Global variable sweep in Monte carlo analysis

Which version of the tools are you using? Which tool are you using, in fact? The guidelines ask you to provide this kind of information for a reason...

I'm guessing you're using ADE XL, because ADE Assembler supports this. If you really have to use the older ADE XL tool, you can put the variable sweep in the corner definitions and sweep that way.

Regards,

Andrew.

Hi Andrew , I am using IC 6.1.7-64b.500.23  ADE XL. 

Thanks for the reply,

How can I ensure that I have included all the required model files for MC simulation ?

In my case MC simulation is running even without including any additional model from the PDK (I have the same model setup that was used by me in normal transient simulations).

Hence I am a bit worried if I am not missing anything ?

Monte Carlo won't run if you have no statistics blocks in the model files - but of course there's no way for the simulator to know that you've included the correct files. It can't possibly know that, unfortunately. You just need to check carefully that the right files and sections are included, as you should for any simulation.

Andrew.

Monte Carlo won't run if you have no statistics blocks in the model files - but of course there's no way for the simulator to know that you've included the correct files. It can't possibly know that, unfortunately. You just need to check carefully that the right files and sections are included, as you should for any simulation.

Andrew.

Thanks for the reply Andrew. 

How can I export the entire ADE XL setup in Assembler so that all the global variables/corners/sweeps are also transferred as it is ? 

I tried to load ADE setup into Assembler but I see that the global variables are all de-selected as contrast to when i loaded it into ADE XL  (there are more than 50 of them).

You would normally go over the adexl view in the library manager, use Right Mouse->Open With, pick "Assembler" and get it to create a new maestro view. Nothing should be deselected.

Normally if you're creating a new maestro view from scratch, then design variables in each test do not get automatically promoted to global variables (which they do in ADE XL), mainly because you can sweep local (design) variables in Assembler so there's less need to make them global. However, you can easily convert them to global variables by selecting them under the test and picking Right Mouse->Create/Update Global Variables (there's an option to change this default behaviour of not promoting them, but mostly in the past everything was made global because it was the only way of varying them).

Andrew.

Thanks for the steps Andrew. 

While doing the Monte carlo simulations with the below settings , I am getting an error as indicated below for some of the MC points (8 out of 100)  - does this appears to be a DC convergence issue ? If so how to go about resolving this ?

Warning from spectre during info `dcOpInfo', during Monte Carlo analysis `mc1'.
WARNING (SPECTRE-16002): Monte Carlo iteration 36 terminated prematurely because of the following error(s):
Error found by spectre during info `dcOpInfo', during Monte Carlo analysis `mc1'.
ERROR (SPECTRE-25):
Message from ADE: Simulation stopped by ADE because the Simulation Run Timeout has exceeded its specified value.L

There's not enough information to go on, from purely one message in the spectre log file.

I suggest you contact customer support; debugging a convergence issue in a public forum tends to be pretty difficult because of not being able to see enough data.

Andrew.

HI Andrew, there is one particular net where the matrix calculations seem to be  facing an issue. 

Zero diagonal found in Jacobian at `net0130' and `net0130'.

are there any diagnostics that i can try around this node ?

That may suggest that the node is floating (or near floating).

Andrew.