AnsweredAssumed Answered

89441A Vector Signal Analyzer + Jitter Measurement

Question asked by igouzouasis on Apr 6, 2013
Latest reply on Apr 11, 2013 by DaveHornbeck
Hi

I posted the same message in another category, but I believe this is the right one. 

I have some questions on the old 89441A Vector Signal Analyzer. I know that the device is old, but hopefully someone may give me a little help on the matters that trouble me.

I used the VSA with the RF section to measure phase noise of an oscillator at 650MHz. There is a very useful Agilent step by step tutorial to perform phase noise measurement:

http://www.home.agilent.com/upload/cmc_upload/All/PhaseNoiseMeasurementsonthe89441A.pdf

I want to use the Phase Modulation method with noise subtraction and the math equation to be inserted in the instrument is F3 = PSD1-(D1/K2) (page 13), where K2=0.05 - more details are in the tutorial. 

However, application note "Agilent PN 894400-2" page 4, describes the same method, same equation, where K2=20. 

http://cp.literature.agilent.com/litweb/pdf/5091-7193E.pdf

Has anyone else identified this problem? Which is the right constant to use? I tried to find a solution logically, but couldn't.


The second question I have is about jitter. Ather using Phase Demodulation, I took a plot of Power Spectral Density. Measurement bandwidth is 5MHz. I then took the time domain of the demodulated signal for a jitter measurement. Consider attached image. Using band power markers I get the rms phase deviation in rad^2. Then, a computation to rms time jitter is as simple as squaring the rms phase deviation and dividing by 2*pi*f, right?


Also, in order to get the p-p jitter, should i set two markers to min and max phase deviation, get the p-p phase deviation and divide by 2*pi*f?

A situation that troubles me is that another way of computing jitter is by taking a histogram in an oscilloscope. Setting trigger at zero crossing, the oscilloscope (an Infiniium) takes many period-to-period samples and store the hits as histogram. Peak to peak is easy to be observed in the oscilloscope as the end points of the histogram. The problem I have is that the value taken by the histogram is considerably different by the computation (p-p phase)/2*pi*f.

Am i getting something wrong? 

Thanks,
I can provide more information if want to.

Edited by: igouzouasis on Apr 6, 2013 10:32 AM  

Attachments

Outcomes