Hi all,
I need to run some MonteCarlo experiments on a circuit, while sweeping some of its parameters.
However, this is not possible within the current ADEXL (BTW, why?)
I am not very SKILLed...so here's where I need your help. The sweep space is not small, spanning some 90 pts, so the idea I come up with is to use some command line approach to work around the lack of a "GUI" interface.
Basically, I would run the MC simulation 90 times, with a 'same circuit' changing only for the value of two Design Variables.
In my overly simplified mind, I would end up with a script containing some 90 lines of this sort:
spectre (myCircuitSetup, DesVar1=1,DesVar2=1)
spectre (myCircuitSetup, DesVar1=1,DesVar2=2)
spectre (myCircuitSetup, DesVar1=1,DesVar2=3)
....
Now, the questions are basically HOW to implement such thing, provided this is a viable solution (which is question number zero):
1 - I have saved the Ocean script from ADEXL interface. This script is somewhat bulky (for instance, it contains the Corner definitions although there is no actual corner selection for simulation). Being ignorant as I am, I would then blindly create 90 copies of this with the modifications needed (the different DesVar declarations) in the right places. This would involve quite some work on my part to find an effective way to 'program' this text file manipulation. I am not skilled in any of the Phyton or any other shell programming but I will find my way through it. However, this just sounds too cumbersome, therefore:
2 - Is there another method to more elegantly pass to the simulator a same netlist, with same settings all the way thru, BUT for the value of some DesVar? Some for loop?
3 - What would be the suggested method to 'distribute' this huge job efficently? Some batch thingy? Basically I would like to reproduce the intelligent batch behavior you can have running behind the scenes when you couple sweeps with distributed computing.
Any hint would be really appreciated, thanks in advance! :)
Michele