This is only a preview of the May 2023 issue of Silicon Chip. You can view 57 of the 112 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 "Dual RF Amplifier":
Items relevant to "UVM-30A UV Light Sensor":
Articles in this series:
Items relevant to "GPS-Disciplined Oscillator":
Items relevant to "Wideband Fuel Mixture Display, Pt2":
Items relevant to "Songbird":
Purchase a printed copy of this issue for $11.50. |
Vintage Radio
Astor’s first transistor radio –
The APN
By Ian Batty
While Astor was beaten to the transistor radio market in Australia by
AWA, first is not always best. The circuit used in Astor’s APN became the
template for many Australian transistor sets that followed it.
With a smaller case and superior
electronic design, the APN is a notable ‘first outing’ for the famous Astor
brand. As with other Astor sets I’ve
reviewed, mass production and popularity don’t necessarily mean cheap
and sloppy design. This radio’s performance is comparable to other contemporary sets, while its visual design is
unmistakable.
While I can’t claim the misguided
genius of Viktor Frankenstein, my
radio (see Photo 5) is an assemblage of
parts. It’s an all-transistor APN chassis
in a case taken from an all-valve BRQ.
100
Silicon Chip
Luckily, Astor didn’t make too many
changes to the BRQ’s case when they
reused it for the APN, so unless you
know what you are looking at, you
probably won’t notice the substitution.
The APN is one of a family of radios
released in the changeover from valves
to transistors. These continued the
valve sets’ visual designs but popped in
transistor-based circuitry. A good number used identical cases. Geoff Trengove and Jim Greig published part one
of a two-part series in the January 2023
issue of Radio Waves that included a
complete survey of such sets.
Australia's electronics magazine
The only difference between the
case for the APN and the BRQ is that
the former lacked the hole for the On/
Off/Battery/Mains switch on the right
side, and the power cord cutout in the
rear flap (because the BRQ was mains-
powered but the APN is a battery set).
The APN also bears the label
“TRANSISTOR” in place of the BRQ’s
“SPORTSTER” beneath the speaker
cutout. There’s also a label above the
volume control on the BRQ labelled
OFF ON VOLUME, which is missing
from the APN since it has no such
switch.
siliconchip.com.au
Internally, the APN uses a separate
sub-chassis for the audio and RF/IF
sections – see Photo 1. The RF/IF section is on the right, where the BRQ
signal circuitry was located (Photo 2).
While three transistors occupy a lot
less space than four valves, the APN
adds a third IF can, so the space savings are not huge. The three-transistor
audio section sits in a previously-
unused space at the top of the case
(Photo 3).
The APN replaces the BRQ’s mains
power supply (with its transformer
and large filter capacitors) with a
parallel pair of 276P 9V batteries.
Aside from the transistors, all components are of similar size to those
in the BRQ.
Circuit description
You might think that the APN circuit (Fig.1) looks pretty much like
any other six-transistor set. In fact,
the APN set the template for Australian transistor sets, with a self-excited
converter, two IF stages and a diode
demodulator with AGC to the first IF
stage. The audio section comprises an
audio driver and transformer-coupled
Class-B output.
The APN’s performance rivals that
of the look-alike BRQ four-valve set.
Astor’s APN showed AWA’s 897P to be
a mediocre design, as the 897P needed
seven transistors to give only marginally better performance.
Astor drawings simply number components in order. Items #1 to #22 are
resistors, #26 to #45 are capacitors, #50
to #56 are inductors, #57 is the battery
and #58 is the speaker. I’ve preserved
this scheme to prevent confusion;
however, I’ve numbered the transistors
Photo 1: the interior of the Astor APN is divided into two separate chassis for
the audio & RF/IF sections. The audio section is primarily above and around the
Rola speaker, while the RF/IF section is located on the right, as shown by the
large IF cans mounted horizontally.
Photo 2: the APN chassis metalwork is based on (and nearly identical to) the allvalve Astor BRQ, with the BRQ shown here for comparison.
Photo 3: a closer look at the audio section of the APN. Astor decided to utilise the empty space below the ferrite rod
antenna to house the components.
siliconchip.com.au
Australia's electronics magazine
May 2023 101
Q1~Q6 and the demodulator D1, as
Astor omitted such labels.
The local oscillator uses collector-
emitter feedback, operating the oscillator transistor in a grounded-base
configuration, guaranteeing reliable
oscillation across the broadcast band.
As the base is not in the oscillator circuit, local oscillator radiation via the
antenna circuit is minimised.
Q1’s forward bias seems too low
at only around 70-100mV, but that’s
because Q1 runs in Class-B, giving it
the nonlinear operation vital to the
mixing function.
Australian manufacturers generally
used tuning gangs with identical sections, necessitating a padder capacitor to get the LO to track the antenna
circuit. In the APN, this is #30, a fixed
310pF capacitor.
The converter feeds the tuned,
tapped primary of the first IF transformer, #52. Tapping the primary
allows the transformer to exhibit a
high Q factor without its tuned circuit being damped by Q1’s relatively
low output impedance, typically
under 50kW.
The first IF amplifier (Q2) uses simple capacitive neutralisation thanks
to 6pF capacitor #35. This eliminates
the feedback effects of its inherent
collector-
base capacitance (see the
panel for more details).
The voltage drop across 330W emitter resistor #5 indicates a standing
collector current of around 0.6mA
(600μA). This will fall as the AGC circuit acts to reduce the first IF’s gain on
strong stations.
Q2’s bias circuit uses a high-value
resistor from the supply (#12, 100kW)
so that the AGC voltage (supplied via
#11, 2.2kW) can effectively control
Q2’s collector current and thus, the
stage gain.
The second IF stage (Q3) uses fixed
bias, with a standing collector current of just over 1.3mA. This stage is
not neutralised, perhaps due to the
demodulator loading the third IF transformer (#54), giving a lower gain.
Both IF stages have their bypassing
(base and IF transformer) tied back
to their emitters. This single-point
method gives highly effective bypassing and reduces the component count
by eliminating the usual emitter
bypass capacitor.
Diode demodulator D1 feeds
demodulated audio to 5kW volume
control potentiometer #13 and, via filter resistor #11, to the AGC line. The
AGC line is filtered by 15μF capacitor #33, removing any audio signal
and producing a simple DC control
voltage.
The audio signal feeds to audio
driver transistor Q4 via 2μF coupling
Transistor Neutralisation
Some textbooks describe neutralisation in terms of feedback. Capacitor #35
applies positive feedback from the collector’s tuned circuit to the base.
I verified this by increasing #35 to 10pF. That doubled the sensitivity compared to the recommended circuit, confirming that the neutralising capacitor
applies positive feedback.
It was tempting to ‘hot up’ the APN to equal the 897P’s superior performance
this way, but I resisted.
Consider the effect of the transistor’s collector-base capacitance; since the
collector signal is an amplified, inverted version of the base signal, collector-
base feedback is negative. The point of the positive feedback from capacitor
#35 is to cancel this out. So you can think of neutralisation as adding a balancing circuit that nulls out the effects of collector-base feedback.
I addressed the matter of anode-grid feedback in valves in my article on
the Grebe Synchrophase radio in the February 2018 issue (siliconchip.au/
Article/10977). The same principles apply to transistor circuits, except that
some designs account for transistor feedback’s complex nature.
While a valve feedback’s phase angle is ideally 180º, transistor feedback
deviates from this as the internal feedback contains resistive and capacitive
elements. A simple capacitive circuit cannot totally counteract such a complex feedback effect. Full correction demands a resistive-capacitive neutralising circuit, properly known as ‘unilateralisation’.
With unilateralisation, the signal in the amplifying circuit flows only from the
input to the output and never in the reverse direction. Regency’s TR-1 (described
in the April 2013 issue; siliconchip.au/Article/3761) uses such a design.
102
Silicon Chip
Australia's electronics magazine
capacitor #40. Q4 uses combination
bias, with a voltage divider formed
from resistors #14 and #16 and 1.8kW
emitter resistor #17. There is a feedback path from the speaker connection via resistor-capacitor combination #18/#43 and series resistor #15.
Transistor Q4 feeds driver and
phase-splitter transformer #55, with
top-cut provided by 4.7nF capacitor #41.
The output stage Q5/Q6 operates
in Class-B, with around 150mV of
bias provided by resistive divider
#20/#21 and thermally-compensated
by NTC thermistor #19. Q5 and Q6
share emitter resistor #22 and drive
output transformer #56, which in
turn drives speaker #58. 47nF capacitor #44 applies top-cut to the output
transformer.
Class-B operation provides better
efficiency than Class-A. Of the transistor sets I’ve tested for Silicon Chip that
use Class-A, only one manages even
30% efficiency (the GE P807).
Terman (siliconchip.au/link/abje)
quotes Class-A’s maximum theoretical
efficiency as 50%, with typical values
of 20-35% (p391). The same source
puts Class-B’s maximum theoretical
efficiency at 78%, with common values of 50-60% (p393).
The APN’s Class-B output stage
gives a maximum power efficiency
of around 46% for full audio output,
which may not seem like much of an
improvement over a good Class-A
stage. But the APN’s full output comes
with a battery drain of around 63mA
on peaks.
A 250mW Class-A output stage
(with an efficiency of 30%) implies
a standing power consumption of
siliconchip.com.au
Fig.1: a redrawn circuit diagram for the Astor APN. It uses just six transistors, one less than the competing AWA 897P. The
Class-B output stage and dual 9V batteries gives a typical runtime of 200 hours.
750mW. This would give a constant
battery current approaching 85mA,
resulting in under 50 hours of operation from the pair of 276P batteries. Such an output stage would also
demand extensive heatsinking and
very precise biasing to prevent thermal runaway.
As the set will rarely be run at full
volume, the resulting average battery
drain is much lower. Average listening levels allow a battery life exceeding 200 hours.
Why two 9V batteries in parallel? I
suspect two reasons – first, there was
enough space, given that they removed
the mains power supply used in the
previous valve model. It would also be
a marketing point, as the APN would
give about 20 times the battery life of
the previous BRQ valve set.
Restoration
The case cleaned up nicely, with
the oddity that it appeared to be a
case from the previous valve model
(BRQ). The electronics were another
story. It did work – just. Sensitivity
was very poor, and it only seemed
to tune from about 700~800kHz to
around 1500kHz.
The original metal can (TO-5 package) 2N484 converter transistor had
been replaced by an all-glass OC44,
and the original CK872 TO-5 audio
driver was replaced by a TO-1 package 2N406.
The audio output was distorted, so
I first checked the output stage bias,
which was too high. I tried removing
the bias thermistor #19, but one lead
siliconchip.com.au
broke off from the resistive body. It was
not repairable, so I replaced the bias
circuit with a diode-connected transistor (see Fig.2). This has the advantage of giving the correct bias voltage
that tracks correctly with temperature
changes.
The audio output was still low,
going into clipping at under 100mW,
and I wasn’t getting the expected
50mW output with 5mV at the audio
input. The original CK878s showed
very high leakage, so I replaced both
with AC128s. I was able to disconnect
the CK878s and leave them in place,
preserving some visual originality.
The volume control coupling capacitor (#40) measured low in capacitance, so I replaced it. I could then get
50mW of output with only 4mV input
– about right for a three-
transistor
audio amplification stage.
I then looked at the IF channel.
I’ve previously warned against using
paint/wax/other stuff for sealing
adjustment slugs. This set had wax
poured into the tops of the three IF
cans, and the slugs were held tight.
Maybe it was still in alignment, and I
was just being fussy.
Still, I thought the sensitivity was
low, and I measured 2.5V DC at Q2’s
emitter. It should have been about
0.15V; the problem was excessive collector leakage in Q2. So I replaced both
Q2 and Q3 with OC45s.
That fixed the excessive emitter
voltages, which should have meant
that the IF channel was working correctly again. The remaining low gain
prompted me to try removing the sealing wax, so I removed and dismantled
the three IF cans.
Whatever the ‘foreign’ wax was, it
had a much higher melting point than
the manufacturer’s wax used to seal the
coil windings. My heat gun had the
wax on the windings dripping while
the wax in the coil cores was only just
softening.
Rather than overheat the windings,
I boiled a kettle, poured the water
into a jug, and dunked the coil. This
worked well enough with IF2 and IF3
Fig.2: I replaced the NTC thermistor with a diode-connected transistor (right) to
provide the correct bias voltage with respect to temperature.
Australia's electronics magazine
May 2023 103
Photo 4: the LO coil is
not easily adjustable on
the APN (shown at far
right), despite it having
an adjustable slug. I had
to spend quite a bit of
time cleaning the wax
out of the three other
coils that someone else
had added, so that they
could be adjusted.
to let me carefully extract the adjusting slugs with several ‘treatments’. I
visited a machinery shop and came
home with a ¼-inch, 26 thread-perinch (TPI) tap and die.
The tap worked a treat. Held with
no more than finger tension, I was able
to clear the coils’ internal threads of
wax gradually. Heating and swabbing
the threads with cotton tips was not
an option because I didn’t want to risk
damaging the coil windings, and heating the wax would have allowed it to
coat the internal thread evenly, worsening the problem.
Curiously, although the tap seemed
a correct fit to the coil thread, the
adjusting slug would not drive into
the matching die. There was definite
interference, so I resorted to a fine wire
brush to clean the slug threads. I managed to get IF2 and IF3 adjustable, but
IF1 defied all my attempts. Luckily,
the slug was well out of the coil, with
the IF resonating at close to 520kHz.
I first tried the easy way – bridging an extra capacitor across the primary of IF1 (in this case, 68pF). While
this brought the resonance down to a
bit below 460kHz, the resulting gain
appeared low.
That makes sense; Q = (1 ÷ R) × √L ÷
C, so a larger C, for the same L, reduces
Q and thus, stage gain. My back-ofthe-envelope shows an expected Q
reduction of about 15%, close to what
I measured.
So instead, I recovered a suitable
slug from an old TV coil and popped
it in. Luckily, the jammed slug was at
the bottom of the IF, so the new one
screwed easily into the top of the winding, and I could bring the IF down to
455kHz.
I still didn’t have the gain I expected.
104
Silicon Chip
The AGC filter capacitor (#33) was
open-circuit, so I replaced it. The second IF amplification stage showed a
low gain; the culprit was #37, the emitter bypass. It’s unusual to find a paper
capacitor open-circuit, but I did, so be
alert to that possibility.
With the IF going, I looked at the
converter stage. The ferrite rod’s leads
must have broken at some point, as
they were soldered to single-strand
hook-up wire. I replaced the connections with flexible stranded wire and
protected the joins with heatshrink
tubing.
After replacing the existing OC44
with one from my spares box, I found
that the local oscillator would not
work. I suspected the emitter coupling capacitor, #29. Remembering
the faulty capacitor #37 in the IF strip,
I replaced #29. The oscillator would
still not work, and after much faffing
about, I pulled my substitute OC44 and
tested it. Its current gain (β or hfe) was
only about 30.
So I tested all the OC44s I had on
hand and selected one with a β over
100. That got the set going at last.
I was surprised to find that the oscillator transistor’s gain was so critical.
OC44 specifications show a β range
of 45~225, with 100 typical. Yes, my
replacement had a β of only 30, but
I’d have expected the designers to be
pretty liberal and allow for low transistor gains.
As with valve sets, it looks like the
converter is the stage most sensitive
to device performance. Perhaps they
selected the OC44s for gain at the factory.
With all that done, I was able to
finish the alignment and complete
my tests. The ferrite rod has a small
Australia's electronics magazine
auxiliary winding that can slide along
its length to adjust the antenna circuit
at 600kHz. While this works, I’d be
careful not to ‘exercise’ it too much, as
I expect the coil wiring to be delicate.
One final niggle: the LO coil cannot be adjusted on this set. Yes, it
does have an adjusting slug, but it’s
obscured by a ferrite rod mounting
bracket (see Photo 4).
Transistor failures
I’ve never had to replace every transistor in a radio. The APN is a reminder
that transistor technology was advancing rapidly in the 1950s and 1960s, and
didn’t really mature until silicon transistors became mainstream. You can
still buy OC44/45s online, but you’ll
likely get a better deal from the HRSA’s
Transistor Bank (visit hrsa.org.au for
more information).
While it’s often possible to rejuvenate valves by over-running the filament/heater, I’ve not found any similar technique for transistors.
That makes sense: valve emission
depends on the chemical composition
of the cathode coating, so it’s possible
to ‘boil off’ contamination by overrunning. However, a semiconductor junction is intimately fused in manufacture, and degradation that increases
leakage is unlikely to be remediable.
There are two significant measures
of leakage, ICBO and ICEO. ICBO is the
current flow measured from collector
to base (“CB”) with the emitter not connected (“O”), while ICEO is the current
from collector to emitter (“CE”) with
the base not connected.
An ICBO of, say, 10μA might seem
trivial, but it’s a base current, so the
transistor’s current gain can magnify
this to a collector current of 100μA or
siliconchip.com.au
considerably more. This would affect
the ICEO. The leaky 2N484 in this set
had an ICEO exceeding 10mA with a
Vce of 10V.
How good is it?
For a first outing, it’s pretty good.
The best comparison is AWA’s first
transistor set, the 897P, which I previously reviewed (April 2015 issue;
siliconchip.au/Article/8458).
The 897 used seven transistors, with
two interstage audio transformers for
maximum gain in the four-transistor
audio channel. This appears to be in
compensation for the low overall gain
of the RF/IF section. The 897’s audio
gain is over ten times higher than that
of the APN, so let’s keep that in mind.
The APN’s RF sensitivity is 375μV/m
at 600kHz and 200μV/m at 1400kHz.
Both readings showed signal+noise to
noise (S+N:N) ratios better than 20dB.
Compared to the 897’s 250μV/m and
150μV/m, and discounting for the
897’s extra audio gain, the APN’s RF/
IF section has more gain overall.
The APN’s actual performance is
certainly on par with valve portables
of the day. My favourite ‘distant’ station, Horsham’s ABC Western Victoria (3WV) on 594KHz, rocked in at
full volume.
The maximum audio output, at
10% total harmonic distortion (THD),
is 260mW. At 50mW, THD is about
5%; at 10mW, it falls to 3%. The -3dB
audio response from the volume control to the speaker is 260-4600Hz,
with a peak of about +4dB at 1050Hz.
From the antenna to the speaker, it’s
150-1900Hz.
For a +6dB output rise, the signal
increase was around +25dB, about as
good as can be expected with the simple AGC used in the APN. It went into
signal overload at around 25mV/m,
which is a strong signal.
-3dB selectivity is ±2.5kHz; for
-40dB, it’s ±14.5kHz. This selectivity
compares well with the 897’s figure
of ±13kHz at -60dB, allowing for the
897’s double-tuned IF transformers.
I tested it at only -40dB because
much over this put the APN’s converter into overload. It gave reliable
results at -40dB, and the ±14.5kHz
skirt selectivity is enough to reduce
interference from any adjacent channel radio station.
Its low battery performance is good.
Although its sensitivity reduces with a
supply voltage of 5.5V, it still exceeded
50mW output with low distortion.
This low distortion justifies my bias
diode replacement for the failed voltage divider/thermistor circuit.
Purchase recommendations
I’m looking for a good original
case with a wrecked chassis to de-
Frankenstein my example (see Photo
5). If you have AWA’s 897 in your collection and don’t have an APN, consider getting one.
It’s a bit smaller, with – to me – a
more interesting visual design. As an
engineer, I appreciate its comparable
performance to that of the AWA, especially given that it has one less amplifying stage.
Jim Greig restored a genuine APN
(described in the HRSA Radio Waves
magazine, October 2020) and found
similar faults to mine. Jim’s method
of fault-finding is a valuable reminder
that different repairers use the basic
principles differently.
According to the Radio Waves article in January 2023 referenced earlier, Geoff and Jim have only discovered one issue of this radio, in the red
case (see the lead photo). So if you see
an APN in a different case, it’s likely
another Frankenstein’s monster.
Special handling
The tuning and volume knobs are
a press fit onto the capacitor shaft. I
recommend that you don’t use screwdrivers or other levers to remove them.
I was able to use finger pressure; if you
can’t get them off that way, run strings
under the knobs and use a gentle pull
SC
to remove them.
Photo 5: my
‘Frankenstein’
Astor APN came
in a leather case
originally for a
similar valve set. It
is different from the
‘original’ red leather
case shown in the
lead photo.
siliconchip.com.au
Australia's electronics magazine
May 2023 105
|