MonteCarlo simulation is taking much longer time than expected

There are two different "nominals" here (unfortunately ADE XL uses a rather misleading naming convention). In the Corners section in the Data View assistant there is something called "Nominal". This actually would be better named "Test Models" or something like that - in essence it uses whatever models have been set up in each test rather than the corner setup. Whether it's nominal depends on what was in the test - you could have picked the fast model section or something like that in the test.

If you have this "nominal" corner enabled, it will run with that set of models, and assuming there are statistical parameters defined (with mismatch in your case) in the model files, then it will run a mismatch analysis around those default models. If you have any other corners enabled, it will run mismatch (or process) centred around the models in that corner too. So as I said before, you have two corners, so that's twice the number of runs - both will be statistical analyses though.

Secondly there is a choice on the monte carlo options form called "Run nominal simulation". What that does is run a single non-statistical simulation at the beginning of each monte carlo run (i.e. with whatever models you've given it, just without any statistical simulation). Originally this was done as a check - if any of  your measurements failed, it would then not bother to run the rest of the monte carlo. However, that behaviour changed a while back (during IC616) and anyway it was never possible to plot the results of this nominal simulation. So in IC617 this choice has been removed in ADE Assembler and Explorer (it's still there in ADE XL because we didn't want to touch it; however, I think it's a waste of time to use it).

As for your last question, it will be centred around whatever models you're using - and that will depend on the model setup in the PDK you're using. You may be able to run mismatch-only simulation about any specific corner you pick, or you may have a specific corner for monte carlo which may be based around the typical process corner.

Hope that helps,

Regards,

Andrew

Dear Andrew,

Thanks a lot for trying to clear my doubt.

But, still I am NOT convinced.

Let me put it clear about  my understanding from  your FIRST statement -->

"If you have this "nominal" corner enabled, it will run with that set of models, and assuming there are statistical parameters defined (with mismatch in your case) in the model files, then it will run a mismatch analysis around those default models. If you have any other corners enabled, it will run mismatch (or process) centred around the models in that corner too. So as I said before, you have two corners, so that's twice the number of runs - both will be statistical analyses though." <--

as below :- 

Suppose I have two MonteCarlo (Statistical) model files  each for MOS and Capacitors. i.e. MOS_model.process.mc.scs and CAP_model.process.mc.scs.

MOS model file shows the following options: - TT, SS, FF (process options) and MC (statistical option).  and CAP model file shows the following options:- TYP, MAX, MIN (process options) and MC (statistical option)

Now for "Nominal" corner let's  say, I kept MOS_model.process.mc.scs --> SS  and CAP_model.process.mc.scs --> MAX.

What I have understood is:- With each "Nominal" corner run, a MISMATCH analysis  is performed centered around MOS-SS and CAP-MAX corner. 

For MC corner setting :-  

I kept MOS_model.process.mc.scs --> MC  and CAP_model.process.mc.scs --> MC.

What I have understood is:- With each "Nominal" corner run, a MISMATCH analysis  is performed centered around MOS-SS and CAP-MAX corner. 

So, MC corner is NOT doing anything new from "Nominal" corner. It is just repeating !!

Now let me put it clear about  my understanding from  your SECOND statement -->

"As for your last question, it will be centred around whatever models you're using - and that will depend on the model setup in the PDK you're using. You may be able to run mismatch-only simulation about any specific corner you pick, or you may have a specific corner for monte carlo which may be based around the typical process corner."  <--

Let's say, now I have realized that "Nominal" is just doubling run. So the Idea is disable it.  The MC corner is set as:-  MOS_model.process.mc.scs --> MC  and CAP_model.process.mc.scs --> MC.

With the above setting, around which process corner does the monte carlo perform the Statistical analysis ?

If I set MC corner as:-  MOS_model.process.mc.scs --> TT  and CAP_model.process.mc.scs --> MIN,  around which process corner does the monte carlo perform the Statistical analysis ?

I think it will be MOS-TT and CAP-MIN right ?

Kind Regards,

I don't think I can answer these questions because it completely depends on how the sections in the model files have been written. It might be that they have statistical mismatch parameters in all sections, and then process+mismatch statistical parameters in the MC section - I have no idea (this is just a wild guess). If it was that, then I'd expect that if the test models were set to:

MOS_model.process.mc.scs --> SS and CAP_model.process.mc.scs --> MAX

and the corner was set to:

MOS_model.process.mc.scs --> MC and CAP_model.process.mc.scs --> MC

Then I'd expect that a mismatch-only simulation with both these enabled would do mismatch around SS/MAX for the "nominal" and mismatch around typical (assuming that the "MC" section is centred around the typical process corner). So they may not be the same.

However, this is all guesswork - I have no idea of how your models are set up (since I don't know what technology you're using - and even then I'm not sure I could talk about it here). Presumably the foundry documented what the model sections do and how they should be used somewhere... if not, ask them!

For your second concern, I can't answer as well because I don't know where these MC model sections are centred about (but I'd guess they are centred around the typical process values). So in your terminology that would be MOS-TT and CAP-TYP (not sure why it would be CAP-MIN). However, this too is guesswork.

Regards,

Andrew.

Dear Andrew,

Thanks a lot. I will talk to the foundry people for the exact clarification.

Kind Regards,