Hello All,

I ran pss and tdnoise in spectreRF to estimate jitter for my oscillator. My ring oscillator is a differential relaxation oscillator with frequency of 4MHz. The last stage has differential to single ended converter.

I have following clairifications to be done:

1) Should I perform noise simulation on one of the differential output of the inner-stages of the ring oscillator or should I use the output of the differential to single ended converter output ?.

to get jitter I am integarting noise from 22KHz to 4MHz in tdnosie form. I get 390mV which I divide by Slew Rate of 330MV/s I get a jitter of 1.02nsec. This is in agreement to what I got on silicon in a 30us Time interval Error a measured jitter of 1.4nsec.

If I integrate from 25KHz to 4MHz, I get the noise as 53.6uV which then divided by Slew Rate gives a jitter of only 0.13psec. This is much smaller.

On what basis do I chose the BW for integrationin tdnoise form. ?

Why is the noise so different between the two BWs, 2khz-4Mhz or 25KHZ to 4MHz. ?

The jitter estimated in the BW is what, periodic jtter, longterm jitter or what ?

Any feedback will be greatly appreciated.

ic_engr

Ken

Then when is tdnoise useful ? I was under impression that tdnoise is an easy way of estimating jitter.

My Spectre Models also include 1/f noise models. So would the method of section 10.1 be valid.

ic_engr

tdnoise can be used to computer jitter for driven circuits.

If you are getting the noise in dB rather than dBc, then you are not pushing the right buttons. I think you need to plot the "Phase Noise" to get your results in dBc.

Flicker noise invalidates all jitter metrics except for cycle-to-cycle jitter.

What is confusing you?

-Ken

Ken,

You were right I did plot the "Phase Noise" and I did get the dBc/Hz. What I got was a sloping plot with fc at 20kHz. After fc the phase noise is basically flat at -130dBc/Hz i.e L=1e-13. My fo=4MHz

Using eqn 74 and 78 with different delta f I get different jitter numbers.  
If I use Delta f of 30kHz I get jitter=1.18pS.
If I use Delta f of 500kHz I get jitter =19pS
If I use Delta f of 2MHz I get jitter=79pSec.

Can you please explain me what these different Jitter numbers mean when I use different Delta f.

With 1/f models included in simulation what do these jitter numbers mean.  

To het Time Interval Error in 30uS which of the above numbers should I multiply with SQRT(30us * fo) ?

I will really appreciate your input.

ic_engr.