Why does the frequency drifts in this design?

[liberal]

pll drift

My synthesizer is 880MHz with ADF4113,the reference is a 10MHz OCXO,and the N is 88.
1.when i test the OCXO with spectrum analyzer,it is 10.000002MHz,i think if locked,it will output 880.000176MHz,but the spectrum analyzer is 880.000471,why?
2.when i reduce the SPAN to 1kHz,i see the frequency will drift slowly,i will be 880.000490MHz after half an hour(maybe it will drift with time,but i don't measure),why drifting?and when the loop bandwidth is 10KHz it drift quicklier and farther than 100kHz loop bandwidth why?
3。when vco output is 880.000490MHz, divided by N=88 ,it is 10.0000056MHz,if the PFD detect one or some Hz?

 

[RFDave]

Re: pll drift

What is the reference of your spectrum analyzer? It's likely that the two are drifting slowely relative to each other.


In addition, a spectrum analyzer isn't really the right tool to use for measuring frequency. You really need to use a counter to get the Hz resolutions that you are looking at.

Dave
www.keystoneradio.com

 

[biff44]

Re: pll drift

You may have a drift in the actual synthesizer, but like Dave said, you probably do not. You are lucky in that you are using a 10 MHz reference in the PLL. Your spectrum analyzer most likely can output 10 MHz from its back panel. Use the spectrum analyzer's 10 MHz reference out to replace your PLL 10 MHz ocxo, and you will probably notice ZERO drift. If not, call us back!

 

[RFDave]

Re: pll drift

Biff44 has a good suggestion. If you do that, don't be concerned if the phase noise get's worse using the spectrum analyzer as a reference. I've had trouble with 10 MHz references having lot's of phase noise in the past.

Dave

 

[liberal]

Re: pll drift

yeath,spectrum analyzer drifts............

 

[madengr]

Re: pll drift

Talk the boss into buying a GPS stabalized source. It's a 10 MHz OCXO (about as good of phase noise as you can get) that is stabalized by GPS. So you have excellent long term drift with good phase noise. Much better phase noise than rubidium clocks. You can then run this to each bench via fiber.