[time-nuts] HP 3586A/B/C entirely referenced to 10MHz: A solution

[Bert, VE2ZAZ]

Hi Everyone,

Some recent tests I have made on my HP 3586B Selective Level Meter have 
confirmed that the detected audio drifts a lot as a function of ambient 
temperature. A day/night change of 1.5 degree Celcius was clearly visible on 
Spectrum Lab samples. I could actually find out how many times and when the 
central home furnace had cycled in-out overnight. It gave a neat plot which I 
post here: https://www.onlinefilefolder.com/3sFcUl8Z8i8zy7

So I have been looking at modifying my HP 3586B SLM so that it becomes entirely 
synchronized off the external 10MHz reference. As you probably know, all stages 
exept the final one, the SSB LO (BFO),  are derived from the 10MHz reference. 
The SSB LO chain is made of two free-running crystals used for LSB and USB 
detection. In my unit (3586B, option 003), the detection LO frequencies are 
13775 Hz and 17475 Hz. These two frequencies are not directly math-related to 
any other internal reference (at least I could not personally find any). BTW, 
the IF LO rate is 15625 Hz.

I have substituted a function generator instead of the internal oscillators just 
to see how much the off-centering would affect the audio quality, and found out 
that this works well as long as it is not too far off the designed LO rates. The 
closest 10MHz-derived integer rate I could find that will work for audio tone 
measurement is divide-by-726 (13774.104... Hz) in LSB. The closest USB rate 
would be divide-by-572 (17482.5... Hz), much farther than the LSB one. So the 
LSB rate difference of around 0.9 Hz can be neglected when keying in the 
frequency and listening to regular radio signals. When making absolute audio 
measurements, subtracting that delta is easy to achieve in post processing. 
Besides, the frequency and amplitude measurement capabilities of the instrument 
are not altered by this mod. To me this would seem like an acceptable compromise 
that is simple to implement and would add long-term stability. My intent is to 
have a single chip (in this case, an 8-pin PIC divider) do the trick, as we are 
dealing with tTL level signals here.

But can I do better?
How much more complex?
Can I avoid PLLs?
All mixer stages become synchronized to a single source. Is this an issue for 
reliable audio detection?
Am I missing something here?

As always, I truly enjoy your feedback and am quite convinced I will learn 
something new once more.

Thanks,

Bert, VE2ZAZ