problems on a system's phase noise evaluation

Dear reddevil011,

>CaseB: evaluate phase noise separately. run PNoise analysis with A, select noisetype=source. Use the simulation result "output noise"  to evaluate oscillator's output. Then run PNoise analysis with B, select noisetype=timedomain. Translate the simulation result "Jee" to phase noise. The two items add to "OUT"'s phase noise.

Initially, it appears as if this will not include the impact of B loading A. This can have an impact on an oscillator's steady state amplitude and frequency. Further Jee is an integrated parameter and does not include the specific frequency components of the noise sources. AS it is a jitter parameter, it also includes only a portion of the phase noise characteristic. Hence, I am not sure how you are combining the phase noise result from A and the Jee of B to determine the phase noise result of A and B combined.

If the added simulation time of performing a phase noise analysis of the combined A/B circuit is not excessive, I would recommend running the phase noise simulation on a netlist containing both A and B.

Shawn

Thank you for your advice, Shawn.  

Firstly, when simulate block A,  the equivalent load of B is considered. So the oscillator's  steady state amplitude and frequency will not change.

Secondly, I wanna make a correction, select noisetype=jitter when running the PNoise analysis of B, the simulation result "Jee" is in a unit of "second/sqrt(Hz)"  and translate to phase noise by multiplying (2*pi*f1), which f1 is the output frequency.

CaseB is based the following consideration: Considering ideal Noiseless B, it is just a buffer, the system's output phase noise is equal to A's phase noise. When B's noise is considered, the total output phase noise is A's phase noise plus B's phase noise.   A is an autonomous circuit, so I select noisetype=source to run PNoise analysis; B is a driven block, so I select noisetype=jitter to run PNoise analysis.

Simulate both blocks together. In the ADE pnoise form, select "Noise Type: jitter" and select PM instead of the default FM. Then use Jc for the desired number of cycles as your result. Jee in principle is infinite for an oscillator, in practice its value will depend on the lowest frequency in your pnoise setup. See also http://community.cadence.com/cadence_technology_forums/f/33/t/17940.

Thanks for your reply, Frank.

I've read the thread you offer, and still have some problems.

The virtuoso version I am using can not get PM jitter for autonomous circuit yet. So, I should select Noise Type=timedomain in PNoise analysis, and use tdnoise to get jitter, by dividing the slope of output signal. 

Is my understanding correct?  Thanks.

Yes, this should be possible, you just need to make sure that you specify the correct timepoint for the pnoise analysis. Another possibility would be to write the analysis commands for the pmjitter analysis into a textfile that you define as an include file in your simulation.