Anyone done trim-and-sim with Monte Carlo?

 Try using a transient anlysis to do this.  Make a verilogA block to model the tester.  In the first part of the transient analysis use this to measure and trim the part.  Then in the last half of the tran run, sweep the temp using the swept parameter option in spectre tran runs.

Is this possible in 5.1? I don't see an option to change the temperature during a transient run. I probably need two sims - one at room temp, to trim, and then a sweep.

Cadence has opened ER CCMPR00790644 for this.

 The ability to do this sweep only depends on spectre versions. If you have>spectre 6 I think you can do it.  You just need to use additional arguments in ADE transient options to feed the sweep options to spectre.  Do spectre -h tran and look for dynamic parameters

 > spectre -V
@(#)$CDS: spectre  version 7.1.0 32bit 04/21/2009 11:50 (sjfdl189) $

I don't see this on the options form though.  My options are :

 Sim interval parameters : start, outputstart

Time step params : step, maxstep

Initial Condition Params : ic : dc,node, dev,all ; skipdc : yes, no, rampup, autodc, waveless, sigrampup; readic

Convergence params : readns, cmin

State file params : write, writefinal, saveclock, saveperiod, savetime, savefil, recover

Integration method params : method : euler, etc

Accuracy params : realref: pointlocal, alllocal, etc; iteratio

Annotation params : stats, annotate : no,title,sweep, status, steps

 Output params : save : selected,lvlpub,lvl,allpup, all; nestlvl; oppoint : rawfile, screen,logfile, no; skipstart, skipstop, skipcount, strobeperiod, strobedelay, compression, infotimes, flushpoints, flushtime, flushofftime, actimes, acnames

Newton params : maxiters, restart 

Captab params : captab

Additional params : additionalParams 

 Not surewhy it is not there.  You may need to ask customer support.

 Not surewhy it is not there.  You may need to ask customer support.

 Not sure why this is the case.  You may need to check with customer support

It can be (also) done by running a pair simulations for each monte carlo run. The first simulation would be a "trim" simulation to find out  the variable values. The second sim would be  the real sim run with these values.

The key is that you use exactly same "seed" and "run" number for each "trim" and "sim" pair. As long as seed/run numbers are the same, the netlist would be identical for each pair of sims.

This can be a lot of manual labor, but it might be possible to automate it with an oceanScript. You can have an outer for loop that increments run numbers. Within this loop, you run 2 monte carlo analyses in sequence, each with number of runs=1 and starting run=loop variable. You process the results out of the first monte Carlo and then set your variables etc before you start your second monte carlo.

You probably got a general idea. However you may have work on getting the ocean-script code right.

 You can do this with spectreMDL.  The tutorial is the best thing to look at.  You can set up two simulations and set the parameters for the second simulation based on the results of the first simulation.  You can set this up to run over temp.  There are examples of this in the spectreMDL tutorial.  This should let you do what you need.