This is only a preview of the August 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:
Articles in this series:
Items relevant to "180-230V DC Motor Speed Controller, part two":
Articles in this series:
|
AUDIO
OUT
AUDIO OUT
L
R
By Jake Rothman
Balance with Baxandall
P
eter Baxandall was possibly
Britain's greatest discrete
analog audio circuit designer.
He combined physics, magnetics,
acoustics, music and electronics
into one with great creative flair. He
initially worked at the Royal Signals
and Radar Establishment in Malvern.
The creator of stereo and the longtailed pair, Alan Blumlein, was also
based there, albeit before Baxandall,
so they never met. I like to imagine
what they would have come up with
if they had.
Like Blumlein’s work, a Baxandall
circuit was always original, backed
up with practical research, and costefficient. He is most famous for his
*Typically Sowter
10kΩ:10kΩ
*Typically Sowter
T1*
10kΩ:10kΩ 1:1
T1*
1:1
10kΩ
balanced
10kΩ
input
balanced
input
3 –
Female
3 – +
XLR
socket
Female
2
+
XLR socket 1
2
1
Jinxed!
I wanted to interview Baxandall
in Malvern for Practical Electronics
when Mike Kenward was editor. Sadly,
Baxandall died in 1995. Then I thought
I’d try to interview John Linsley-Hood
in Taunton instead, and he died too.
Rf
10kΩ
Rf
10kΩ
RIN
10kΩ
RIN
10kΩ
Rf115pF
27kΩ
–
–
+
+
Output
(Non-inverting
Output
(Non-inverting
2
+
2
1
1
Insert 1000µF non-polarised
capacitor for low DC offset
Insert 1000µF –
non-polarised
capacitor for low DC offset
– RIN3
27kΩ
Rf
*600Ω 1:1 mumetal +15V
10kΩ
modem
transformer
R
*600Ω 1:1
mumetal +15V
f
10kΩ
transformer
½Rmodem
IN
7
T1*
5kΩIN
½R
1:1
7
T1*
5kΩ
–
1:1
–
NE5534
NE5534
8
½RIN
+
5kΩ
½RIN
+ 5 8
4 5
5kΩ
22pF
4
–15V
22pF
0V
–15V
0V
3 –
3 – +
Output
Output
Differential
virtual earth
input
Differential
virtual earth
input
3
3 –
Input
1
2
+
2
1kΩ
–
½LF353
+
1
Output
RIN
10kΩ
0V
Metal work
Fig.2: a commercial example of a Baxandall balanced
line input (not very well optimised).
40
+–
Output
NE5532A
0V +
Output
NE5532A
15pF
0V
Rf215pF
27kΩ
–
+
+–
NE5532A
0V (Signal
ground)
+
NE5532A
Fig.3(a): a balanced virtual earth used
in the full-spec Baxandall
0V (Signal ground)
balanced input
in the KEF KM15.6kΩ
active
monitor Inner
speaker.
winding has
5.6kΩ
Sec
lower DC resistance
–
Inner winding has
To differential
lower
DC resistance
virtual earth
–
input
3 –
180pF
5.6kΩ
To differential
5.6kΩ 180pF
5-65pf
+
+
virtual
earth
Input
trimmer
input
2
180pF
1 3 –
Custom-wound 1:1 transformer
5.6kΩ
5.6kΩ
+
Mumetal 187 core (Belclare)
+
Input
25mm wide, 0.72cm² stack, twin bobbin
2
Metal work
1000
turns
on
each
side
1
Custom-wound 1:1 transformer
Metal work
Mumetal 187 core (Belclare)
25mm wide, 0.72cm² stack, twin bobbin
Metal work
1000 turns on each side
Metal work
5.6kΩ
180pF
Rf
10kΩ
Termination
resistor
–
Pri
560pF
Sowter
5340
1:2
RIN3
27kΩ
Rf1
27kΩ
Rf2
27kΩ
+
Fig.1(b):Metal
thework
Baxandall balanced input configuration. The
Metal work
input resistor(s) and input transformer positions are flipped.
RIN
10kΩ
Normally, a differential op amp
15pF
Fig.1(a): an inverting amplifier with balanced input transfomer.
Input
Input
Transformer input
Inverting
op amp
Inverting
op amp
0V
0V
Metal work
Metal work
Not to give up, I went to Newport
Radio Rally to interview the Reverend
George Dobbs, designer of the Ladybird
book transistor radio, but he cancelled
due to ill health and died. I then went
to interview John Birkett in Lincoln,
owner of the longest-lived component
shop in the UK. You can guess what
happened. Am I cursed? You be the
judge...
Anyway, back to stuff we can
control. Following on from the
previous investigations of obscure
audio circuits, we’ll look at Baxandall’s
balanced input circuit.
bass-treble tone control, which has
adorned almost every Hi-Fi amp since
the 1950s. He won a competition with
it and never got a royalty, just a $25
watch.
His series on power amplifier design
in Wireless World in the late 1970s
inspired Douglas Self’s series in the
1990s. All engineers, like Newton,
build on the shoulders of giants.
5.6kΩ
5-65pf
trimmer
Pri
Sec
Fig.3(b): the transformer stage feeding the above amplifier.
Most 1:1 600Ω to 2kΩ small audio transformers will work well.
Practical Electronics | August | 2025
circuit would be used for a balanced
input because it is linear and low in
cost. A transformer input, however, is
fully isolated from Earth (floating) and
has a huge common mode rejection ratio
(CMRR). This means it is very useful
in areas of high interference, such as
broadcast studios and electric transport.
The common-mode rejection ratio is
the amount by which a common signal
to both inputs (such as mains hum) is
attenuated at the output of the circuit.
So with a 100dB CMRR and a 100mV
unwanted common-mode signal, only
1μV of that 100mV will make it through
to the output.
I remember a situation with a shipbased pirate radio station, where the
RF induced a voltage in every bit of
metal, and an op amp balanced input
would have been saturated. The ‘oldfashioned’ transformer inputs on the
desk worked fine, even with 50V of
common mode signal on the positive
and negative inputs.
There are expensive problems
with audio balancing transformers;
it typically costs £12-40 for one that
has good frequency response and
low distortion. The good news is
that with Baxandall’s circuit, a cheap
transformer can give excellent results.
circuit improvements can simply be
to swap over the positions of two
components. This was the key to the
success of Baxandall’s design, which
was based on an inverting op amp
circuit.
Normally, the input balancing
transformer feeds the input resistor RIN,
as in Fig.1(a). Having the transformer
feed the inverting input directly and
putting the input resistor first, as
in Fig.1(b), makes the transformer
more linear. This gives a much
better frequency response and lower
distortion.
This is because the inverting
input is a virtual ground, meaning
there is almost no voltage across the
secondary. Since a transformer reflects
what is happening onto the other side,
there is no voltage across the primary
either; all the input voltage is dropped
across RIN.
This setup makes the transformer
operate in current mode, and because
this current is set by the input resistor,
which is inherently linear, the currents
in the transformer windings are forced
to be linear. With a transformer
operating in normal voltage mode,
saturation and hysteresis effects are
much more pronounced.
Inverting the inverted
Practical circuits
Sometimes one of the simplest
To optimise the CMRR, RIN is split
Photo 1: a typical expensive
input transformer in a Mumetal
can.
into two individual resistors, which
positions the transformer’s capacitance
symmetrically, helping the balance at
high frequencies. The first example of
Baxandall's circuit I saw was in a line
input for a Sonifex call box used in
broadcast systems, shown in Fig.2. I
had to service one in the early eighties.
This circuit strangely used an
expensive Sowter transformer (similar
to the one in Photo 1) with a noisy JFET
op amp. It also had an odd termination
resistor, RT, to ensure stability, which
increased the noise gain. It would have
worked better if it had had a small
capacitor in series with RT, making it
into a Zobel network.
The circuit reached a new level of
sophistication in the input stage for
KEF’s KM1 active studio monitor,
for which Baxandall designed the
electronics. It was made for the BBC
and Abbey Road studios. A differential
virtual earth amplifier was used, as in
Fig.3(a), which gave another layer of
common-mode signal rejection.
The transformer input stage shown
in Fig.3(b) also split the input resistors
to include an HF trimming network.
A big electrolytic capacitor was
included to minimise the DC offset,
which was otherwise around 300mV
because of the high input bias currents
of the NE5532 op amps. This caused
a low-frequency (LF) hump of 5dB at
0.8Hz by resonating with the secondary
inductance. This could be fixed by
putting in an LF step network in series
with the capacitor, or with the offset
removed entirely using a servo.
There is another variation of the
circuit (by Barry Blesser, on p297 of
the Journal of the Audio Engineering
Society, May 1972) where the feedback
resistor is fed into a third winding to
cancel the flux, which shrinks the
transformer even more.
Winding your own
Baxandall was highly adept at
designing signal transformers, a lost
skill today. He used to give full winding
details for many of his transformerbased designs, such as his Wireless
World March 1957 Inexpensive HighQuality Amplifier.
This input stage was no exception,
and the details are given in Fig.3(b).
Since the primary has series resistors,
it can be wound with very thin wire.
The secondary should be wound with
thicker wire on the inside winding for
lower resistance, giving lower noise.
Sadly, one can’t get Belclere
transformer kits any more, which
included the laminations and bobbin.
But I have found there are Mumetal
600Ω 1:1 modem isolation transformers
available inexpensively on the surplus
Practical Electronics | August | 2025
41
Rf
10kΩ
AC negative feedback
Rf
10kΩ
470kΩ
AC negative feedback
DC negative feedback
1:1 600Ω
modem
transformer
CIN
2
1:1 600Ω
R
RIN IN
modem
1.1kΩ
1.1kΩ
transformer
3
1
4
2
+
2
1
RIN
1.1kΩ
3 –
Input
Metal work +
2
1
3
0V
+
C
VBiasIN
V+ input
Cascode
8Ω
2200µF
– power amplifier
+
3 –
Input
V+
470kΩ
2200µF
DC negative feedback
–
4
+
RIN
1.1kΩ
(see last month)
+
VBias
1
Zobel network
Cascode input
power amplifier
(see last month)
8Ω
0V
Fig.4(a): two weird circuits joined
into one. Adding the Baxandall balanced input
0V
0V
Metal work
to last months
complementary cascode
amplifier.
Unfortunately, the inductance of
10kΩ
Zobel network
the transformer resonates with the capacitor CIN and the AC feedback path.
V+
–
2kΩ
3 –
Input
1
+
2
3 –
Input
Metal work +
2
1
2kΩ
2kΩ
VBias
+
V+
0V
–
VBias
Output
15Ω
capacitors
e.g. LS3/5A
omitted
0V
V–
15Ω
e.g. LS3/5A
10kΩ
+
+23V
–
Metal work
+23V
10kΩ
–
VBias
Output
capacitors
omitted
+
VBias
2kΩ
+23V
10kΩ
V–
0V
+
+23V
0V
Fig.4(b): using the Baxandall circuit with two complementary cascode amplifiers
to make a more powerful bridge set-up. The distortion should, in theory, be lower
than one amplifier.
market that do the job (see Photo 2).
I bought 200 of them just so you can
have some as well! Note that the inner
winding is pins 1 and 2.
Ferrite pot cores are not normally
used for audio signals, but the
Baxandall circuit compensates well
for their shortcomings. If I was
designing a balanced input for, say, a
railway public address system, where
maximum fidelity was less important,
I would use them. They are wellshielded with the winding enclosed,
and don’t need a screening can like
laminated transformers.
It is well worth reading Baxandall’s
chapter, Microphone amplifiers and
transformers, in the Microphone
Engineering Handbook by Michael
Gayford, published by Focal Press.
Performance
Using the circuit shown in Fig.1(b)
with Rf = RIN =10kΩ, using a NE5534
op amp and the modem transformer
at 0dB (line level), the total harmonic
distortion plus noise (THD+N)
measured 0.0018% between 100Hz
and 20kHz. It rose gently to 0.028% at
20Hz and then was flat down to 5Hz.
Using the transformer in the
conventional way gave over 5%
THD+N at 20Hz due to saturation,
falling to 0.01% at 1kHz.
The Baxandall circuit basically
reduced the core flux density to a
minimum at LF. The >100Hz THD+N
reduced to 0.001% when the resistors
were reduced to 5.1kΩ because of the
lower Johnson (thermal) noise – see
Fig.5.
Note that to ensure HF stability, Rf
needs a 47pF phase-lead capacitor
across it, and the NE5534 needs a 22pF
compensation capacitor between pins
5 and 8.
Putting it all together
Photo 2: this modem phone-line isolation
transformer (25D8310MO1) is ideal for the
Baxandall circuit. It is unscreened and must be kept
away from hum sources, like mains transformers.
42
The complementary cascode power
amp described last month, being an
inverting configuration, could combine
nicely with the Baxandall balanced
input, as shown in Fig.4(a). There is an
interaction between the transformer’s
inductance and the loudspeaker output
and input capacitor (Cin), causing a
hump in the LF response that needs
resolving, though.
Another interesting project would
be to make a bridged version to
more than double the power output,
like in Fig.4(b), using the balanced
drive from the transformer. Since the
complementary cascode amplifier
is asymmetrical, the even harmonic
distortion components should cancel
out, dropping the THD.
In theory, the output capacitors
could be left off (avoiding the LF hump
Practical Electronics | August | 2025
problem) since both output pins are
sitting at the same voltage, but I would
be worried about transient DC currents
flowing through the speaker. There’s
only one way to find out, and that’s to
build it. I’ll do that as soon as I have
some proper PCB designs ready to go.
Analog audio design is like cooking;
it’s how you combine the elements and
their interactions that matter. Very few
circuits are truly new; just different
combinations.
Total Harmonic Distortion (%)
0.5
0.2
0.1
0.05
0.02
0.01
.005
.002
.001
.0005
.0002
.0001
Minor erratum
20
50
100
200
500
1k
Frequency (Hz)
2k
5k
10k
20k
Fig.5: the Baxandall balanced input distortion plot with a modem transformer (768
25D8310MO1) and NE5534 op amp at full output (27V peak-to-peak into 600Ω).
In last month’s article, I mentioned
that the Bailey and Quad 303 amplifiers
used rare Motorola transistors that had
minimal Early effect. I should have
written that they were RCA transistors
PE
instead.
Techno
Talk
puzzle
This question was posed
on the bottom of page 5,
asking readers to spot the
difference between two
images of circuit boards,
one a reference and the
other a productiton part,
which was used as a test
for an image difference
analysis system.
The system detected an
extra part, which was added
as a test (upper red box). It
also detected foreign object
debris, which could become
a problem at some point in
the future (lower red box).
FIND ALL YOUR ELECTRONIC
COMPONENTS IN ONE PLACE
BASIC MICRO
E L E CT R O N I C S C O M P O N E N T S U P P L I E R
w w w . basicmicro . co . u k
High-quality, genuine parts
Practical Electronics | August | 2025
43
|