I did a comparison script that compare Xyce and ngspice results for different type of sweeps. I tried dc sweep for sky130_fd_pr__nfet_01v8 for tt corner simulating all Bins w and L in the model and i found max error between the two simulators was 0.033% Here is the link for comparison script and results. https://github.com/mabrains/xyce_sky130/tree/main/testcases/xyce_vs_ngspice @User @User @User @User

@Ali Elshorbagy Thanks for providing such a detailed bug report. I can’t look at this today, but will look at it tomorrow.

@Ali Elshorbagy thanks! good work. @Eric Keiter do you expect them to match? I see the error is small at least for the image shown above.

@Ali Elshorbagy, similar to @mkk I am curious; did you post this to indicate a problem, or that you thought that the comparison had been successful? Also, are these comparisons for DC, AC or transient? If they are for transient these differences don’t concern me.

@Eric Keiter The intent of these test to align the two tools both on the tool and model card. What we are trying to show above that the current 2 model cards match ngspice and we believe it’s useable on DC run. The above is only DC. We are now working on the AC and later we will work on Transient. I believe the above it’s successful comparison. @Ali Elshorbagy please mention in your post that either you believe it has been successful comparison, tool issue, model card issue.

@Amro Tork thanks for the info. I think that the binning issue (which I mentioned in another thread) makes it so that the current version of Xyce should not be used to sweep L and W parameters for binned models. But otherwise, my experience has been that we generally will match ngspice (and other codes) very well for this process node.