Using models from MOSIS for BSIM3 simulation

BSIM3 question

Hi

plz tell me somebody to how much extent the NCSU CDK can be used. While simulating the transistors and MOS BSIM3 which model files to be used. CAn they scale to .18u M technolgy??

Re: BSIM3 question

gunturikishore said:

Hi

plz tell me somebody to how much extent the NCSU CDK can be used. While simulating the transistors and MOS BSIM3 which model files to be used. CAn they scale to .18u M technolgy??

Click to expand...

I believe the hspice model files of NCSU CDK were measured from the TSMC process directly. The only drawback is that they does not support different process corners.

Re: BSIM3 question

gunturikishore said:

Hi
plz tell me somebody to how much extent the NCSU CDK can be used. While simulating the transistors and MOS BSIM3 which model files to be used. CAn they scale to .18u M technolgy??

Click to expand...

Forget about the spice models coming with NCSU CDK. For one thing, they actually come from mosis.org, i.e., roughly fitted models by MOSIS based on MOSIS's test structures. For another, it is for TSMC 0.35 um CMOS.

If you want to do any serious design, you have to get TSMC's 0.18um Spectre/SpectreS models via MOSIS. I used them to design two chips for 1-7 GHz. They work OK.

Re: BSIM3 question

I am presently working on the BSIM3v3.2 parameters. What are Corner Parameters?

Re: BSIM3 question

DoctorX said:

If you want to do any serious design, you have to get TSMC's 0.18um Spectre/SpectreS models via MOSIS.

Click to expand...

How can I get TSMC's models from MOSIS? I'd like to know more about it. Many thanks.

Re: BSIM3 question

Tilhu:
**broken link removed**
download it. its free.
srivats

Re: BSIM3 question

MOSIS add one or two test structures on some wafers for each batch (so called "run") of fabrications, and they extract SPICE models from them, and post the models at MOSIS's website as free models. So called "run-specific models".

On the other hand, vendors, such as TSMC, do the same (every year or so), and release the models as "vendor models", and their distribution is controlled.

The MOSIS models are more up-to-date, reflecting the speed difference of each run; the vendor models are more accurate. For analog design, you should primarily use the latter.

Big companies, when using outside fabs, typically fabricate their own test patterns and extract their own SPICE models. University workers generally cannot afford to do that.

Re: BSIM3 question

DoctorX wote:
.."The MOSIS models are more up-to-date, reflecting the speed difference of each run; the vendor models are more accurate. For analog design, you should primarily use the latter. "....

If you look at the MOSIS website, there is something called Corner parameters. What are those? Also vendor release of models are little bit expensive. So if there are no choices ofmodels except the MOSIS models, then what are the probabilities of failure when using them?
thanks in advance..
srivats