How to setup + or - n*sigma in monte carlo hspice

[hoatre]

Hi all,

I want to setup + or - n*sigma only in monte carlo hspice.

Can anybody help me how to do setup?

Thanks,

 

[ic98]

what do you mean by pos/neg sigma? do you know the mean of Gaussian distribution?

 

[erikl]

I want to setup + or - n*sigma only in monte carlo hspice.

See e.g. HSPICE® User Guide: Simulation and Analysis B-2008.09 , p. 318 .
If you just want MC runs for e.g. -3σ and +3σ, you could use (for a transient analysis):

.tran 10p 1n sweep sigma -3 3 6
.tran 10p 1n sweep monte=100

 

[palmeiras]

Hi erikl,

Your first sentence: variation of -3 to +3 sigma. But what does mean "6"?
By the way, after my Monte Carlo simulation, the HSPICE simulator does not show the mean and sigma found. As shown in this picture described in the manual.
Do you know how to get this "summary"?

Thank you very much,

 

[erikl]

Your first sentence: variation of -3 to +3 sigma. But what does mean "6"?

Simply the step size between -3 & +3 , so only these 2 analyses are run. Of course you could use a POI list instead.

after my Monte Carlo simulation, the HSPICE simulator does not show the mean and sigma found. As shown in this picture described in the manual. Do you know how to get this "summary"?

Not really :-(
.PRINT set?
.OPTION MCBRIEF not set?

 

[palmeiras]

Hi erikl,

.OPTION MCBRIEF is already equal to zero. :-(
What does you mean with ".Print set"?
Thank you very much.

 

[erikl]

What do you mean with ".Print set"?

Just the standard .PRINT statement. Try:
.PRINT MC statistics

 

[hoatre]

Hi erikl,

The .measure statement matches the last analysis command in the netlist before the .measure statement.That mean in your comment, .meas matches the second .tran statement.
But I test with only command:
.tran 0.001ns 10ns sweep sigma 0 3 0.05
.meas tran td_in_to_out1 trig v(in) val='0.5*vsup' td =1ns rise=1 targ v(out1)
+val='0.5*vsup' td=1ns rise=1
the parameter sigma variant,but values td_in_to_out1 don't change.
I don't know, how to parameter sigma affect to model ?
I want run monte carlo and can control sigma , but I don't find the way.

Thanks for your comment

 

[erikl]

... the parameter sigma variant,but values td_in_to_out1 don't change.
I don't know, how to parameter sigma affect to model ?

I think HSPICE doesn't find any sigma-dependent parameter in your netlist (or any .INCLUDEd file): sigma is not an intrinsic parameter of HSPICE; its effect has to be defined by a parameter definition, which changes some relevant model parameter value(s), like, e.g. Vt_mos(sigma)=AVt_mos/sqrt(WL)*sqrt2/2*sigma
where AVt_mos is the process-dependent Vt spread parameter (published by the foundry), W*L the Pelgrom size dependency, sigma is the spread value to be considered (for yield) in your analysis, and Vt_mos(sigma) the actual sigma-dependent Vt value to be applied in the analysis.

Foundry PDKs use appropriate definition files with quite a lot of sigma-dependent parameters together with other mismatch and corner-dependent variations, which have to be .INCLUDEd into the main simulation file.

The MC simulation then runs the given count of analyses by selecting random numbers between your given sigma range (like -3 .. +3).