Why there is no convergence in PSS and transient if I enable "Calculate initial conditions (ic) automatically"?

Dear Clara,

Clara Dong said:

Could you tell me what's the difference between "Calculate initial conditions (ic) automatically" and set  initial condition through Convergence Aid by users? Is that reasonable if I disable "Calculate initial conditions (ic) automatically" and give initial condition through Convergence Aid?

As you probably know, Cadence recently changed the default behavior of a pss analysis such that the "Calculate initial conditions (ic) automatically" radio button in the GUI is now, by default, checked to enable the feature in a pss analysis. The motivation for this change in default behavior is outlined in the Troubleshooting article at Cadence URL:

support.cadence.com/.../ArticleAttachmentPortal

As detailed in the note, Cadence reports that its experiences suggest many designers do not how to set appropriate initial conditions to expedite the transient portion of the pss analysis and hence a good estimate of the steady-state waveform is not achieved at the end of the tstab portion of the pss analysis which results in convergence issues for the pss portion of the analysis.

However, when this feature is selected (oscic=lin), and you do specify a set of initial conditions to the simulator using ADE, they will be ignored as initial conditions and used, instead, as nodesets.

Therefore, to enforce your set of initial conditions in the simulation, you must verify the "Calculate initial conditions (ic) automatically" radio button is not selected.

When you do have "Calculate initial conditions (ic) automatically" radio button checked, spectre performs a linear eigenvalue analysis of the DC solution. This essentially provides an estimate of the poles and zeroes. Spectre uses this information in an algorithm to expedite any simulation time required to achieve an estimate of the steady-state waveform (amplitude and frequency) for the subsequent pss analysis. However, Cadence also notes that for some oscillator characteristics (such as oscillator Q) and some types of netlists, the use of this feature may lead to more convergence issues.

More details on the initial conditions and how Spectre estimates them is provided page 352 of its reference document "Spectre Circuit Simulator RF Analysis Theory" (August 2021 version, product version 21.1)  at its URL:

support.cadence.com/.../techpubDocViewerPage

Anyway, I hope I understood your question and this provides some helpful insight or information Clara!

Shawn