Hi all,

I have a quesiton on how to simulate jitter?

I want to simulate the clock jitter from my clock generation circuit. Say, if I apply a ideal sinewave to my clock generation circuit, I will get a clock signal after a few digital gates. All device noise, supply nose and substrate noise will add jitter noise to the final output. I want to simulate how much jitter my final clock have.

Can any one tell me how to do this simulation?

thank you so much,
jimmy li

Andrew:

My circuit is driven by a peridical signal, not like VCO. I choose "time domain" for "Noise type" in PNOISE form and then simulate the voltage noise at the zero-crossing poiint. After simulation, I can plot the spectrum at the zero-crossing point of my output clock. Then I do integration to get the noise power. The time jitter at the zero-crossing point equals to the noise power divided by the square of slew-rate at the zero-crossing point.

Do you think I am right or wrong?

Thank you so much for your help,
jimmy li

Frank,

Thank you, I have read your replies. They are very useful to me. I also have
read the two documents on the website and got more sense. Now I still have some
small questions. Hopefully you can give me a hand.

1. You said PSS simulation can find tc.I used transient to do this and found these two
simulations show different result. I mean, the zero-crossing point of my output clock are
different from these two simulations. I don't know why. And as you said, tc is"the time where the signal crosses the theshold level of the following stage", so do you think I shouldn't use zero-crossing point as tc? Do you think I should connect the following stage to my clock otherwise I can't find tc? I think the sizing of the following stage will affect tc value.

2. In the direct plot form, when I chose tdnoise/integ output noise/total noise/spectrum/magnitude, I didn't
see any fields allowing me to input starting and ending frequencies. When I clicked "plot" then, I got an error
message "fprint/sprintf: format spec incompatible with data nil". I don't know why. The cadnece i am using is 5.0.0_MSR

3. If I chose 'output noise", not "integ output noise", I got the spectrum after clicking plot. The unit of Y-axis
is V/sqrt(Hz). Do you mean I should integrate this curve from 0 to Fs/2? I think the result will be different than
if I integrate power then take square root. I mean, integ(V) is not equal to sqrt (integ(Power)), am I wrong? I am a little confused here on the equation (53) and (54) in document "predicing the phase noise dn jitter of PLL....".

4. Since the jitter is in the order of ps, Do you think the tc I used should be accurate enough, for example, 21.12345n, not just 21.1n? And do you think tc should be within 0 and T. I found PSS only generata a transient waveform from 0 to T

thank you so much,

jimmy li