Dear All,
On top of the well documented (see
http://www.designers-guide.com/documents.html and example in
http://www.designers-guide.org/VerilogAMS/) PSS issue with hidden-state in veriloga behavioural models one faces the following issue.
Known veriloga phase models for transient analysis (
http://www.designers-guide.org/Analysis/PLLnoise+jitter.pdf) calculates the oscillator absolute phase(t) by integrating oscillator instantaneous frequency function of input voltage VCOin(t). Such models indeed do not converge in PSS analysis as the oscillator absolute phase is not a periodic function of time or phase(t+T0) # phase(t).
Therefore if one looks to simulate for instance -PLL loop stability (Phase margin, Gain margin) using PSS-pac analysis including effects from discrete nature of PFD sampling (assumed to be modeled at circuit level), or, - PLL Phase Noise using PSS-pnoise including folding in dividers and print contributors one must use full detailed circuit level in the analysis with associate long simulation run and possible some PSS convergency issues.
I am looking for OSCillator or VCOscillator models useable in Spectre-RF PSS-pnoise, PSS-pac, PSS-... analysis; it should be a kind of "most simple oscillator circuit" (L, C, with N-shaped non-linearity?) with parameters for pull characteristics (fmin, fMax, non-linear pull, ...) and phase noise (flicker, thermal, ...) as for their .tran models counterparts. It can not be a phase based model as all state variables must end-up periodic in time for PSS analysis. It is a model because it takes abstraction of a lot of details in R, L, C, MOSFET's ... still maintaining the behaviour using most simplified circuit in PSS analysis.
Thank you for your help herewith.
Regards,
ppg.