I did a simple test as follows:

(1) 4GHz squarewave drives a hard switching buffer.
(2) pnoise simulation with enough sidebands, compare simulation with (timedomain at threshold) and (source).

The flick noise contribution is increased by almost the same factor as white noise. Flicker noise corner is at 100K

What I don't understand is :

flicker noise corner is much lower than f0/2=2GHz, so theoritcally there would be no folding for flicker noise, then why would the flicker noise contribution also increases dramatically?

Any reply is highly appreciated.

Hi,

The strobed noise and the time avged noise tell different stories. The former is used to describe discrete noise while the other is describe the continuous time noise. Strobe noise only gives the noise at the threshold voltage and ignores the noise at the other time of the period, but the time avged noise means the time avged noise of the total noise in the whole period. That's why you got a larger noise (both thermal & flicker) from the time avged pnoise simulation.

-Ryan

Thanks guys, I think now I undertand better.

But in my simulation, I always get higher strobed noise (after converted as shown in Kundert's jitter+noise) than average noise, not like Ryan said. So which noise should be higher exactly?
I think that is related to the situation that, now not only the noise upconverted to the fundamental tone affect the zero-crossing, but also those around harmonics. While with average noise, we count only the noise around the fundamtenal tone. Am I right?

Also with ripple counter, I tried with several different settings

(1) simulate it as whole with strobed feature.
(2) Simulate it stage by stage, and sum it up using the equaiton given in Kundert's jitter+noise

These two simulations give similar results, however they are all much higher compare to simulate it as whole with 'source'. So is it necessary to simulate the ripple counter stage by stage as Kundert suggested?

Hi Ryan,

What I mean by "avged noise count only the noise around the fundamental tone", is the noise after conversion. And 'number of harmonic'  counters only for the mixing (up conversion) effect. Bbut when you plot your phase noise around the foundamental, it is actually only the noise around the fundamental tone after conversion. Whereas the timedomain noise simulation really counts all the noises around the harmonics which affect thr zero crossing.

I have done a number simulation with quite basic circuits like inverters and so on, always get a higher noise with strobed simulation. But give it a second thinking, I think wether the stobed noise should be higher or lower really depends on the slew rate, it you have a infinite slew rate, then strobed noise will be actually 0 after converted to phase noise. As you said, avergae noise and strobed noise are differnent stories, so we should not compare them anyway. One is physical, one is not.