[sdiy] Oscillator tracking

[Paul Schreiber]

a) before you 'blame' the VCO, you have to also be sure of 2 things:

1 - your voltage source is accurate to less than 300uV, and that you can 
*measure* it accurately to 50uV
2 - your frequency counter is accurate to at least 0.1Hz

b) the CEM3340 was *specifically designed* not to be a 'stand alone' VCO IC. It 
was designed to be "pretty good" and have a uP measure the output frequencies 
and then have "on the fly" correction voltages added periodically. Because the 
tempreature compensation used is based on a "just OK" bandgap (not like it's 
5ppm or such). This is referred to as the Autotune. Doug Curtis expected that 
the CEM3340 would have to be measured/compensated via CV corrections every 12min 
or so. There is *no* HF tracking compensation in the 3340 to speak of. Because 
there is no *reason to*, the uP would simply add/subract an offset from the 
"perfect" 1V/Oct output to make it track (yes, this is in a way cheating, but 
again, the 3340 was designed to be in a closed-loop environment). Doug told me 
that his goal was the 3340 was 90% of the work, the CV loop compensation was the 
other 10%.

c) I had to spend lots of (pre-eBay) $$$ to get the equipment to take care of 
(a). But the 1 thing eBay has done is provide 80-90% off original prices for 
test equipment. I have seen $2400 HP DVMs sell for $45. My Keithley CV reference 
is accurate to 18-bits and has a 3ppm voltage reference in an oven. My frequency 
counter uses an oven crystal VCXO with a 68020 processor doing on-the-fly 
adjustments to better than 0.25ppm. Yeehaaa. You can get all sorts of great 
stuff on eBay if you look.

Paul S.