HP 3325a Synthesized Function Generator - odd spectrum?

[SilverSolder]

Playing with a new toy...  an ancient spectrum analyser brought back from the dead (HP 3571A)...

I sent a 1Khz test tone into it and swept from 200Hz to 2200Hz captured the output with a scope (see attachment).

Everything looks good but I am having trouble understanding the result...

I "get" the harmonic at 2KHz, probably harmonic distortion...  but what is the reason for the low level tones present at 500Hz, 333Hz, and 250Hz?  Are they some kind of artifacts created by the way the 3325A signal generator works?  Or do harmonic distortion products go down, as well as up, the frequency spectrum?

[DimitriP]

What does it look like when you "feed" it 2, 3, 5 KHz ?

[amspire]

Could be the way the 3325A works. Just not sure of your spectrum analyser units. If the grid is 10dB/div, both the sub 1kHz and 2kHz signals look too high. If it is 20 dB/div, they look fabulous.

The 3325A uses a fractional N divider that then controls a PLL oscillator. The output of the fractional N signal has jitter which can lead to a cycle to cycle variation. The PLL will try and smooth out this jitter, but it will still be there to some extent.

Richard

[SilverSolder]

DimitriP: Same pattern of f/2, f/3, f/4, f/5.... 

Amspire: The vertical axis was not calibrated but the top of the 1KHz signal (at 0dB) is about 80dB above the noise floor.  The 3325A appears to be meeting its specifications in terms of the amplitude of the harmonics;  I was just wondering about the odd behaviour of the spectrum...  your explanation makes sense, thank you.

[amspire]

It is easy to assume that 1kHz should be easy to synthesize - the fraction N divider's clock is probably an exact multiple, so why is there jitter? The thing is when it is synthesising 1Khz, the fractional N divider is actually trying to output 30.001MHz. It doesn't divide into anything.

[DimitriP]

DimitriP: Same pattern of f/2, f/3, f/4, f/5.... 

...and what about the 250, 333, 500Hz ? 

[SilverSolder]

DimitriP: Same pattern of f/2, f/3, f/4, f/5.... 

...and what about the 250, 333, 500Hz ?

At these lower frequencies, the same pattern occurs - there are peaks at f/2, f/3, f/4 etc. 

[Someone]

DimitriP: Same pattern of f/2, f/3, f/4, f/5.... 

...and what about the 250, 333, 500Hz ?

Its likely a very simple modulation scheme thats driving the frac-n, so there are many subharmonics in the demodulated output:
http://www.seas.ucla.edu/brweb/teaching/215C_W2013/Synthesizers13.pdf
There are some examples of the VCO spectrum after frac-n synthesis in there, then consider how that looks once its demodulated.