[HBR] Yet Another HBR Project -- Chapter 3

[hbr@mailman.qth.net]

The tunable oscillator part of the job is about wrapped up.

Tracking the dial has been somewhat of a challenge.   I pulled the 
tuning cap apart (48 itty-bitty ball bearings, total, not 100!) and filed 
the rotor plates to (I thought) fix the 'kink.'  Two hours from starting, it 
was back together, bearing preload correct, and working -- the 5 kcs 
kink was only about 1 kcs smaller because I had misjudged the 
problem.   So after staring at the thing for a while I got out the Dremel 
and what they call an rotary file but is actually a sort of end mill.   I 
stuffed paper towels over the bearings, Sharyn held the vacuum 
cleaner nozzle, and I milled off the necessary part of the plates, a 
little at a time with the oscillator running so I could check progress 
frequently.  Got it down around a 500 cps error and decided to call it 
quits.   

Following that I started bending plates -- very little required, except 
near the low frequency end where there's another kink of about 1.5 
kcs.  (Pretty much an inevitable result of using a series padder to 
correct the curve of a cap designed for SLF over a 2:1 frequency 
range -- a three point fit is going to leave a pretty good size error 
somewhere.)  It may not be possible to bend plates to fix that 
because the slots in the plates are in the wrong place.   However the 
Dremel technique might work again.   That is, it would work but the 
plates are less stiff there and I'm not confident of being able to do the 
job without ruining the cap.   However I'll probably give it a shot in the 
next day or two and then do a final pass on linearization.   Except at 
the low end kink it's around +/- 200 cps, end to end -- good enough 
for now, since the real tuning range (which affects linearization) 
depends on the crystal filter frequency which won't be known 
precisely until the receiver is built.   

The crystals are nominal 6.5536 Mcs clock crystals.   That's the 
parallel resonant frequency, the series resonance will be below there, 
and the passband of a series-crystal ladder filter (I think this is right) 
starts at the series resonance and extends upwards toward the 
parallel resonance.   Does anyone have a good description of the 
process of building a ladder filter?   I've seen discussions but they 
always seem to leave out a couple of critical facts.  I've done it before 
but it'll go faster if I have directions!

Then I went through and rewired the oscillator, shortening leads and 
generally getting rid of the haywire.   On the first test of drift, the 
oscillator went down a couple of hundred cps, then up over a few 
hours by several hundred.   I had finished up the linearization with 
most of the coil space-wound so that was no surprise.  I took two 
turns off the coil which made it necessary to close wind most of the 
rest; drift was then down 200, back up 200 and more or less stable.   

(With windings about 2x diameter in length, a closewound coil on 
polystyrene form will increase in inductance as temperature 
increases; with about 1/4 or 1/3 of coil spaced slightly it will have a 
near zero coefficient, with more than that spacewound it will have a 
negative coefficient.   Because they don't expand much with temp 
increases, ceramic coil forms are *not* more stable for ordinary ham 
VFO purposes -- the form stays put but the wire expands and 
increases inductance.   But if you had to build an oscillator for an 
extreme temperature range (A B-17 on the ground in England maybe 
50 degrees, at 28,000 ft. about -10 ...) you'd want to go ceramic.)

200 cps down and back is perhaps tolerable, but could still be better. 
I poked around with my digital thermometer probe and found that the 
only tuned circuit part that heats very much was the coil -- about 2 
degrees.   That seemed to be divided between some self heating and 
(mostly) heat conducted from the 39k resistor connected to the 
center tap.  (1 mA, 39k, 39 mW -- it doesn't take much if it's 
connected to a critical part!)    I split the resistor into 10k for isolation 
located in the box and 27k below the chassis, bypassing the 
junction.  (I didn't formerly have proper isolation there but obviously 
have to have it to use the oscillator in a receiver.)   Also reoriented 
the resistor from long axis vertical to axis horizontal, for better 
cooling.   Warm up drift seems to be less, but I'll have to let it get 
stone cold to be sure.   

The only other part that heats at all is a fixed capacitor of 15 mmf 
that's across the coil.   This is an NP0 cap, but those are almost 
never precisely '0' so it's best to keep them as cool as possible.  I 
only had a tiny one; I can either order a physically larger unit or use 
four 3.9's in parallel.  Humm ... I also could use a larger cap on the 
other side of the series padders, that is, across the tuning cap.   
That's much futher from the tube sockets so less temp increase and 
also a lower voltage point.

The downward drift is the tubes.   Once the upward drift is minimized 
I can put a compensating cap right on the tube socket.

The next question is "What's the rest of the front end going to look 
like"?   

My idea of a mixer with one input push-pull and the other single 
ended, output single ended, won't work.   A mixer can be thought of 
as a circuit that multiplies the two inputs so (say) negative x negative 
= positive.   Thus output of one mixer can be added (hooked in 
parallel with) that of a second mixer with positive x positive inputs.  
But negative x negative added to negative x positive -- the outputs 
cancel.  (Jim, I think you tried to say this ...)   Basically two of the 
three connections to a two-device (singly balanced) mixer have to be 
push pull and the other must be single end.  I have to either to go 
push-pull for one input and single ended for the other with output 
push-pull, or use two push-pull inputs and single-ended output.

I only have one tuning cap section available to tune the output, and 
the signal mixer I was planning (from the G2DAF Mk II) expects 
single end input.   That favors a premixer that's push-pull on both 
inputs, allowing single end output.   But I don't know about a Hahnel 
circuit with push-pull output ... that would likely require more than 
one envelope.

Alternative:  Go with the push-pull VFO, a single ended Hahnel input, 
take the premixer output push-pull using a separate winding for 
tuning.   More band switching that way.   Could either use single 
ended output or push-pull.

In between alternative -- if you ground one grid of a (say a dual triode) 
mixer, float the cathode, and drive the other grid, you get much of the 
effect of a push-pull input.   That is, all the phase relationships work --
the only thing you don't get is full cancellation of the input in the 
output circuit.  (The total cathode current has to vary at signal rate to 
get opposite grid-cathode voltage changes, so the total plate current 
must vary similarly.

Another interesting fact -- the 6BN6 gated beam tube has 
outstanding linearity on either input.   I'm guessing they can't be 
used as signal mixers because of high noise levels but they ought to 
work fine as premixers with 100 mV or larger ignals from both 
oscillators.   And unlike the beam deflection tubes, the electrode 
voltages are reasonable.   So two 6BN6's with G3 driven push-pull 
(the VFO) and one G1 driven single end (Hahnel) with the other G1 at 
signal ground could have the plates in parallel.   The VFO gets the 
full benefit of balanced mixer design, the Hahnel only gets some, but 
those signals are always 5 Mcs on the high side of the desired 
output frequency -- fundamental is relatively easy to get rid of and 
harmonics are a non-issue.  And you have single ended output.

There is outstanding info on 6BN6 performance in a product detector 
in the article "Some Ideas in a Ham-Band Receiver," Arnold & Allen, 
QST, May 1960, reprinted in "SSB for the Radio Amateur", 1965 
copyright.

Definitely time for the next round of headscratching!

Walt
KJ4KV