[HBR] Yet Another HBR Project -- Chapter 4

[hbr@mailman.qth.net]

A change of plans ...

After some further work, the VFO is nearly finished.   When turned 
on it goes down maybe 150 cps in 10 minutes, then drifts very 
slowly, remaining within +/- 50 cps indefinitely.   The variation seems 
to be due to changes in room temperature; it's lower at night, higher 
in the day; the reversal of direction since last report is caused by 
redoing the coil so it's mostly not spacewound.    I could probably 
rewind it yet again and come even closer, but this will do for now.   
Changing the capacitor compensation is easier -- I may do that.

I will again disassemble the tuning cap so I can trim the rotor plates 
slightly to get rid of the 1.5 kcs kink near the low end.   

Now to the matter of the crystal part of the premixed injection signal.  
Sometimes just thinking about the problem works as well as 
grabbing a soldering iron.   The trouble with locked oscillators is that 
it's darn near impossible to be sure they'll lock on the right multiple if 
that multiple is more than two or three.   I need about 8 at the low 
end, about 24 at the high one.   *That* was the problem solved by the 
Hahnel circuit -- since nothing is locked, it can't lock on the wrong 
frequency.  The worst that happens is you produce more than a 
desired minimum of output on an adjacent multiple and slightly less 
on the right one.

However, the key distinction (between locked and Hahnel designs) is 
that in the Hahnel circuit, oscillation entirely stops for every cycle of 
the controlling (crystal) signal, and then restarts.   The time required 
for this is proportional to Q of the oscillator tank circuit -- lower Q's 
are desired for a quicker stop and restart so you can actually get 
some cycles of the desired frequency before the next stop time.  

In other words, oscillation at the desired frequency is 100% 
modulated by the controlling signal.   That means you are producing 
half the total power on other frequencies -- the quicker the stop and 
start, the more big sidebands are out there.   To get rid of those, you 
need high Q circuits, but the oscillator tank can't be one of them.   
So, you need amplifiers and ganged high-Q circuits to follow.   The 
handbook I mentioned shows a circuit which is claimed to keep 
sidebands 40 db below the desired frequency with a single tuned 
amp stage.  Maybe, but it sure doesn't look as easy as it once did.   

There's a reason the military paid for this work and then never did 
anything with it.   All the multi-channel sets I've studied either used 
individual crystals (the WW-II approach) or some form of crystal 
synthesis, usually direct in the early days.   Indirect (a tunable 
oscillator phase locked to a crystal synthesized signal) began to 
take over around the early 60's, as SSB became the norm for long 
range military voice communications. 

I've done a little work with locked oscillators -- it's no problem to get it 
on the right multiple but getting it to lock there again when the set is 
switched off and then switched on again after getting cold, is a real 
challenge.   And if I'm going to have to use two narrowband tuned 
amps following the Hahnel circuit (they have to be narrow if they're 
going to cut out those sidebands) ... 

I think I'm not going down either path.  What are the alternatives?    

One would be the standard ham design -- a separate crystal for each 
band.   This has the great advantage of simplicity.   I need a crystal 
frequency for each band -- say six or seven if I don't cover all of 10.   
To use the oscillator I've just built, these would range from about 14 
to about 40 mcs.   These can be all third overtone and as special-
order crystals go, aren't terribly hard to get.   Seven by $15 or $20 is 
a little more money than I want to spend, but not totally impossible.   

I could redo the oscillator so as to be compatible with a crystal 
series I have (FT-101, maybe the KWM-2A), but I don't have any real 
spares and I don't particularly want to redo the oscillator. 

Thirdly, I could go to a synthesis approach.   For that, I may have 
crystals -- I have three or four junk ARC-27's that produced 1750 
100kcs channels 225-399.9 Mcs.   I also have some other UHF sets 
with similar output but probably somewhat different conversion 
schemes.   

(The ARC-27 was the UHF transceiver in the F-4H Phantom, the 
main Navy fighter of the Vietnam era and was used in other  high 
performance aircraft of that era.    It was heavy (around 40-50 
pounds) and very expen$ive.   A non-pressurized version (ARC-55) 
was used in large support aircraft and very different (lightweight) 
equipment was adapted or developed for rotating-wing aircraft.)  

The main crystal groups in the ARC-27 are 9.15 Mcs to 8.25 Mcs by 
0.1 Mcs steps (there's your 100 kcs channels) and 25.7-34.7 by 1 
Mcs.   Combining these (and an IF signal) they cover 20-29.9 Mcs by 
0.1 Mcs steps.   Then a multiple of 10 Mcs is added to get the final 
frequency.   I could use two of the 0.1 Mcs step crystals at 0.5 Mcs 
spacings, the ten 1 Mcs step crystals, and a 10 Mcs crystal to get 
about anything needed.   

This approach is more complex -- it requires combining three 
crystals to get each band.   Of course more crystal signals means 
much more potential for spurs.  If going that route, I'd be strongly 
tempted to change the concept to general coverage rather than ham 
bands only.  (And spurs are a certainty in a general coverage design -
- the goal is to minimize the strength and keep the worst out of 
critical bands.) 

In addition to combining three crystal frequencies, I might have to 
somewhat retune the VFO -- which for a general coverage design, I 
might be willing to do.  Of course with general coverage, the 
switching also becomes more complex.   Lacking a 60 position 
bandswitch, I'd have to use three switches -- 0.5 (2 positions), 1.0 
(10 positions), and 10 (3 positions) Mcs increments.   With some 
risk of confusing the operator the 0.5 and 10 Mcs switches could be 
combined -- but I'd probably get the hang eventually.

We're not just switching crystals there, either -- the front end and 
premixer output also must be tuned according to the band.   In a 
general coverage design, the number of front end bands rarely 
matches the number of crystal step bands and all that has to be 
figured out and made to work.   Typical front end bands might be 1.5-
4-8-16-30 Mcs -- a total of 4 bands -- note that those don't map 
neatly to the two or three switches for the crystals.   

In a general coverage design, the front end tuning cap has to be 
*large* and have a good ratio of max to minimum capacitance.   Then 
you get into Q problems at the high end of each range ... 

A larger chassis would be required.

Or -- third alternative -- I could probably design a hamband-only 
synthesis scheme from the beginning to use just two crystals at a 
time and probably not more than three or four, total.   The 1.75 Mcs 
crystal I already have would be one; reasonable multiples of that, 
added to one of maybe three other frequencies might do the trick and 
would not require retuning the VFO.   However this very likely keeps 
the awkward parts of the original scheme (problems covering 160 and 
10) without its seeming simplicity.

I'll probably doodle a while (particularly on the third choice) before 
deciding.   Told you this was going to take all winter.

Walt
KJ4KV