2 Meter SSB unit refurbishment
I designed and made this
unit back in 1979 or early 1980 as part of my 2m transceiver
almost without any test equipment, probably relying on what I
could hear on a 10MHz receiver so I'm unsure of how exactly it
will perform. It uses the phasing method of generating SSB.
The pcb is mounted in
an aluminium chassis screwed very securely to the main chassis
of the 2m rig. Clearly not intended to be serviced because, once
fitted it was wired into place.
By cutting three coax cables
and a few wires I was able to remove the pcb.
Lots of additional components
are added so I must have had some trouble with balancing things.
There are two relays added and I guess these must be associated
with receive/transmit? As I've had to strip out all the rigs
scruffy wiring plus a pair of large relays and the mode switch
it means redesigning the way the various modules are interconnected.
The underside was etched with copper
chloride solution and the pcb has four screws to space it from
its box together with several more screws on which to mount a
screening panel, now missing.
I suppose I needed this
old chassis to prevent stray pickup of RF as the final amplifier
was pretty potent.
The chassis has evidence of
previous use, which I recall was a top band transmitter using
6C4 valves made in 1961, following a visit to the shack of Alan
Melia G3NYK (another Liverpool University student) back in 1961.
Clearly the chassis was pretty old when I re-used it.
On the right is the regulator
providing 11 volts for powering all the modules, many of which
include their own 5 volt or 6 volt regulators.
The good news is the special
miniature audio transformers all appear to have intact windings
so testing can proceed. Lots of tarnishing is present but soldering
looks in good shape so let's hope the corrosion is merely cosmetic.
Below are a pair of drawings
that I found of the original design (and now marked up in Red).
I can see the three audio transformers and 7 transistors on the
first drawing and 6 on the second and on the pcb there are 13
transistors so the pcb looks like it's based on the circuits
As far as I can make out
(corrosion permitting) the transisistors are as follows:- HP
5-410 (TR1/TR4/TR5), 2N5087 (TR2),
BC109 (TR3/TR6/TR7/TR8/TR9), 2N4393
(TR10/TR11/TR12/TR13). From memory I think the HP code for a
FET was "5" but I can't see what a 410 or perhaps 41C
is, but as it's in an audio circuit I guess its spec isn't too
L2 will tune to the incoming
10.7MHz crystal oscillator to minimise harmonics. L3 likewise.
I'm a little concerned the input circuit around TR8 might easily
overload and produce square waves, something I probably wouldn't
have checked back in 1980. I powered the pcb from it's design
voltage of 11 volts and it drew only 10mA. I connected 1KHz audio
and 10.7MHz and looked at the output, initially using my oscilloscope
and found audio was present at the collector of TR2 and 10.7MHz
at the output of L3. Tuning L2 proved it peaked but only just
and after some checking I found TR9 had a broken collector lead.
Some transistor legs were badly corroded and TR9 was worst. I
fitted a new BC109 and the RF output tuned and I was able to
reduce the signal generator output by 20dB. I'd been looking
at a signal from leakages instead of via the circuit. Despite
further checking I could see no evidence of modulation at the
output so turned on my HF receiver tuned to 10.7MHz and sure
enough found it was an unmodulated carrier when connected to
L3. After a few minutes I found that flexing the pcb in the audio
area suddenly produced a 1KHz AM signal at the receiver so I
swapped the lead over to my spectrum analyser (see below). Clearly
either a dry joint or another corroded transistor leg needs fixing.
This picture shows 10.7MHz
double sideband with only 10dB carrier suppression. You can also
see harmonics of the 1KHz audio. At this stage I haven't looked
at the amplitude of the audio input so it may be far too high
and producing distortion in the amplifiers.
There are lots of loose components
around the filter areas where previously I'd temporarily soldered
parallel bits and pieces so at this stage I'm not too worried
about balancing etc.
Before proceeding I need to
locate and fix the intermittent fault.
At this point I replaced
the pair of transistors TR6 and TR7 (TR7 had the corroded leg)
with a pair of new BC109, but I took the precaution of testing
a batch first. I found their gain varied from 250 to 600 so chose
a couple showing a gain of close to 420 in case this helped to
get a better audio phase shift match. I also had to familiarise
with the different trimmers and pots so I could more easily make
appropriate adjustments and mark up the old circuits. With the
audio transistors replaced I retested with the results below.
This is a picture of the
output after adjusting the twin carrier balance pots. Not too
easy because noise induced when twiddling the pots needed time
for scans to stabilise. It's certainly improvable but this is
a fair example with an audio input of around 1KHz (f1) giving
equal carriers 1KHz either side of the RF input of 10.7MHz (f0).
Unfortunately the signal is
well nigh perfect DSB so I'll need to check out the audio circuitry.
Carrier suppression here is
about 29dB and the 2nd harmonic is about 30dB down.
Then I reduced the test
tone to 474Hz and the two signals are about 1KHz apart with an
indicated carrier suppression of around 31dB.
As you can see there's absolutely
no suppression of either sideband.
Disregard the signal strength,
as in most measurements I make with my Rigol, because I'm using
an uncalibrated high impedance probe, and at the sort of levels
I'm dealing with here and the unscreened setup, the baseline
probably includes all sorts of local signal pickup (making it
tricky to make suppression adjustments).
I added test points when I made the
SSB pcb so these should come in useful to make the adjustments.
As I understand the design there is an incoming amplified microphone
signal which is split into two differently phased outputs exactly
90 degrees apart. By using an oscilloscope I should be able to
adjust the pots to line these up so with a sinewave input they
are close to the correct phase difference, then use the RF output
to make final adjustments. The changeover switch should be able
to flip from USB to LSB and vice versa.
These pictures were taken
at the inputs to transformers T2 & T3 from TR6 & TR7
emitters, taking care to limit the audio input to the pcb to
prevent visible distortion. They are at 3KHz, 1KHz and 300Hz
as shown with the amplitudes nicely peaking at around 1KHz and
the balance pot set to equalise their amplitudes. Although the
phase difference between the pairs of signals seems to be about
90 degrees, the balancing pots have no discernible effect and
both RF sidebands are equal, so something is amiss.
This is the scan when
3KHz audio is input to the pcb.
Carrier suppression seems to
be dependent somewhat on the amplitude of the 10.7MHz input from
the HP8640B so the balance pots will need setting up when the
proper local oscillator is used.
Carrier suppression looks like
This picture was taken
with the same pcb settings at that for 3KHz but with 300Hz audio
and a scan bandwidth of 3KHz.
Switching off the audio results
in visible leakage from the cables and circuitry at the RF input
and no doubt this leakage is superimposed on the scans which
implies the supression might perhaps be around 40dB.
After searching for information
on the audio phasing circuit I selected back in 1979 I found
it was called a B & W 350 2Q4 circuit and it's precise resistor
and capacitor values matched those marked on my original sketch.
Because if the difficulty of winkling out precise resistors from
my junk box I'd used slightly lower values and added small pots
to fine tune their values. My next step should be to measure
each one in situ and adjust the corresponding pot to read the
precise value. This should be possible as the four resistors
are buffered by capacitors and the two FET gates. It would seem
that not only are the values critical, but the inputs to the
two halves of the circuit should be driven from different voltages,
explaining why R13 (110 ohms) and R14 (390 ohms) are not matched.
For the moment I'll leave the capacitors because I'm fairly sure
I was able to select USB or LSB all those years ago.
I decided to check the audio
circuit and adjust the resistors to their correct theoretical
values. Surprisingly they were all pretty close but twiddling
the four pots set them precisely. I checked the circuit between
the twin output transformers and the four FETs in the balanced
mixer and thought I'd found a mistake.. it turned out I'd altered
the connections to the USB/LSB switch (making the FETs go to
the switch wiper instead of T2 output) but the alternative was
fine. I've modified the sketch above to reflect this change.
Everything checks out so the next step is to look at the waveforms
in the balanced mixer to make sure the audio signals are the
correct amplitude. The phase of the transformers must be OK as
they can be switched round using the USB/LSB switch and in the
early tests the position of this didn't change the DSB output.
As you can see I finally
saw some LSB rather than DSB, albeit with only a 16dB difference,
and second harmonics from the audio test tone.
Just to be shaw, I flipped the
USB/LSB switch and saw the picture below...
At this point I managed
to save this picture but then the thing reverted to its DSB condition.
I poked around and tapped everything
but when I bent one of the FETs slightly SSB came back and I
was able to make adjustments.
Still only 16dB sideband
suppression and 28dB carrier suppression, but at least it's now
stable and I can start to make proper adjustments. I reduced
the audio test tone and the signal cleaned up nicely.
Swinging the audio from 300Hz
to 3KHz didn't change the comparative amplitudes of the signals.
I was very impressed with the
stability of my repaired HP8640B and I was able to tune its output
to position the traces on the screen when the Rigol was looking
at a scan bandwidth as low as 5KHz at 10.7MHz.
As the Rigol isn't looking
at valves and high voltages I might connect it directly to the
output (the coupling coil in L3) and measure the true output
across 50 ohms instead of using the high impedance probe in the
diecast box above. The signals may be easier to check if the're
not buried in noise. On the left, above, is my audio attenuator.
Because the mic amplifier has a very high gain, the output from
my cheap audio oscillator swamps the input,even when turned to
zero, so I'm using the attenuator which gives me up to 80dB of
attenuation. I incorporated a gain control on the pcb when I
first built it but decided to extend the wiring to a front panel
control later, leaving a second, preset gain control on the pcb,
to enable me to match the microphone output. The setting of this
preset is important because too much audio signal results in
distortion leading to spurious RF sidebands.
I was busy for a while but returned
to the job of testing the SSB generator. An intermittent fault
was plaguing operation with the loss of one audio channel at
the balanced mixer. I noticed the audio output transformer appeared
to be slightly loose and with some trouble removed it from the
circuit board. I found one of its pins was broken between the
transformer and its solder joint. The evidence pointed to a repair
I must have done when I'd first fitted it. Not only was the pin
broken off but a thin wire had broken from a second pin. Because
the two output transformers need to be matched I had to make
a repair so soldered 5 wires to the original pins, repaired the
wire fracture and refitted the transformer. Thankfully it worked
fine and I now have reliable, switchable sidebands, albeit at
best with only 28dB supression. After experimenting I think the
phase-shift resistor values are too high so that the pots don't
null out the unwanted sideband. Either that or one or more of
the high accuracy capacitors is out of spec so I now need to
remove these and test them for accuracy.
The RF phasing works fine
with a respectable nulling point of the carrier at each of the
two adjusting pots but the audio phasing varies between only
18dB and 28dB. As shown here the values of the four resistors
are preset by using a small pot plus a fixed resistor. Assuming
the capacitors are OK the resistors (top to bottom) are calculated
to be 487K, 770K, 198K and 125K. The pcb has test points so I
used these to set the resistance values from the nominal values
shown here. Amplitude balance is made via VR6 shown in the picture
This was about the best
performance I got for USB, although there were lots of interactive
factors including the 10.7MHz and 1KHz test signal amplitudes.
I swapped the fixed capacitors in the phasing circuit and in
doing so discovered the 30pF capacitor (part of C13=430pF) read
open circuit. At one point I think the suppression of the carrier
and unwanted sideband scraped down to 40dB.
Once the crystal oscillator
is connected the results will be more dependable.