[HBR] Another Receiver Project -- HBR-4, Part 15

[waltah at earthlink.net]

Slow progress ...

The front panel is painted and marked.   It's not the best job I ever 
did -- lots of troubles with the darn rub-on letters not sticking 
properly and some of the layout was tricky -- but it turned out okay.

The solution to no-stick letters was to buy new, from OSE.   Why 
didn't I think of that earlier?   Now I have to track down some 1/2" 
letters for the nameplate.

Radio Shack has the %$#@ little bulb for the (ex-FT-101) tuning 
dial and index.   Every FT-101 owner in the world hates that bulb, 
because you have to remove the front panel to change it and to 
remove the panel you have to take off the dial.   Reassembly is a 
mess ... But it's done now.

The 25 kcs calibrator works.  It's a 6J6 astable multivibrator, plate 
coupled, with synch to the 6EH7 100 kcs calibrator via 5 mmf from 
the 6EH7 plate to one of the 6J6 plates.   There were some 
interesting problems.  

Depending on how the synch was done, some configurations 
generated a lot of broadband noise -- basically phase noise, I think, 
coming from a too-broad synch pulse in some configurations.  

At first I had the mv plate resistors so high that there wasn't time to 
fully charge the coupling capacitors.   What an unstable mess!   

But once the circuit was right, stability, surprisingly, wasn't a 
problem.   My Navy 'Handbook -- Preferred Circuits Navy 
Aeronautical Electronic Equipment' (1963) says that a properly 
designed multivibrator has better stability than a blocking oscillator 
and is preferred as a radar pulse repetition frequency generator.

For best stability the grid resistors return to the plate supply rather 
than to ground.   That way the grid voltage rate of change at the 
moment of triggering is much greater and the frequency jitter 
proportionately less.   Hummm ... I turn the calibrator on/off by 
switching the cathode, probably meaning that (with the grids 
returned to HV) the heater-cathode voltage rating is exceeded.

On the high bands, the amplitude of the markers varies 
considerably -- they may jump from S-7 to S-9+20, then fall back.   
This is because the strength of the markers depends on the 
multivibrator waveform which depends on the exact phase angle at 
triggering.  Thus the strength of the markers depends on the phase 
angle between the two oscillators.   Although the phase angle is 
closely fixed at the fundamental (the 25 kcs *is* synched, after all), 
the waveform difference can be enough to cause a big change in 
the harmonic output on the 1000th+ harmonic (10M).  And it's not 
constant because the receiver plate voltage varies about 3 volts 
from zero signal to S-9+20 and that shifts the free-running 
multivibrator frequency, meaning the phase angle shifts.   So tuning 
in a marker changes the phase angle slightly which changes the 
marker strength.  

It works fine, but it's not very polished.   I should clip the 
multivibrator waveform to make the triggering distortion go away.   
There's something like 100 volts p-p signal, so there's plenty to 
work with.  But probably not now.

Another round with the VFO.  The audio clarity still isn't as good as 
I'd like and one of the main places to look for problems is the VFO. 
Improving the B+ filtering helped slightly.  I then rewound the coil 
using #25 instead of the #30; that made a slight further 
improvement.  At the same time I went two more rounds with 
temperature compensation.   It's getting there -- after the first 
minute or two you really can't tell that it drifts at all for about 20-30 
minutes when an SSB conversation is perceptably off.   The drift is 
very steady though, so it can still be improved.

Right now there's a 7.5 mmf N150 cap between the 6J6 plates at 
the socket and 8+3 mmf N150 between the plates inside the tank 
circuit box.   It needs very little more and moving the inside caps 
nearer to the tube connections so they'll heat more quickly could 
be enough.

The voice quality problem is a tough one.   Instability of any of the 
three oscillators will do it, ditto any parasitic (all these tubes will 
oscillate into the VHF), any clipping due to overload of an IF stage, 
and non-linearity of the detector or audio stages.   Parasitics 
(usually ultrasonic) in the audio section are another possiblity when 
lots of feedback is used.   Inadequate filtering of B+ will cause 
trouble on any oscillator or any triode amplifier -- pentodes are 
more tolerant.  

Most of that stuff has been eliminated either in the design or the 
subsequent testing.   The main target right now is the VFO.   It's 
unavoidably a compromise design:  Obtaining linear tuning with the 
FT-101 tuning cap requires a fairly low-C tank circuit which is the 
reverse of what you want; the high output needed for the beam tube 
premixer creates the risk of problems with one or more of the 8 or 
so ceramic caps that form the tank circuit; there can be VHF 
parasitics; and since the design is 'hot cathode,' heater-cathode 
leakage is an issue.

It can be hard to pin down the problem and not all problems can be 
solved without redesigning the radio but the list above is about all 
that's possible -- B+ ripple and coupling to the signal have been 
eliminated.   

Most likely bet: heater-cathode leakage.  You can hear a 60 cps 
warble on the high harmonics of the VFO.  There's about a +5-volt 
cathode potential with about 2V p-p ripple at 10 Mcs.   

The trouble is that a small DC current flows from the heater to a 
positive cathode -- basically this is a poor-quality diode.  That 
current is modulated by the AC on the heater.   If there's a signal 
on the cathode, it also will be AC modulated by this diode 
modulator. 

One way the problem can be eliminated is by shorting the heater 
and cathode together for signal purposes.   In this circuit you'd 
need to put chokes in the heater to keep the signal from going all 
over the set and such chokes must be quite large to handle the 
450 ma current and give enough reactance at 10 Mcs.  Additionally 
they would introduce another reactance into the oscillator circuit 
bringing a new set of stability issues.  Not a good choice.

The command transmitters handled the problem by using a 
filament winding on the Hartley VFO coil, as I recall.   But that trick 
has no equivalent in this circuit.

Another approach would be to bypass the cathode to ground.  In a 
push-pull oscillator it needn't be hot.   However, bypassing creates 
high grid current pulses when the grid swings positive and the 
resulting harmonic generation leads to really bad spurious signals.  
Letting the cathode float (there's a high cathode resistor and a 
series choke) makes the oscillator operate in a much more linear 
fashion.    

Really serious high-purity oscillators are operated class 'A' with an 
AGC circuit to set the operating signal level rather than letting it get 
clipped by non-linearity in the tube.   (Wein bridge is an example.)   
This one is just "pretty close to class 'A.'"

By adding another filament transformer it would be possible to bias 
the filament positive with respect to the cathode.   That technique 
back-biases the heater-cathode 'diode' and was often used in high 
gain audio amps; it reduces leakage considerably.

By applying negative bias to the cold end of the cathode resistor it 
should be possible to do the same thing without the separate 
filament transformer.  Only 2 ma is needed so the necessary 
voltage might be tapped from the -75 volt bias supply or if that 
doesn't work, rectified from the filament line itself.   Since the peak 
negative heater voltage is about 10V and the cathode is at +5V, the 
cathode should return to something like -20 VDC and the grid leaks 
must also go there.

The next test is to watch a frequency counter.   Any jumpiness at 
all (even the last digit) would confirm the VFO as the problem.  In 
addition to looking for 'jumpiness,' just poking around with a probe 
often reveals a VHF parasitic.

And sure enough the frequency counter does show some single-
digit jumping around, so there does seem to be an oscillator 
problem.   Single digit doesn't sound like much at 5 Mcs but that's 
single digit when averaged over a full second:  Looking at the length 
of one audio cycle (say 1000 cps) it would be many times greater if 
the jumpiness is random with regard to the oscillator signal as 
leakage would be.   And an oscillator with a spectrum even a few 
tens of cps wide will certainly deliver lousy SSB audio.

No sign of VHF signals so the betting is on H-K leakage.   

Walt 
KJ4KV