Ocean script not converging in transient simulation

Milind,

I'd check the resulting input.scs files and compare them - there must be a difference if the simulator is behaving differently (you should be able to diff the two files).

One likely explanation is that saving an OCEAN script does not (by default) save settings which are default in your Virtuoso environment, and if those defaults (set by cdsenv variables) are different from the default defaults (!), this could lead to a difference. You can save a more complete OCEAN scripts (which potentially as a lot of redundant statements in it) by using:

envSetVal("asimenv.misc" "saveDefaultsToOCEAN" 'boolean  t)

(you might need to set this before opening ADE L, I can't remember, sorry!) - and then save the OCEAN script.

Andrew

Thank you for your response. Yes I did that based on my other searches through this forum. The files are exactly the same till the first time point of the transient step. The only difference before that are things like time, license checkout time, process ID, etc.

So now I tried doing the envSetVal command that you suggested. I restarted the Cadence Session and then ran this before opening the ADE session. Ran the simulation in ADE. Saved the Ocean script and tried the ocean script. The script failed convergence, although I do see a lot of additional parameters in the ocean script.

Now I tried the command once more and resaved the ocean script. This time it worked! I will experiment a bit more and follow up here.

That sounds as if you’ve compared the output log (spectre.out) rather than the input netlist (input.scs). If the results are different the input really must be different because it is calling the same executable in both cases (assuming too that the command-line options for the simulator are the same).

Andrew

I think I spoke too soon. Tried it again. Doesn't converge. Also when it did somehow the spectre.out had a difference. The one from ADE-L was clean at the starting of the transient steps, for the one in ocean it had a Notice saying solution failed to converge and disaster recovery algorithm is enabled, after which it was able to continue. Now it is not able to get there as well.

However I see that it converges in the script when I have APS++ enabled. But still it is not the same output log as ADE-L output log.

I checked the input.scs and it is the same same in both cases. The only difference is the order in which of the design variables are listed on the top (also the ocean script version the variables are written in scientific format while in ADEL it is using the micro-u or M symbols). The other difference is the order in which the save elements are written at the end. Other than that the input.scs that I get after running the simulation are the same.

I can't really see how Spectre can do different things given the same netlist (if the command-line options are the same). I suggest you contact customer support so that an application engineer can take a look.

Andrew

Ok, I have initiated that through the CAD group here. Thanks.