Stats

  • Replies 1
  • Subscribers 132
  • Views 32
  • Members are here 0

Potential Simulator Artifacts: Glitches in Isolated Blocks during Transient Analysis

SpiceMonkey

I am experiencing an issue in Spectre (version 20.1) while simulating channel isolation in a testbench containing two completely isolated blocks.

Setup Details:

  • Both blocks are placed in the same schematic testbench.

  • All control signals are provided by independent ideal vdc sources from analogLib.

  • There are no shared nets or signals between the two blocks.

  • The issue: Block 1 is meant to remain static, but a glitch appears in Block 1 precisely at the moment Block 2 is enabled.

I am concerned that this is a simulator artifact rather than actual crosstalk, as these blocks are configured to be electrically isolated. I have already attempted to mitigate this by setting relref = alllocal and tightening the tolerance to reltol = 1e-6, but the glitch persists.

It appears that the convergence of the static block is being affected by the large signal transitions occurring in the second block. Given that relref = alllocal has not resolved this, I would appreciate your insights on the following:

  • Is this a known behavior in Spectre 20.1 regarding global vs. local convergence when large signal transients are present?

  • Are there additional simulator configurations (e.g., specific errpreset settings, integration methods, or advanced convergence controls) that can help suppress these numerical artifacts?

  • Could this be related to how Spectre handles the matrix solution when multiple blocks are active, even if they are topologically disconnected?

Any advice or suggestions on how to isolate the cause of these glitches would be greatly appreciated.

Thank you!

  • Andrew Beckett

    It's hard to comment on the details of this without seeing it, but some general observations:

    1. There is only one matrix being solved. The fact that they are in separate subckts doesn't alter that - it doesn't solve each subckt separately. I am assuming that the glitch must be small enough to be below the error criteria for convergence. Circuitry in one part of a design can certain impact others (e.g. it might need to take more time steps because of activity in one block, and there's only a single time axis) - Spectre (unless we're talking about Fastspice approaches using Spectre FX) is not a multi-rate engine.
    2. Do you have the supplies and grounds properly connected (this is more likely to be noticeable when you have some element of the circuit that is floating).
    3. It's unlikely to be anything specific to a particular version of Spectre.
    4. The best solution here would be to provide the testcase to customer support so that it can be checked to ensure that the mechanism is understood and to give a better explanation.
    5. Setting the tolerance to 1e-6 seems overkill to me (that's a good way of killing the performance of the simulator) - unless you really need such high accuracy. It's more likely that there is some circuit issue which is influencing this.

    Andrew

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.