Designer’s Guide Community

Pages: 1

Transient Noise to build Phasenoise curve (Read 4424 times)

Stefan Senior Member Offline Posts: 124	Transient Noise to build Phasenoise curve Sep 21^st, 2007, 2:15pm Hi, has anybody of you used a transient noise simulation to calculate a phasenoise curve for an Oscillator or PLL ? I know that this is error prone and RF measures would be more appreciated, but in this special case we simply can do only transient simulations. Anyone has experience with this ? Regards, Stefan
Back to top	IP Logged

Sarig Junior Member Offline Posts: 31	Re: Transient Noise to build Phasenoise curve Reply #1 - Dec 6^th, 2007, 10:10am Hi, I have the same problem. did you find the answer ? Erez
Back to top	IP Logged

neoflash Community Fellow Offline Mixed-Signal Designer Posts: 397	Re: Transient Noise to build Phasenoise curve Reply #2 - Dec 16^th, 2007, 6:34am Since phase noise is a random process, it requires very large amount of data to do statistical analysis. So, if you run transient, you make sure you do very long time and have enough of data for jitter analysis. I don't think it is feasible way.
Back to top	IP Logged

Sarig

Junior Member

Offline

Posts: 31

Re: Transient Noise to build Phasenoise curve
Reply #3 - Dec 17^th, 2007, 1:32pm

Hi,
We need to do Tranisent noise and find the phase noise because for example for Fractional N this is the only why to find the Phase noise.
Also for VerilogA models you need to do many changes i order to run PSS+PNoise.
So if you want you can for PLL-> Down convert to I and Q -> Filter -> bring to Matlab using:
x=I+jQ
find the angel -> unwarp -> fit to line and PSD it.

you are right that you need long long sim (1mS for GHz VCO)
Sarig.

IP Logged

kDaniu

New Member

Offline

Posts: 6

Re: Transient Noise to build Phasenoise curve
Reply #4 - Apr 30^th, 2008, 4:55pm

neoflash wrote on Dec 16^th, 2007, 6:34am:

Since phase noise is a random process, it requires very large amount of data to do statistical analysis. So, if you run transient, you make sure you do very long time and have enough of data for jitter analysis.
I don't think it is feasible way.

you have to say - How much that your 'for a long time'?
that transient noise in Spectre has a behavior to degenerate (in the case of statistical quality (randomness))

IP Logged

Pages: 1

‹ Previous topic | Next topic ›