Noise separation in phase noise simulation

[du_hast]

Hi

Im doing a phase noise simulation in cadence (using MMSIM70) Im using the option "noise separation" to determine the amount of noise introduced by each instance in my circuit. However, when i compare the total output noise (in the band of interest) with the sum of the individual contribution of each instance (in the band of interest) the curves differs. What is wrong? Any idea?

thanks !

 

[biff44]

Phase Noise similations are always questonable.

But assuming the program is correctly predicting noise, how are you summing the individual components to get the total output noise? Are the noises of the components correlated or uncorrelated to each other (makes a big difference in how the summation is done). Is there am to pm and am to am compression going on--also changes how the summation is done.

Rich

 

[du_hast]

Hi biff44, thank you very much for you valuable answer...

Well, i dont know exactly how cadence determine the output noise. However, assuming that the circuit is considered by the simulator as a LTV system, the total output noise is just the sum of each output noise contribution, isn't it? Take a look to this paragraph at cadence site:

"The shooting Pnoise fails to simulate the effect of the supply noise on phase noise because it solves the VCO’s small signal response in two separate steps. First it solves the phase-modulation (PM) response using the phase sensitivity function calculated from the extraction of the dominant eigen-mode. Second it solves the amplitude modulation (AM) response by subtracting out the PM contribution from the perturbation equation. The method is problematic for perturbation, which decays slower than the oscillation period, such as the noise from bandgap voltage reference with large RC-time-constant. For the slow-decaying perturbation, there is no frequency-independent separation between AM and PM responses, thus they need to be solved simultaneously.

To address this issue and accurately simulate those type of circuits , Cadence has enhanced harmonic balance Pnoise analysis. Using the "augmented" option available in MMSIM71 ISR5 and above, the end-user can now predict accurately the phase-noise of such circuits. The same solution will be available in MMSIM71 ISR6 for the shooting method.
"

In my case, im simulating a complementary CMOS LC VCO... i guess that the output noise is uncorrelated as it comes from uncorrelated sources. As far as i know at low frequency offset the phase noise is almost PM, however, because of the varactor there is a AM-PM noise mechanism....but i guess that this mechanism is already considered and the system remains linear (but time variant) thus the output noise could be summed ?