same schematic and analysis - different results?

Hi Joris,

the final truth for a simulation is always in the netlist. So I recommend to compare both netlists and search for the difference there.
When you have it there it's not that hard to find the origin of the difference in the schematic or ADE setup.

Kind regards,

Marc

Hello Marc,

thanks for the reply. I used a difference detecting tool to find the differences between the netlists: the only difference is the version number (that I chose).

This is strange...

Greetings,

Joris

Dear Joris,

Did you compare the netlist files or the input.scs files? Spectre calls the input.scs file which contains the parameter values and simulator settings. I think you might want to compare the two input.scs files not the netlist files to make sure they are identical.

Shawn

Hi Joris,

if you get different outputs than there must be a difference in the inputs. If the netlists are identical....do you work with the same Simulator version?
What about the device models, the same?  If you can't figure it out you should try customer support ( if you have access), or you could ask a second
unbiased person to have a look. In my experience, after working on something for a longer time you sometimes become "blind" for otherwise obvious things.

Regards,

Marc

Hello all,

the input.scs files are identical (apart from the  version number).

One version simulates correctly, the other one does not simulate at all (neither pss or tran ...)

The problem lies in a quarter wavelength transmission line (an mtline-component and an s2p-file (the s-parameters

were generated in another simulator)). In the original schematic, no errors occur.

In the copied schematic, it does not converge:


Warning from spectre during initial setup.
    WARNING (CMI-2504): I75: Diagonal elements of G matrix should be positive.

...

Warning from spectre during IC analysis, during periodic steady state analysis `pss'.
    WARNING (SPECTRE-223): Detected NaN at node I75:t_ib2_1.

Kind regards,

Joris

Joris,

This makes no sense. Given that all that happens is that spectre runs on the input.scs, if the input files are the same, the results should be the same too. By the time spectre sees it, it has no knowledge of the schematic.

If you're using a new enough MMSIM version, the spectre output log should also reference all the files that it is reading. You should check that too in case you've got a different path passed to the simulator somehow (you could also compare the runSimulation scripts in the netlist directory for the two cases - but the output logs would be a good start).

Andrew.

Hi Joris, did that issue got resolved? Actually I am also facing the same problem.

Dear amrvisi,

JorisLambrecht said:

I copied an original schematic  via the Library Manager and copied the corresponding

ADE L states myself (in the .artist_states folder). I then simulate this copy of the schematic with the copy of the state,

but I get a different output than in the case of the original schematic+state, even for a DC analysis?

All variables are set to the same values.

The only potential issues that I can think of that may lead to a simulation result of an identical netlist are:

1. If your PDK has been changed between the two simulations. This will potentially change model parameters for devices in your netlist and lead to different results when compared to our "original" simulation results.

2. If the version of spectre you are using has changed. In even this case, any difference in results will be quite small.

If neither 1 nor 2 are true, then the two netlists (i.e., original and "copied" version) must be different. You can easily do a comparison of the two input.scs files using tkdiff or a UNIX diff to examine potential differences.