; readers over 60
will remember them being advertised
in a two-page spread in PE. The Sterling
Sound SS125, designed by ex-Sinclair
engineer Richard Torrens in 1977,
inspired this circuit.
It intrigued me because it had a
0.02% THD rating rather than the
typical 0.1% or more at the time. I
used it as a midrange driver amplifier,
along with Ben Duncan’s Hi-Fi News
active crossover in 1980, to make a very
successful sound system. I was only 18 at
the time, and we had great parties until
my Dad got fed up.
Loads of ladies who wanted to be
medics turned up, but they paired up
with those men destined to become
dentists. Sadly, they got paid much more
for drilling holes in teeth than I did for
drilling PCBs.
Bias generator
Let’s now look at the features of the
complementary cascode power amplifier
circuit I’ve designed, shown in Fig.4.
For low-level circuits such as preamps,
a resistor-based potential divider to
generate the half-rail bias is fine. For
a Hi-Fi class AB power amp running
on a typical unregulated supply, that
is not good enough. This is because
the inevitable power supply voltage
fluctuations can be fed back into the
amplifier via the bias voltage, causing
it to modulate itself.
Output inductor
When feeding long loudspeaker leads,
as in typical domestic Hi-Fi setups, the
capacitance can cause power amplifiers
to oscillate. A damped inductor network
(L1, R21) is often used to isolate this.
If the leads are short, say in a guitar
amp combo or active speaker, it can
be omitted. The Zobel network alone,
C12 and R25, will suffice to maintain
stability.
A single supply rail
The traditional LTP is perfect for
dual-rail DC coupled-amplifiers, where
the speaker is connected directly to the
output. With single-rail amplifiers, a DC-
+50V regulated
8.5mA
R8
12kΩ
LED1
Red
3.2mA
2.5mA
4.5mA
R1
2.2kΩ
C1
2.2µF
R3
47kΩ
TR1
BC546B
R12
6.8kΩ
19.5V
C6
470nF
Bias
TR2
BC556B point
C4
2.2nF
R6
820Ω
+
R7
4.7kΩ
R23
330Ω
C5
10µF
25V
IC1
k TL431
Adj
R9
27kΩ
TR6
ZTX651
mal
Ther
link
T
lin her
k ma
l
21V
R5
160Ω
Iq adjust
23V
output
bias
C9
6.8µF
6.3V
Tant
CW
C8
10pF
15-37mA
+
VR2
5kΩ
R10
2.2kΩ
TR3
ZTX694B
a
TR7
ZTX751
R24
13Ω
R17
100Ω
ZD1
43V
R2
10kΩ
58
C7
33pF
R19
0.1Ω
TR9
BD911*
4.5mA
0V
R20
470kΩ
C10
2200µF
50V
LED2
Orange
CW
DC bias
adjust
R18
0.1Ω
C11
220nF
Clip
VR1
5kΩ
+54V unregulated
(drops to +46.5V
on full power)
TR5
ZTX300
R11
5.6kΩ
C2
470nF
*On 6.8°C/W
heatsink
TR8
BD912*
R16
100Ω
R15
33Ω
0V
Input
4.5mA
R13
15kΩ
0.75mA
C3
100nF
R14
220Ω
TR4
BC556B
+
R4
470Ω
Clean
ground
DC negative feedback
AC negative feedback
Practical Electronics | July | 2025
This raises distortion and causes
the speaker cone to flap about at
low frequencies, an effect I call ‘bias
pumping’. If you look at the data sheets
for some power amp chips, it will give a
higher distortion figure for single-rail use
compared to dual-rail (unless it’s been
measured with a stabilised bench power
supply). The solution is to regulate the
bias voltage with a zener diode.
In this case, a TL431 ‘programmable
zener’ IC is used because it is adjustable
(with VR1). If a regulated supply is used
for the input stage, a decoupled resistive
divider will suffice.
It’s always a good idea to design an
amplifier, its power supply and the
speaker to be used as one system. I
trim the bias for symmetrical clip at the
lowest dip in the impedance curve of the
speaker. Do it quickly to avoid burning
out the driver.
Input and VAS power supply
This might be considered an overthe-top upgrade, but it gives an audible
difference to me. As I’m now 62, I
can no longer hear the extreme highfrequency distortion harmonics that most
complex power amplifier circuits aim to
minimise. Despite that, I still measure
high-frequency distortion. I’m also very
aware of hum and its modulation effects.
It’s very difficult to regulate the power
supply to the whole amplifier; it’s often
more complex than the amplifier itself,
and it reduces the maximum output on
peaks by 30%. Luckily, regulating the
low-level input and voltage amplifier
stage parts of the circuit is quite easy.
In this design, the output noise and
ripple was reduced from 10mV to 2mV by
adding regulation for the early, low-power
stages. Use of a more complex LM317
high-voltage regulator circuit rather than
the simple zener scheme (ZD4, ZD5) used
here will reduce it still further.
If you want to add rail regulation to a
dual-rail amplifier, two regulators will be
L1
10µH
C12
100nF
100V
R25
4.7Ω
1W
Fig.5: a suitable power
supply for the amplifier,
with a voltage doubler
built around diodes
D1 and D2 acting as a
charge pump. TR2 pulls
the regulated rail down
quickly for fast muting
on power-off.
Output
Mains
input
R22
2.2kΩ
0.5W
LS1
8Ω
F1
2A
PTC fuse
T1
60VA
18V
36V
230V
R3
22kΩ
D3
1N4148
C4 +
2.2µF
63V
C2
22µF
100V
6A
200V
Bridge
L
–
Faster switch-off
One of the problems with the
traditional linear power supply setup
is that it does not switch off quickly; it
carries on working until the capacitors
discharge. This can let through thumps
from previous stages. By turning off the
low-power regulated supply quickly, the
power amp can be muted.
This is accomplished by a discharge
circuit (TR1 & TR2) that pulls down
the rail when the AC from the mains
transformer goes away. This is driven by
its own rectifier diode, D3, with minimal
smoothing provided by capacitor C4,
so it powers down before the rest of the
circuit.
The power supply circuit is shown
in Fig.5.
An 18-0-18V 50-60VA transformer is
more than sufficient for one channel.
A top-notch approach to prevent Earth
loops in a stereo amplifier, would be to
use a 35V + 35V 100VA dual secondary
R4
22kΩ
13V
C5
470nF
D2
1N4002
D1
1N4002
Dirty
ground
Fig.4: the complete power amplifier
circuit. The currents through the
driver transistors are about 4.5mA
because the output transistors are
not fully switched on under quiescent
conditions. R3 and C2 form an LF step
network to prevent the LF hump.
Practical Electronics | July | 2025
78V
R2
1kΩ
2.5W
WW
R6
220Ω
0.5W
+50V regulated
+54V unregulated
N
+
C1
10,000µF
50V
R1
1kΩ
0.5W
TR1
BC337
+
18V
Speaker
ground
TR2
BC639
or BD139
R5
10kΩ
+
R21
3.3Ω
1W
series to indicate clipping. The power
is still reduced from 30W RMS into 8Ω
to 26W, though.
Since low-power amplifiers, like
this, will inevitably clip occasionally
in real use, this modification is clearly
audible and sounds excellent as a guitar
amplifier. It also allows us to use most of
the available voltage from a given power
supply transformer.
With the transformer being the most
expensive component in any power
amplifier, it more than pays for itself.
The Hi-Fi enthusiast’s technique of using
a 100W power amplifier to avoid dirty
clips when driving small speakers seems
a very expensive solution.
needed, but with a single-rail amplifier,
you need only one. The famous Quad 303
single-rail amplifier had full regulation,
although it was mainly used to limit
the maximum voltage on the output
transistors. The complexity of regulation
means it’s rarely done.
(Dual regulated audio power amp
supplies are available, such as those
by Hypex, but being SMT switchmode designs with over 100 parts,
I’ve found them to last only last five
years and be uneconomic to repair.
That is unacceptable in the recording
industry.)
Also, with current mirrors and other
improvements to the input and driver
circuitry, the common-mode rejection
of the amplifier can be made very high.
However, this only works until the
output transistors saturate at clipping;
then all the power supply noise is let
through, giving very dirty-sounding
clipping.
This horrid effect can be avoided by
clean-clipping the driver stage first, but
this will again reduce the maximum
output power on peaks by around 30%.
This is because the clipping level must
be set below the lowest voltage that the
unregulated power supply sags to on
kick drums and other peaks.
As usual with electronics, there’s
a nice solution: since the current
requirements are only around 10mA, a
simple voltage doubler can be used with
a zener regulator, and we can then set
the rail voltage a few volts higher than
the output stage supply rail.
A defined ‘clean soft clip’ level before
the output stage can then be set by a
zener diode (ZD1) with one LED in
C3 +
220µF
100V
D4
27V
1.3W +
D5
24V
1.3W
C6
100µF
63V
Clean ground
Loudspeaker ground
Dirty ground
E
C7
10nF
59
toroidal transformer driving two separate
PSU boards.
Voltage amplification stage (VAS)
This is just the standard configuration,
a common-emitter stage with a dynamic
load. The load here is a constant-current
source, rather than a bootstrapped
collector resistor, because it gives lower
crossover distortion. Its higher voltage
loss is obviated by use of the higher
voltage rail from the voltage-doubled
supply.
One problem with the standard
common-emitter VAS is that it is
subjected to the full rail voltage swing
between emitter and collector (VCE).
With bad examples of some older types
of transistors, such as the BFY51 and
BC107, this could result in distortion
from the Early effect. This is where the
gain and capacitance of the transistor is
modulated by its VCE voltage.
I remember having to buy special
Motorola transistors that avoided this
problem. These were used in Bailey
and Quad amplifiers. Another source of
good audio transistors was Ferranti in
Oldham, near Manchester. Their ZTX
E-line series ‘Zetex’ transistors were
used in the groundbreaking PE Gemini
and Orion amplifiers. They are still in
production by Diodes Incorporated 50
years later!
2A 60V ZTX651 and ZTX751
transistors are used in the driver stage
here. The Orion VAS transistor was a
Ferranti BFS61 or ZTX451. I have found
the ultra-cheap BC546B only works okay
for swings up to 30V. For this amplifier,
I used a ZTX694B with a 13Ω emitter
resistor, which exhibits little Early effect.
It has a VCE rating of 120V.
High-voltage transistors always have
less Early effect, but often the Hfe is low;
the ZTX694B has an Hfe of 400. Adding
a small amount of emitter resistance
(R24) also helps reduce distortion at high
levels. Conversely, adding resistance
increases distortion at low levels due to
open-loop gain reduction.
I did try a cascoded VAS stage based on
the Gemini circuit, but I found it offered
no improvement on the ZTX694B.
D. S. Gibbs and I. M. Shaw, who
designed both PE amplifiers, were
Ferranti application engineers at the
time. Shaw was also the inventor of a
diode dodge to improve the symmetry
of quasi-complementary output stages.
Later, this was improved by Baxandall,
then adopted commercially by Rogers
and Naim.
Low gain for low noise
Most power amplifiers have too much
voltage gain for use in active speakers;
typically 20× to 40×. This boosts the
noise from the active filters, creating a
60
Photo 1: a power amplifier
and its supply on a bit of
wood! A technique I often
use for prototyping.
Photo 2: a close-up of the
driver transistor heatsinking
and thermal coupling
arrangement. A bit of thermal
paste in the ‘sandwich’ helps.
noticeable hiss. Peter Baxandall reduced
the gain of the Quad 405 amplifiers
while increasing the crossover level in
the pioneering KEF KM1 active speaker,
resulting in a system as quiet as a passive
speaker.
So with this amplifier, I also decided
to go for a much lower gain. I’ve even
made one unity-gain stable so you can
use it as an active filter itself, or put an
active gain control around it.
I also made it inverting, which allows
a fully filtered half-rail bias generator
to feed the non-inverting input (base
of TR2). Inverting amplifiers generally
have more noise than their non-inverting
counterparts, but this is only if you need
a high input impedance.
In this case, the input impedance is
kept low, at 2.2kΩ; this is high enough
for most op amps and reasonable
discrete circuits to drive. The inverting
configuration is also useful for bridging
with another power amplifier and for
adapting to a balanced input. Now we
have something sufficiently different
from other amplifier circuits to make
the design work worthwhile.
Output stage
I’ve opted to use the complementary
emitter follower pair (CFP) output
configuration here since it works well
for powers below 50W, where the output
transistors can be fairly small, with low
capacitance. For powers above this, and
where devices need to be paralleled, I
use the Darlington emitter follower (EF)
topology, which gives lower distortion
and has better HF stability.
By Darlington, I don’t mean monolithic
Darlington transistors, where there is a
driver and output transistor on the same
chip, since this exacerbates thermal
instability. The driver transistors have to
be separate from the output transistors
to avoid being heated up by them.
Philips used monolithic Darlingtons
successfully in their active speakers
by putting a thermistor in the V BE
multiplier.
For small amplifiers, the CFP gives
Practical Electronics | July | 2025
Photo 3: the prototype stripboard is slightly messy because the board is too short.
Photo 4: the star wiring to capacitor pins is visible on the power supply underside.
lower quiescent dissipation. Also, the
VBE multiplier transistor only needs to
be in thermal contact with the drivers for
thermal stability, avoiding the hassle of
mounting it on the main heatsink.
The crossover distortion of the CFP
is more ‘spiky’ than the Darlington EF
and occupies a smaller voltage swing. It
also has poorer stored charge removal,
hence the HF distortion is a bit higher
at 10kHz. The collector load resistors
on the drivers are quite important for
low distortion at high levels. Generally,
for amps of this size, 100Ω is optimum,
setting the driver current to 4.5mA.
The output transistors TR8 and TR9
are not critical. I’ve used old TIP41/2C,
BD911/912 and even the venerable
TIP3055/2955 complementary pairs. Of
course, if you want lower HF distortion,
you could use modern, faster types.
Building and testing
The complete prototype assembly
is shown in Photo 1. Yes, it is a real
“breadboard”; a nice metal enclosure is
Practical Electronics | July | 2025
not warranted for an initial prototype (the
mains wiring has to be well-insulated!).
I always build prototype power
amplifiers on stripboard; breadboards
are just too prone to bad connections
and short circuits, which could cause it
to blow up. Normally, stripboard is only
suitable for simple circuits, which is
true if one is going to build it in one go.
The key to building a complex, fragile
circuit on stripboard is to build it in testable
sections. As always, electronics can be
easily partitioned into its constituent
stages, and this is the key to its success as
an engineering discipline. Try doing this
when building a railway bridge, though,
and life becomes much more difficult!
Always use a bigger board than you
need. If you run out of holes, you have
to start soldering onto component leads,
which quickly becomes messy. The board
can always be snipped down to size with
side-cutters (or cut with a hacksaw) later.
Although a good PCB layout designer
does not usually place components
in the same physical position as the
schematic, with strip board prototypes,
it makes it easier to test if it’s wired like a
circuit diagram. That is with the positive
rail is at the top, the negative rail on the
bottom and the signal flow from left to
right. The stripboard is shown in Photo 3.
If you have to move something to
optimise the layout, always test it
before adding something else. Once
you have more than two errors, the
time taken to get things working rises
disproportionately.
Add loops of tinned copper wire to
attach test probes and crocodile clips.
Veropins don’t work; the spring-loaded
test leads can ping off. An extra-big loop
should be used for the Earth, since there
may be up to five leads attached.
For this amp and many others, I’ve
built and tested the circuit blocks in
the following order. Always ramp up
the voltage slowly using a bench power
supply and set it to a low current limit,
around 250mA:
1. Create the current source bias voltage
using LED1. This tells you circuit is
powered.
2. Add the current source, TR4. Make
sure the current is correct; in this
case, 4.5mA.
3. Add the input transistors, TR1, TR2
and VAS TR3.
4. Test the gain with feedback resistors
R1 & R2 wired into the circuit.
5. Add VBE multiplier transistor TR5 and
test that the VR2 preset works. Set the
voltage across its collector and emitter
to a minimum.
6. Add driver transistors TR6 and TR7.
Since this is basically a push-pull
emitter-follower, the feedback can be
connected and a basic test performed.
7. Finally, add output transistors TR8
and TR9. Gradually increase the
current limit on the supply.
To thermally couple TR5, TR6 and TR7,
sandwich VBE multiplier TR5 between
the two drivers (TR6 and TR7), wrapping
them all together with a copper strip (see
Photo 2). This also acts as a low-power
heatsink. I used another Zetex transistor
(a cheap low-spec one) for TR5 because
the flat E-line shape allows them to fit
together with maximum surface coupling.
Adjust VR1 for symmetrical clipping
with a 50W 8Ω load resistor and VR2 for
the lowest distortion at low powers (3V
peak-to-peak). This will typically give a
quiescent current of 15-37mA.
The great thing about getting a
stripboard circuit working is that you
can guarantee the next developmental
stage, a proper PCB, will work better.
It will have lower parasitic capacitance
and resistance.
Power supply construction
Stripboard is less than ideal for
building power supplies since it is very
61
0.5
Total Harmonic Distortion (%)
difficult to make star connections on
capacitor pins as shown for C1 in Fig.5
This technique is necessary to reduce
the hum induced by charging pulses
and prevent noisy Earth currents from
impinging on the audio. Power supplies
also need low resistance tracks.
It is possible to buy 0.2-inch (5.08mm)
Veroboard, which is usable, but it’s rare.
It’s best to use perfboard (which has no
copper) with tinned copper wire. This
technique can give better results than
a PCB, probably because the tinned
copper wire has a substantially larger
cross-section than a PCB track with
standard thickness. I occasionally resort
to soldering tinned copper wire onto
PCB tracks to lower their resistance. The
underside is shown in Photo 4.
0.2
0.1
0.05
0.02
0.01
.005
.002
.001
.0005
.0002
.0001
20
50
100
200
500
1k
Frequency (Hz)
2k
5k
10k
20k
Fig.6: the distortion vs frequency curve for the power amplifier. This was done at
low power (2W into 8Ω).
Performance
The distortion vs frequency curve of
an amplifier is always revealing; this
one is shown in Fig.6. The THD+N is
around 0.003% midband at low powers,
so the complementary LTP works almost
as well as a traditional LTP. If a single
input transistor stage had been used, the
distortion would have been about 0.01%.
There is little increase in distortion at
20Hz, which illustrates the effectiveness
of NFB around the output capacitor.
There is a rise to 0.02% at 10kHz at
all levels, which is not so good. This
suggests the amplifier is running out of
open-loop gain at high frequencies. This
is because the compensation, layout and
output stage need further optimisation.
The complementary LTP’s subtraction
for the negative feedback also possibly
doesn’t work so well at higher levels
compared to a properly matched LTP.
The LF/mid-band distortion of the
amplifier steadily increased to 0.005%
at 12.5W and to 0.018% at 20W, so we
were back to the stated specification of
the Stirling Audio module. However,
there could be some other problems,
such as in the component layout.
I’ve effectively reinvented the wheel
here, and it does not give a Douglas Self
Photo 5: the power supply. Note the connections going directly to the pins of the
big reservoir capacitor, C1.
level of performance. However, it’s still
worth adding the complementary LTP to
the armoury of discrete building blocks,
along with this small power amplifier
system suitable for active speakers.
These days, linear power amp chips
are often single-sourced and expensive
compared to a handful of discrete
transistors, so if you have the time, roll
PE
your own amp!
1591 ABS flame-retardant enclosures
Learn more:
hammondmfg.com/1591
uksales<at>hammfg.com • 01256 812812
62
Practical Electronics | July | 2025
Parts List – Complementary Cascode Amp
1 6 × 5in (150 × 125mm) piece of stripboard
2 5kΩ trimpots (VR1, VR2)
1 10µH 5A suppression inductor (L1)
2 6.8°C/W flag heatsinks
1 10mm length of 0.5mm-thick copper strip (from builder’s merchants)
various screws, nuts etc
Semiconductors
1 TL431 programmable voltage reference IC, TO-92 (IC1)
1 BC546B 65V 100mA NPN transistor, TO-92 (TR1)
2 BC556B 65V 100mA PNP transistors, TO-92 (TR2, TR4)
1 ZTX694B 120V 1A NPN transistor, E-Line (TR3)
1 ZTX300/ZTX108 45V 100mA NPN transistor, E-Line (TR5)
1 ZTX651 60V 2A NPN transistor, E-Line (TR6)
1 ZTX751 60V 2A PNP transistor, E-Line (TR7)
1 BD912 100V 15V PNP transistor, TO-220 (TR8)
1 BD911 100V 15V NPN transistor, TO-220 (TR9)
1 3/5mm red LED (LED1)
1 3/5mm orange/amber LED (LED2)
1 43V 400mW zener diode (ZD1) [eg, BZY88C43V43V]
2 100Ω
1 33Ω
1 13Ω
1 4.7Ω 1W
1 3.3Ω 1W
2 0.1Ω 1-2.5W
(wirewound)
Power supply parts
1 4in × 2in (100 × 50mm) piece of perfboard
1 18-0-18V 50VA toroidal or 37V 1.6A EI-core transformer
1 2A 60V PTC thermistor (‘solid state fuse’)
All 60 issues from Jan 2019
to Dec 2023 for just £49.95
PDF files ready for
immediate download
Q ty
1
1
1
2
2
1
1
1
1
1
2
2
1
1
1
2
1
2
1
1
1
1
Value 4-band code 5-band code
470kW
47kW
27kW
22kW
10kW
12kW
15kW
6.8kW
5.6kW
4.7kW
2.2kW
1.0kW
820W
470W
330W
220W
160W
100W
3 3W
1 3W
4.7W
3.3W
Semiconductors
1 BC337 45V 800mA NPN transistor, TO-92 (TR1)
1 BD139 80V 1.5A NPN transistor, TO-225AA (TR2)
1 200V 6A bridge rectifier (BR1)
2 1N4004 400V 1A power diodes (D1, D2)
1 1N4148 75V 200mA signal diode (D3)
1 27V* 1.3W zener diode (ZD4)
1 24V* 1.3W zener diode (ZD5)
* a single 51V zener could be used if available
Capacitors
1 10,000µF 50V solder-tag electrolytic (C1)
1 220µF 100V electrolytic (C3)
1 100µF 63V electrolytic (C6)
1 22µF 100V electrolytic (C2)
1 2.2µF 63V electrolytic (C4)
1 470nF 63V polyester/MKT (C5)
1 10nF 50V X7R ceramic (C7)
Resistors
2 22kΩ ±5% ¼W
1 10kΩ ±5% ¼W
1 1kΩ ±5% 2.5W wirewound
1 1kΩ ±5% ½W
1 220Ω ±5% ½W
Practical Electronics | July | 2025
5-year
collections
2019-2023
Capacitors (all ±20% tolerance unless noted)
1 2200µF 50V low-ESR electrolytic (C10)
1 220nF 50V X7R ceramic (C11)
1 10µF 25V electrolytic (Al or Ta) (C5)
1 100nF 100V polyester/MKT (C12)
1 6.8µF 6.3V tantalum (C9)
1 100nF 50V X7R ceramic (C3)
1 2.2µF 63V polyester/MKT (C1)
1 2.2nF 50V X7R ceramic (C4)
1 470nF 63V polyester/MKT (C2)
1 33pF ±5% NP0/C0G ceramic (C7)
1 470nF 50V X7R ceramic (C6)
1 10pF ±5% NP0/C0G ceramic (C8)
Resistors (all ¼W ±5% or better)
1 470kΩ ±1% metal film 1 4.7kΩ
1 47kΩ
1 2.2kΩ ½W
1 27kΩ
2 2.2kΩ
1 15kΩ
1 820Ω
1 12kΩ
1 470Ω
1 10kΩ
1 330Ω
1 6.8kΩ
1 220Ω
1 5.6kΩ
1 160Ω
NEW!
2018-2022
All 60 issues from Jan 2018
to Dec 2022 for just £49.95
PDF files ready for
immediate download
2017-2021
All 60 issues from Jan 2017
to Dec 2021 for just £49.95
PDF files ready for
immediate download
2016-2020
All 60 issues from Jan 2016
to Dec 2020 for just £44.95
PDF files ready for
immediate download
See page 33 for further
details and other great
back-issue offers.
Purchase and download at:
www.electronpublishing.com
63
|