Bill,

OK...  Here are the results of the latest test.  This time I didn’t do any averaging.  It took some the time to capture the data and do that huge FFT so the relative shift in frequency makes the peak level look lower than it really is. If you reduce the FFT block size back to 8K then it will display correctly.

Here’s the crystal oscillator (I know that 120 Hz is not coming from the oscillator)

Here’s the HP 8662A – Just slightly noisier.  You can see that this oscillator is more stable and the peak is narrower.  I feel confident that this is due to less drift during the FFT processing.

And here’s the HP 8656B – a much less expensive instrument and fairly nasty, but synced to the same reference.  I’m sure that 120 Hz is still there but you can’t see it anymore...

Here’s the settings you can’t see:

Filter Bandwidth: 5 kHz

6620 IF Gain: 0 dB (the lowest)

RF Gain: -20 dB (the lowest)

FFT Window Type: Hanning

Don’t feel bad about not understanding these settings.  I’ve been fooling with this for several weeks and I still don’t claim to get it. The master oscillator is running at 66 MHz and it’s not a straight crystal.  You can set it to dial in the frequency if you want to calibrate it.  The main spur is always directly in the middle of the display, regardless of frequency.  The gain settings have a huge affect on the amplitude.

I hope this helps!

Juan