The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Spectre Monte Carlo Simulation for tsmc process
https://designers-guide.org/forum/YaBB.pl?num=1036784532

Message started by wael on Nov 8^th, 2002, 9:31am

Title: Spectre Monte Carlo Simulation for tsmc process
Post by wael on Nov 8^th, 2002, 9:31am

Hi

I get this error message when I run Monte Carlo simulation for tsmc 0.25u process:

Error found by spectre during Monte Carlo analysis `mc1'.
mc1: Attempt to run Monte Carlo analysis with process variations, but no
process variations were specified in statistics block.

I know it must be something related to my model setup but I don't know what. DC, AC, tran and all other analyses are running fine with the same setup.

Thanks.

Title: Re: Spectre Monte Carlo Simulation for tsmc proces
Post by Ken Kundert on Nov 8^th, 2002, 11:50am

To use a Monte Carlo analysis in Spectre you also need to add statistics blocks, which are used to describe the random variations in your parameters to the simulator. You can find information on how to do this by typing spectre -h montecarlo into your shell.

Title: Re: Spectre Monte Carlo Simulation for tsmc proces
Post by wael on Nov 8^th, 2002, 11:57am

Thanks a lot for your reply. Yes, I discovered that :) Actually I was wondering if I had to write those "statistics blocks" myself or should they have been included in the design kit? But this question might be beyond the scope of this forum.

Thanks.

Title: Re: Spectre Monte Carlo Simulation for tsmc proces
Post by Ken Kundert on Nov 8^th, 2002, 1:33pm

It is hard to imagine individual users having enough information to describe the statistical variations in the process, so the process blocks should come from the foundries. However, few actually do provide this information. I am aware of at least one large company having statistical information, and the corresponding Monte Carlo process blocks, for the TSMC processes, but I do not know how they acquired them.

Title: Re: Spectre Monte Carlo Simulation for tsmc proces
Post by nus_lin on Apr 6^th, 2005, 10:55pm

dear guys,

i am using analog artist to simulate 0.35um 4metal process by AMS. i am trying to use the monte carlo tools provided, but it looks that the library file doesn't point to the statistic model file and the results remain unchanged every run. so can you tell me the name of model file and how to set up the path?

thanks

Title: Re: Spectre Monte Carlo Simulation for tsmc proces
Post by Paul on Apr 7^th, 2005, 12:39am

Hello Wael,

Sometimes a file with mismatch measurement results is included in the design kit, and it is up to the customer to translate this into simulation parameters. In case you cannot get statistical parameters from your foundry, you can still add some "generic" statistical parameters to give you a feel of the sensitivity of your circuit to device mismatch. Anyway, from my personal perspective, if you rely on the exact values of mismatch parameters to make your circuit work, I guess it won't. In modern technologies, you have phenomena like NBTI, hot carriers and so on which modify the threshold voltage over time and which are usually not included in the mismatch parameters.

You can get an order of magnitude for mismatch parameters from the following papers (ieeexplore is your friend ;) ):

An easy-to-use mismatch model for the MOS transistor
Croon, J.A.; Rosmeulen, M.; Decoutere, S.; Sansen, W.; Maes, H.E.;
IEEE Journal of Solid-State Circuits, Volume 37

Physical modeling and prediction of the matching properties of MOSFETs
Croon, J.A.; Decoutere, S.; Sansen, W.; Maes, H.E.;
ESSDERC 2004

Paul

Title: Re: Spectre Monte Carlo Simulation for tsmc proces
Post by nus_lin on Apr 7^th, 2005, 6:02am

i have successfully done a monte carlo simulation.

thank ken, paul. your comments give me clue to setup the simulation enviroment.