How to correct the device model?

[lennonwang]

hi, everyone, when simulating, we often find that the device model is wrong, so the simulating result is also suspicious. Now I want to know, if we couldn't get the correct one, What I could do for correcting the wrong one? Thanks in advance.

 

[shiowjyh]

Hi,
Can you tell us more about what simulation tools and spice models are you using??
Not enough clues to identify...

Regards,

 

[superluminal]

I don't think that lennonwang means the spice model parameters (like W,L , VTH ..etc) because varing it is just done by overwriting the old values.

I think lennonwang means to change the model equations itself . For example , you can get some negative gm or capacitance when using BSIM3v3 models. This can not be changed at all if it is implemented inside a spice program. Either you make your new model from A-Z or change your spice program. Usually models implemented inside spice are IP protected . One exception is the free models source code available at groups like BSIM online. Spice vendors take these "standards" and modify them (using a faster equation solver algorithm , fixing a bug ..etc) and then put in thier spice.You can get the code and modify it as you want but if you are not VERY MUCH CLEVER , don't even think !

 

[lennonwang]

I mean IBIS models. Because spice models are based on transistor level. To correct it is not such a easy thing. when we do a board-level system simulation, we must first assure that the IBIS models are correct to show its electrical characteristics. Now I found the waveform is not up to the result I have predicted. Even the waveform is not convergent and I couldn't see the result. So I have to doubt the accuracy of the model. I mean how to modify the I-V curve or V-t curve and so on.

Best regards.