This is only a preview of the July 2021 issue of Practical Electronics. You can view 0 of the 72 pages in the full issue. Articles in this series:
|
Practically Speaking
Hands-on techniques for turning ideas into projects – by Jake Rothman
Restoring old equipment – Part 4
T
his month, we conclude the
Offset voltage
series on restoring the Leak Stereo
30, looking at the pre and power
amplfers. Plus, we will suggest some interesting upgrades to the circuit design, and
look at useful component substitutions.
Like all old transistor amplifiers, the output
is biased to sit mid-way at around half
the power supply voltage. This has to
be checked before the output coupling
capacitor (C37) and should be around –20V.
Testing the power amps
Quiescent current
The Leak Stereo 30 power amplifier is shown
in Fig.37. Two words of warning, the power
transistor heatsinks are live (connected to
the collector). One brush with an earthed
lead will destroy them. Ensure the volume
control is at minimum in case the preamp is
oscillating or has a leaky coupling capacitor.
Note that the conductive ‘silver’ washers
(Fig.38) for mounting the output transistors
are made from solder or a soft metal to aid
deformation for thermal conductivity.
All discrete class AB amplifiers have
a quiescent current pre-set adjustment
P7 (see Fig.29). If the current is too low,
crossover distortion occurs; if too high,
you get thermal runaway and destruction
of the amplifier. It is a fiddly fine-spot,
best set up with a low-distortion 300Hz
sinewave, notch filter, a pair of ears and a
‘scope to monitor the residual. If you don’t
have the equipment for such a set-up, a
current from 15 to 20mA is about right.
R 45
R 49
R 46
0 V
– 3 9V
+
C 3 1
2 2 µ F
Iq
Iq
C 3 4
3 3 µ F
+
– 15 V
Replacement transistors
T R 6
G E T 5 3 8
R 5 0
– 2 3 . 6V
It’s worth examining the PE Transistor
Guide from the May 1966 issue, which
contains a list of germanium transistors
– it’s available for download from the
February 2021 page of the PE website.
(Pages 4 and 5 are missing – the AC and
AF series. If a reader has a complete copy
we would like to add them.) Also, Andrew
Wylie has a fantastic transistor history site
at: https://bit.ly/pe-ps21-05
T R 8
A D 140
V A 10 3 9
R 61
– 2 3 .2 V
– tº
Iq
R 5 2
T H 2
R 5 5
0 V
V R 7
T R 4
O C 44
– 4. 2 V
C 3 7
10 0 0 µ F
– 2 3 V
to
– 19V
R 5 3
+
C 3 2
10 µ F
R 5 6
+
V R 6
There are links under the PCB for
measuring the quiescent or standing current
in the output transistors with no signal.
However, I found it easier to measure the
voltage drop across the 1Ω emitter resistors.
It’s prudent to set the preset (VR7) to give
minimum current before turning on. Here,
it’s fully clockwise or minimum resistance. I
set it using the preset at 25mV corresponding
to 25mA. This is assuming the emitter
resistors (R56, R57) are 1Ω. After 50 years
they are likely to have drifted high. Also,
germanium devices exhibit a much greater
risk of thermal runaway. They have higher
leakage currents than silicon devices and
this can increase with age. The driver and
output transistors should just be noticeably
warm. If they get hot, turn off. Sudden
quiescent current increases as the preset
is increased must also be watched for.
This can be caused by an oxidised preset,
faulty driver transistors (TR6, TR7) or highfrequency instability. Also, the current will
rise as the amplifier warms up, so trim it
back a bit after it’s run for 15 minutes.
– 1. 4V
R 43
R 44
C 3 3
15 0 0 µ F +
6. 3 V
R 47
– 3 .8 V
Scr een
– tº
R 48
T H 1
V A 10 5 5
0 V
b
C 3 5
8 2 µ F
+
R 5 1
c
T R 7
A C 12 7Z
– 2 2 .8 V
T R 5
A F 118
T R 9
A D 140
R 5 4
R 5 7
0 V
R 5 8
e
b
e
c
C 3 6
3 nF
R 5 9
Fig.37. Leak Stereo 30 power amplifier circuit. The original Leak diagrams use an early
transistor symbol shown in the inset – this is for a bipolar transistor, not a FET!
38
Fig.38. Mullard AD140 power transistor
and heatsink. Note the soft metal washer
is electrically conductive, so the heatsinks
are ‘live’. Is the date indicated by the
three-character code?
Practical Electronics | July | 2021
Fig.39. (above) GET113 small-signal
transistor; Fig.40. (below) The enigmatic
AF118 (TR5) – a rare germanium highvoltage RF transistor. Its data sheet was
in the Siemens 1970-71 Semiconductor
Manual, p.228.
Fig.41. Poor 1kHz square-wave response
due to substituting a slow, low-frequency
transistor (ASY53) for the AF118, an RF
(radio-frequency) transistor (TR5).
52 for these. If you want a proper British
Mullard transistor, the GET872 is a goodlooking substitute.
Small signal
Voltage amplifier
Fig.42. This ASZ20 from Birkett’s was the
only suitable direct replacement for the
AF118; unfortunately, they only had one.
except for radio use, also often works as
the whiskers grow from the inside of the
case. These whiskers can sometimes be
electrically fused by connecting all three
transistor leads together and discharging a
100µF capacitor charged to 35-50V through
to the case, as shown in Fig.43. This dodge
is also very useful for old radios. In my
case, the AF118 had a base-emitter short,
so it couldn’t be fixed.
It can be hard to find replacements for failed
transistors. The preamp transistors (TR1,
TR2, TR3) and power amp input transistors
(TR4, TR5) can all be directly replaced
with silicon types, since the DC bias levels
can be adjusted by pre-sets. BC559s or
BCY71s (if you want an older-looking
type) or other low-noise PNP small-signal
example will be fine. The noise will be
lower than the original germanium types.
I suspect they get leakier and therefore
noisier with age. (At 60 years, we may be
near the wear-out end of the ‘bathtub curve’
of germanium transistor useability see:
https://en.wikipedia.org/wiki/Bathtub_curve).
The input transistor AC107 was originally
designed for tape heads and is supposed to
have a noise factor of 3dB. Even Leak had to
buy specially selected ones. I replace these
with the NKT214F or 2SA49 devices, which
seem to be the lowest-noise germanium
transistors I’ve got. The GET113 (Fig39) is
not critical, almost anything rated at 25V
will be fine here, such as the AC126 or
CV7001. The OC44 is used because it has
a 15MHz ft (upper frequency limited) and
fairly low noise, but it is too expensive
today. I’ve substituted germanium Japanese
AM radio transistors types 2SA12, 49 or
TR5, an AF118 (shown in Fig.40), is
especially difficult, since it is both a highvoltage (70V), and high-frequency 125MHz
alloy diffused or drift-field device. This
transistor was one of the first to be designed
for the video amplifier in televisions and
was used in the 1963 Perdio Portarama black
and white TV. (See Radio and Television
Servicing, 1963-64 issue, p.523, and also
https://bit.ly/pe-pe21-06). I had to get an
AF118 on eBay and it cost £3.95! I tried an
ASY51, which had the voltage (60V) but
not the bandwidth (1MHz). The amplifier
oscillated at first, so I reduced the feedback
phase-lead capacitor C36 to 330pF. It
worked, but the square-wave response
was poor, as shown in Fig.41.
Some Japanese types, such as the 2SA103
and 2SA358 should work, but that would
detract from the unit’s ‘Britishness’. I did
try a Mullard ASZ20 40V 40MHz switching
transistor (Fig.42) which worked well, but
the VCE was 40V, which gives only a small
margin. This device and the AF118 use
the unusual TO7 case with a fourth lead
(S) for the case/screen lead. This package
is infamous for internal shorts due to tinwhisker growth. Sometimes banging the
unit hard will dislodge them. Disconnecting
the screen lead, which isn’t needed here
The two original output driver transistors
(TR6 and TR7, Fig.37) are also practically
unobtainable, especially TR7, the NPN
device. These are shown in Fig.44. The
AC127Z is an NPN small power device
specially selected for high voltage (50V
Vce) denoted by a ‘Z’ stamped on the top
(Fig.45). A military Texas 2N388A 40V Vcb
(Fig.46) could be fitted, which is better than
an AC127Z and is what Tobey and Dinsdale
used. There is a high voltage AC141, the
AC141H. The Newmarket NKT717 could
also be employed. The 2N35 (Fig.47) was
used in the original Lin design, but only
rated at 25V.
The GET538 (Fig.48) is not a problem
being PNP, and can be replaced by the
ACY17, ACY48, NKT227, OC81Z or
2N1377. Note the unique tapped TO5
heatsink arrangement shown in Fig.49.
The GET538 seems to be a mini power
transistor rated at 1.5A Ic max and is full of
copper heatsinking inside. I suspect such
a highly rated device was used to allow
Fig.43. The TO7 tin-whisker problem can
often be fixed with a capacitor discharge.
This trick has saved many classic radios
using AF114/5/6 or 7 and OC170
transistors. It has to be done every five
years since the whiskers continue to grow.
Fig.44. The two germanium driver
transistors used in the Stereo 30, TR6
and TR7.
Fig.45. The rare AC127Z, the high
voltage NPN TR7.
Practical Electronics | July | 2021
Drivers
39
Fig.46. The mil-spec 2N388A NPN
germanium transistor used by Toby
Dinsdale. The late Bernard Patterson, a
long-term PE subscriber, generously sent
me a few.
for the possibility of high-power, highfrequency testing. For audio applications,
a lower-rated device can be used.
The TD Towers book (Fig.50) is great
for looking up data, but always check the
equivalents suggested because they often
have a lower voltage rating and therefore
may not be suitable. In update 4, the
BC153/4 are also incorrectly listed as NPN.
Fig.49. Tapped TO5 heatsink used for
the GET538. People took great care with
transistor mounting in the 1960s when they
cost as much as a beer and sandwich.
Fig.50. The International Towers Transistor
Selector, a vital book for old electronic
restorers and job-lot buyers of NOS
transistors. The old Mullard pocketbooks
are also useful for quick reference.
40
Fig.47. One of the first NPN transistors
by Sylvania, who specialised in NPN
devices. The 2N35 in its distinctive
SO-4 can from an ancient Lin/Dinsdale
amplifier in an old organ. Introduced in
1953, this transistor cost $4.45 in 1955!
Jack Dinsdale used a similar device, the
Syl 1750 which cost the considerable
amount of 25 shillings (£1.25) from Home
Radio, March 1962.
Selecting for voltage
I’ve used the NPN transistor ASY73 (30V
Vce) in a lower-voltage Dinsdale design. The
same goes for the 2N1302s and 2N1304s. I
use the Peak ZEN50 Zener diode analyser
to test their Vce breakdown voltage to see
if they are suitable by joining the base
and emitter. If the ZEN50 reports a Vce
breakdown voltage greater than 50V it
can go in the Stereo 30. I’m not yet 100%
sure about the long-term reliability of this
technique, but it’s shown in Fig.51 where
VZ is a substitute for Vce. If there’s no lowresistance path, say <1kΩ from the base
to the emitter, or the base is left open, the
voltage rating is reduced to the Vce (open)
rating. The Peak’s current is limited to 2mA,
so no damage to the transistor can occur.
Reducing the power rail voltage below
40V (recommended for 8Ω speakers)
enables easier substitution of transistors.
Fig.48. GET538 (TR6), a hard-to-get PNP
driver transistor, replace with an ACY17.
This is achieved by setting the voltage
selector to 250V (see Practically Speaking,
February 2021). Transistors often have a
higher actual breakdown voltage than
the specification in practice. I’ve done
this for ASY73s to replace the AC127Z
and selected 2SA70s, which have a TO7
case, for the AF118.
Fig.51.Testing transistors for collectoremitter voltage breakdown (Vce) with a
Peak ZEN50 Zener diode tester. This
AC188’s Vce is much higher than its
stated minimum –32V rating, and could
replace the GET538.
Fig.52. The DC bias point presets (P6) for the power amplifiers are hidden in the top
left-hand corner.
Practical Electronics | July | 2021
Power transistors
The Mullard AD140s, as fitted, seem to last
quite well. They have a high Vce voltage
rating of 55V which possibly helps. The
current rating is 3A. The driver and output
transistors should all be germanium. It
doesn’t work pairing a silicon NPN driver
with a germanium output transistor, since
a nasty crossover asymmetry results. A
quasi-comp output stage in silicon actually
has edgier crossover distortion because of
the sharper turn-on characteristic, unless
a Baxandall diode is used, such as in some
Naim amplifiers.
There are plenty of suitable output
transistors, such as OC28, OC36, AD142,
2N2147, AU103, ASZ15, NKT401, AD149
R 2 6
– 2 2 V
+
C 2
2 5 0 µ F
C 1
10 µ F
R 2 1
T R 1
A C 10 7
B C 15 4
– 12 . 8 V
+
R 62
R 13
– 4. 6V
Set D C
o nT R 2
co l l ecto r
I np u t
R 12
R 14
P 1
0 V
F L A T
Sw i tch i ng s i mp l i f i ed
f o r cl ar i ty
R IA A
R 16
R 17
and AL103, but I have a big stock of AD140s
so that’s what I use. The early Truvox amps
used AD140s and later changed to AD149s.
These were germanium’s last fling as output
transistors, before the all-conquering silicon
2N3055 took over.
OC25, OC29 and OC35 can be used in
reduced voltage (35V) low-power Dinsdale
designs. OC22, OC23, and OC24 have good
high-frequency response for germanium
(2MHz), but are only rated at 2A, so they
are only suitable for a few watts. This is
not sufficient for the Stereo 30.
Power output
Testing with sinewaves for more than say 20
seconds at full power into a load can quickly
overheat germanium
– 2 6V
transistors. It’s prudent
to gradually reduce
the load resistance
as testing progresses;
say, from 100Ω to 8Ω.
T o to ne
co ntr o l
Germanium’s maximum
junction temperature is
T R 2
90°C, half that of silicon
G E T 113
B C 15 3
devices. The Stereo 30
will give 10W RMS into
C 7
10 µ F +
15Ω and 14W into 8Ω,
but
not for long. Using
C 8
+ 8 0 µ F
the reduced –38V HT
voltage, it’s 10W into 8Ω.
Try doing this at 10kHz
into a load resistor and
you’ll blow it up. In
R 18
the past, loudspeakers
had an inductive rise
in impedance above
C 5
1kHz, which provided
2 0 nF
protection. Load resistors
and modern speakers don’t have this
characteristic. This is the time to optimise
the DC bias point at around 19V using P6 for
symmetrical clipping (see Fig.52). Do this
into an 8Ω load quickly at 1kHz. Remember,
the heatsinks on most Hi-Fi amps are only
big enough for short bursts of full power.
The preamplifier
The pre-amp channel only uses three
transistors, an exercise in minimal circuit
design, reflecting their cost at the time. This
is shown in Fig.53. There are two parts,
the input + RIAA stage and the Baxandall
tone control/filter. Once the resistors and
capacitors have been checked, it rarely has
faults. A tweak of DC bias pot P1 (Fig.32) is
often needed, which should give a collector
voltage of around 12V on TR2. The AB/
Blore-Edwards pots used are usually okay if
not corroded. Locked spindles can usually
be freed by using power-steering fluid,
basically a light mineral oil with a detergent
in it. WD40 and 151 maintenance spray will
also work. The distortion and noise level are
fairly high. The increase in distortion that
occurs when the bass control is advanced is
subjectively quite nice. The filter is useful
if scratchy records from the charity shop
are played.
Improvements
There are a few reversible improvements one
can make; here are some ideas to play with.
Gain structure
+
The preamplifier line inputs were designed
for low-level 100mVrms outputs from FM
R 63
C 4
tuners. For some of today’s digital sources,
8 nF
such as CD players with 775mVrms level
there is too much gain. Lowering
R16 to 1kΩ can reduce this by
Fig.53. a) (above) The input stage set for the preamplifier RIAA input. Two transistors do not give
almost a factor of four, making a
enough open-loop gain so the distortion is high. The AC107 was considered one of the lowesthigher-level line-input. Even this
noise transistors at the time. b) (below) The tone control stage uses a single transistor. Distortion is
is still too much for a CD player
1% with bass boost, but sounds interesting.
and extra attenuation is needed
C 15
with a potential divider on the
2 5 µ F
B as s
input. The same mod works
R 2 3
L i near
R 2 5
with tape head and microphone
– 2 6V
* N o te new v al u es
inputs, facilities which are not
i n br ack ets f o r l o w er
R 2 6*
mai n r ai l v o l tage s u p p l y
required today. On the eBay
o f ar o u nd – 40 V .
C 9
C 11
O u tp u t
10 0 nF
10 0 nF
amplifier I rewired the switch
– 2 2 V
so the line inputs went into the
C 14*
R 2 7
R 2 8
C 16
R 3 2
R 3 4
tone control input. However,
3
2
µ
F
2 5 µ F
+
this mod is not easily reversible
– 11V
C 13
2 5 µ F
and requires that C13 (polarised
T R 3
C 2 4
C 2 5
G E T 113
O u t
15 nF
5 nF
capacitor) is turned round. It did
B C 15 3
C 12
C 10
12 nF
12 nF
sound much better; even though
– 4. 5 V
In
C 18
the high gain combined with
15
nF
L i near
In
O u t
C 2 6 N o r mal /
attenuation in the preamplifier
R
3
0
C 2 1
C 8
5 nF
Steep
R 2 9
T r ebl e
5 0 nF
+ 8 0 µ F
R 3 3
breaks one of the main rules of
high-quality audio.
+
+
I np u t f r o m
i np u t s tage
0 V
O u t
F i l ter s w i tch : o ny o ne f r eq u ency
s h o w n to s i mp l i f y di agr am
Practical Electronics | July | 2021
R 3 8
In
C 2 7
12 nF
Coupling capacitors
Making the output electrolytics
(C37, as shown in Fig.54) bigger
is useful; 1000µF is only just big
enough for 15Ω speakers, so I
41
Fig.54. The original output capacitors were insufficient in
capacitance and voltage. I installed these big axial Philips/BC
electronics/Vishay electrolytics. The top of these caps is a good
place to do the output half-rail voltage check.
upgraded to 3300µF 40V. I worry about
the original rating of 25V with a 42V rail
should the amplifier fail hard-on. It could
take the speakers out if the capacitor breaks
down. Modern electrolytics are much
smaller and better, so this is an improvement
recommended for all capacitor-coupled
amplifiers.
Make C1 bigger: the circuit diagram in
Fig.53 specifies 10µF, but the 1964 amplifier
originally had just 1µF fitted. The rise in
impedance at low frequencies increased
the noise on the turntable magnetic pickup input. I further raised the value of C1,
and put in a 100µF 6V metal-cased solidtantalum capacitor of 1969 vintage. These
old types of capacitor do not deteriorate
with time, so are an ideal NOS ‘antique’
replacement, as shown in Fig.55.
Silicon sacrilege
Leak did one improvement themselves
a few years later. They replaced all the
germanium small-signal transistors with
what were then the only low-cost PNP
silicon transistors, a BC154 for TR1 and
BC153s for TR2, 3 and 4. They had unusual
TO106 epoxy cases (Fig.56). This reduced
the noise level. TR5 could also be replaced
with a silicon transistor, such as a BC327,
with no real difference in sound quality.
It’s very tempting to try some distortionreducing tricks developed in the silicon
era, such as the Baxandall diode, going
back and applying them to germanium. A
simple mod to the tone control would be
a bootstrapped collector load and emitter
follower. (See G Hibbert, Circuit Ideas, p,89,
Wireless World, April 1980).
Fig.56. Later models
of the Stereo 30
dispensed with
the small-signal
germanium transistors
and used these early
PNP silicon BC153
and BC154 transistors.
42
Fig.55. Old axial solid-tantalum capacitors make a good
replacement for old axial electrolytic types; they seem to last
forever. These ones were in an RAF store for 50 years, all perfect.
Note the much larger, gold-coloured C14 decoupling capacitor.
Constant-current sources
Another ‘high-tech’ addition which I’ve
yet to fully analyse would be to replace
the collector load resistors R21 and R28
with 2.7mA current-regulator diodes (CRD)
reducing distortion by increasing the openloop gain. Another 2.7mA CRD in the
collector load (R50) of TR5 may reduce the
dreaded crossover distortion. It’s essential
to get these CRDs the right way otherwise
they are almost a short circuit and you will
blow your transistors up.
sound as good as a complementary silicon
amplifier, having its own unique slightly
‘woolly’ sound, but it’s good enough for a
second Hi-Fi system used at low volume for
the home office, especially a 1960s-themed
office with an Olivetti typewriter! I’ll be
using it in lectures on amplifier design, so
it will earn its keep.
Next time we’ll have a look at restoring
older silicon designs, such as the millionselling NAD amplifier series.
New sockets
On the eBay amplifier the power cord had
been cut off, so I put in an IEC socket in
place of the mains outlet sockets, another
facility no one uses today. The new socket
can be fitted in the exiating rectangular hole
by just enlarging the screw holes to 4mm.
This mod is shown in Fig.57. New phono
connectors were also fitted with simplified
switching, shown in Fig.58.
Odd distortion
I noticed a DC bias shift with signal level
causing lower cycle clipping after a loud
burst. This was a strange form of ‘delayed
distortion’ that took a lot of tracking down.
The cause of this was power supply droop
on the preamplifier. There is no regulation,
only RC decoupling, so when a loud
transient is reproduced the preamplifier
DC bias can drift about as well. This effect
was reduced by altering the decoupling
components, reducing R26 to 150Ω and
increasing C14 to 1000µF (Fig.53b, Fig.55).
The lower resistance was necessary to
maintain the voltage to accommodate the
reduced rail voltage of 38V. Of course, a
regulator could be installed.
Fig.57. A lot of dumped equipment will
have had their power cords cut off by
overzealous technicians, especially in
the public and educational sectors. It’s a
good excuse to then add a proper IEC
mains connector.
Was it worth it?
Economically, it doesn’t make sense. For
several days work the value has increased
from £26 to £60. However, I love it for its
physical embodiment of early transistor
technology and its build quality. It doesn’t
Fig.58. New phono connectors were
required on the eBay amp, but were a pig
to fit. I had to use the insulating plates
from the old phonos.
Practical Electronics | July | 2021
|