The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Measurements >> Phase Noise and Jitter Measurements >> phase noise to jitter
https://designers-guide.org/forum/YaBB.pl?num=1236788112

Message started by sun on Mar 11^th, 2009, 9:15am

Title: phase noise to jitter
Post by sun on Mar 11^th, 2009, 9:15am

hi,
I simulated my PLL's phase noise at behavior level and got a closed-loop pll phase noise curve. Than I tried to integrate the phase noise to get phase jitter and period jitter.
The follow matlab routine is used to convert phase noise to phase jitter and period jitter:
phasejitter = sqrt(integral(fre,2*Lpll))*T/(2*pi);
periodjitter = sqrt(8*integral(fre,Lpll.*(sin(pi*fre*T)).^2)/((2*pi*f)^2));

where Lpll is the single-sideban phase nosie of the pll, the function integral is:
function integ = integral(fre,L)
% calculate integral
len=length(fre);
for n = 1:len-1
areatrape(n) = (fre(n+1)-fre(n))*(L(n+1)+L(n))/2;
end
integ = sum(areatrape);

I expect the same results of phase jitter and period jitter because the pll locked. but the results are:
phase jitter = 8.31e-11
period jitter = 3.27e-12

Is there any error of my routine or my understanding of phase jitter and period jitter?
in the attached figure, the red curve is the phase noise of the PLL.
Sun

Title: Re: phase noise to jitter
Post by swjtuliwang on Apr 15^th, 2009, 4:36pm

your simulation curve in psd of vco、cp、lpf whether it be through the transfer function?i consider the way through the transfer function is uncertain.

i would like to know specific methods about the psd of vco、cp、lpf
my email:swjtuliwang@hotmail.com

Title: Re: phase noise to jitter
Post by smlogan on Apr 16^th, 2009, 1:24pm

sun wrote on Mar 11^th, 2009, 9:15am:

Hi Sun,

I understand your expression to convert from phase noise to jitter. That appears correct. However, I do not understand your expression for period jitter. I don't think you can obtain that from a phase noise plot as the phase noise plot is computed relative to a specific carrier frequency. Period jitter is a measure of the variation of the period from a fixed period T. A phase noise plot relates the variation from its carrier frequency - which may not be the frequency corresponding to the period T of your period jitter requirement.

A definition of period jitter can be found at URL:

http://www.jedec.org/download/search/jesd65b.pdf

Shawn

Title: Re: phase noise to jitter
Post by chenyan on Apr 17^th, 2009, 6:24am

Hi sun,

Your calculation of period jitter is correct (not rigorous mathematically but good enough practically).

The problem is the definition of your "phase jitter". From how you calculate it, I guess it is like edge-to-edge jitter, or absolute jitter if you wish. For a pll, absolute jitter is unbounded. If you integrate from 0 to infinite, you get an infinite.

I can not think of any reason why period jitter and rms timing error should be equal.

Title: Re: phase noise to jitter
Post by Mayank on Jun 7^th, 2009, 8:53am

hello sun,
Do you mean accumulated Jitter by the term period jitter that you are using ? Because guessing from your simulation results, period jitter values might indicate accumulated jitter. single period jitter is given by delta(vcont) * del(f) / f^2 just multiply ur single period jitter here phase jitter by no. of vco cycles in 1 i/p period, if that matches with your period jitter, it is a measure of accumulated jitter wrt i/p time period

Title: Re: phase noise to jitter
Post by kanan on Aug 13^th, 2009, 8:29am

Hello Sun,
The results you are getting are perfectly correct.
The period jitter you are calculating refers to (tmax-tmin) where tmax and tmin are the max and min clk periods seen the clock
The phase jitter you are calculating refers to how much "off" a particular Nth edge is from the ideal edge. The N depends on the frequncy range over which you integrate it. Typically its fmin to fo/2, where fo is the clock frequency itself and fmin will give you an idea of the N referred to above.
Though not mathematically very rigorous, I can give the following example:
If fo = 1Ghz, and fmin is 1MHz, tau = 1/fmin = 1us.
Thus Nth edge refers to tau/fo = 1000th edge.

Hope this helps