understanding of frequency folding in Cadence pnoise set-up

Best,

The noise from the lower sidebands of harmonics of the oscillator frequency will not contribute to the output noise at the upper sideband of the first harmonic of the oscillator frequency - there is no mechanism for that (I'm not just talking about the simulator here). You'd normally need to simulate the lower sideband too - and the common way of doing that is to use the "modulated" mode of the pnoise analysis which will simulate both the upper and lower sideband and combine the results to give you the PM and AM parts of the noise.

So that's probably what you want?

Regards,

Andrew.

Hi, Andrew

What I’m looking for is to get to understand how noise from 0.5GHz to 1 GHz will contribute to 1.0000002G to 1.5G Hz, If VCO output freq=1GHz and sweep start=200 and stop = 0.5GHz (standard setup for pss/pnoise).

I think it just can’t. Instead Cadence somehow assume VCO's noise profile is symmetric along 1GHz. It means that noise profile between 0.5G and 1G is just mirror of noise from 1G to 1.5GHz. So in simulation we only look into noise from 1G + 200Hz to 1G + 0.5GHz.

Correct me if my understanding is not right.

Thanks!

There is no assumption that it is symmetric. You're only simulating the noise in one sideband though, because the noise in the lower sideband won't transfer to the upper side band - you can equally well simulate the lower sideband if you wanted to. Given that frequency translation can only occur through integer multiples of the PSS fundamental (in real life too, unless there are other harmonics involved, which you'd have to include in the simulation), this is correct and to be expected.

This is why you might use the modulated mode of pnoise, because that will simulate the output noise in both sidebands and then combine them to give you the overall double-sideband PM noise at the output.

So your understanding is not correct.

Try the modulated mode...

Andrew.

Hi, Andrew

Thanks for quick reply.

From your message, I understand that the lower/upper side band (in this case, from 0.5GHz to 1GHz and from 1GHz to 1.5GHz) MAY have different noise profile. The modulated mode is best to characterize such behavior.

Then I raise another question: if I simulate a ring oscillator’s phase noise by PSS + PNOISE (with noisetype=sources), the resulting pnoise curve is definitely single side band. Does that mean the result is somehow not correct or at least not complete? Do you suggest that we should go with pnoise (noisetype=modulated or jitter) for a VCO noise simulation?

Thanks again.

It doesn't mean that it's incorrect - it depends on what you are wanting to look at. However, modulated might be more appropriate - it just depends on what you wish to see!

Kind Regards,

Andrew.