Stats

  • Replies 2
  • Subscribers 131
  • Views 583
  • Members are here 0

Issue with Saved Nodes Being Filtered During Spectre PSS

Alireza Razzaghi

Hi Andrew,
I’m currently using the “selected” option under Save Options → Select signals to output (save) in Spectre. I also include a text file under Simulation Files → Definition Files containing several save statements such as:
save X1.* depth=2
save X1.X2.* depth=3
save X1.X2.X3.* depth=4

In this setup, X1 contains a block with LVS-resistor parasitics, which breaks connectivity along the paths where I want to extract and save parasitic nodes. This combined approach had been working well until recently, when I noticed that in pss and transient analyses Spectre began filtering out some nodes that should have been saved.
To diagnose this, I added the following under Options → Analog → Miscellaneous → Additional arguments:
dump_wildcard_info=yes

This generated the input.wildcard.out file. After examining it, I observed that, for example, Spectre reported:
save X1.* nodes: 35

but only 12 nodes were actually listed under that wildcard. Searching the simulator log revealed this warning:
WARNING (SPECTRE-8544): Ignoring internal parasitic RC nodes `CLK' in save statement because
parasitic optimization will filter out the parasitic nodes by default. Set savefilter=none
if the nodes are not to be ignored.

As suggested, I added:
savefilter=none

also under Additional arguments, and this fixed the issue for a short transient analysis. With this change I was able to see all 35 nodes in both input.wildcard.out and the Results Browser.
However, when I ran a PSS analysis, although input.wildcard.out still listed all 35 nodes, the Results Browser again displayed only a subset of them. It appears that Spectre is dropping some nodes during PSS, but I am not sure why.
Would you please share any insight into how I can ensure that Spectre retains all nodes without applying additional filtering, especially for PSS?
Best regards,
Alireza

  • Andrew Beckett

    Alireza,

    Are you using Spectre X? If so, the default is to do parasitic optimisation (which is nearly always a good thing to do as it optimizes the parasitic network resulting in a smaller matrix to solve, leading to much greater performance). What you can do is either turn off parasitic optimisation altogether via the advanced options on the Setup→High Performance Options form in ADE, or doing it for selective nets - see Spectre Tech Tips: Handling Differential Matching Circuits in Spectre X (this will be better for performance if it's specific nets you want to be able to look at it more detail).

    Note that if you explicitly save particular nets (not via a wildcard) these will be preserved.

    Andrew

  • Alireza Razzaghi
    in reply to Andrew Beckett

    Thank you very much, Andrew, for sharing your insights on how SpectreX optimizes parasitics. I am also using SpectreX, and your suggestion to save specific nets resolved the issue.

    Thank you as well for the article discussing the impact of parasitic optimization on balanced circuits such as comparators. As the article noted, SpectreX’s parasitic reduction can introduce a fictitious input‑referred offset when simulating an extracted comparator,

    Best regards,

    -Alireza

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.