Designer’s Guide Community

James

Community Member

Offline

snowfiled

Posts: 32

Jitter inconsistent question
Jul 10^th, 2005, 9:48pm

Dear Ken,

1, In your jitter+pnoise paper, I use the following fomulars to calculate my VCO jitter:

c=L(delta(f))*(delta(f))²/f0²
J=sqroot(c*T)

my f0=200MHz, that's my VCO oscillating frequency,
delta(f)=2MHz. 2MHz is the bandwidth for my PLL.
T=1/200MHz (is it correct?)

from pnoise simulation in Cadence, I got L(delta(f))=-102dBc/Hz

Then I got Jitter=5.3ps. is it rms jitter or Peak to peak value?

2, in pnoise setup form, I found there is an option "jitter"
in the "Noise Type" cyclic form. I thought I could get jitter by choosing this option. After simulation, I got the jitter value:
rms=3.77p peak to peak=23.2986p

the exact resulet is like this:
J[Second]@2e6Hz:rms=3.77p
J[Second]@2e6Hz:ber=1e-3=23.2986p

I have the following questions:

1. what's ber and 1e-3 in the second jitter result? Why the calculation result and simulation result are different? Where am I wrong? ???

2. Why rms jitter is smaller than p2p jitter? I thought it should be larger than p2p jitter.

3. I also noticed that in the Result-->Main form, after choosing pnoise jitter and function as jitter, there is "modifier" with three choices: Second, UI, ppm. I should choose second I suppose, but what are the other two?

4. the reference freqency is 25MHz, divider ratio is 8, so my VCO is working at 200MHz. To get the widest bandwidth, I choose BW=1/10 reference frequency, i.e. about 2MHz. Sould I use 2MHz as delta(f) when I watching the pnoise result? In other words, should I use the bandwidth as the offset frequency?

Really appreciate your answer

Back to top

James

IP Logged

Sergei

Guest

Re: Jitter inconsistent question
Reply #1 - Jul 11^th, 2005, 2:03am

1) The difference of sqrt(2) is due to the difference between L and phase noise, which is many times discussed in PNoise PCRs. Please make sure that you are are using the PM noise which is computed for "jitter" PNoise analysis and could be accessed on the same Direct Plot form.

2) The second reported number is a peak to peak jitter for particular BER (which is a default 1e-3 in this case).
For the relationship between RMS and P-P jitter, scout the web, it has plenty of references.

3) Other units (modifiers) are used in the industry:
Unit Interval and
Parts Per Million.

The latest SpectreRF user guide has some information about jitter measurements.

Sergei.

Back to top

IP Logged

James Community Member Offline snowfiled Posts: 32	Re: Jitter inconsistent question Reply #2 - Jul 11^th, 2005, 9:43am Thank you very much Sergei, I found a paper on rms jitter an pp jitter. I list it below so somedody else who are not familiar with jitter also can have a look http://pdfserv.maxim-ic.com/arpdf/AppNotes/3hfan402.pdf
Back to top	James IP Logged

Ken Kundert

Global Moderator

Offline

Posts: 2386
Silicon Valley

Re: Jitter inconsistent question
Reply #3 - Jul 11^th, 2005, 10:37am

James,
The jitter you compute is the RMS jitter. I have updated my jitter papers to include a section that describes how to convert Gaussian RMS jitter to peak-to-peak jitter (see Section 8.1.1 in http://www.designers-guide.org/Analysis/PLLnoise+jitter.pdf).

I don't understand why you are getting the rt(2) difference between hand calculations and what SpectreRF produces. I did not understand Sergei's comments on this question.

Once you read Section 8.1.1 I expect that you'll understand why SpectreRF provides a bit error rate (BER) with the peak-to-peak jitter number.

Any RMS value will always be less than or equal to the corresponding peak value, and will be less or equal to half the peak-to-peak value.

-Ken

Back to top

IP Logged