The ddx() of a state variable will be 0. I suspect you want to use something that has a direct algebraic dependence on V(g).

BTW, we work with Prof. Angelov on keeping his model (and the GaN version) up-to-date, see the design kit http://www.tiburon-da.com/design_kits/Angelov_DK.zip

You can see how we get gm in that model. It's an ADS design kit, but you can unzip and then find the models in the veriloga directory.

@ Marekm,
Thanks a lot for replying. That's exactly what i need: the last section. I'm trying to understand it. I have to implement the pucel and pospiezalski equation in angelov. and i would like for verification to calculate each noise model  using the parameter of the other one.

@ ken
i want to expand the angelov model  with the noise equation of pospiezalski and pucel . The Pucel equation look at the gate like
: Cgs0_T * Cgs0_T * 4.0 * `P_K * T * NoiseR / gm.

i have realised gm=ddx(Ids,V(g)) allways return a zero.

kem

Marek,
   Oh, I see. You do know how Ids relates to V(g).

Kem,
   Can you see how nobody but Merek could answer your question, and he could only answer it because he had detailed knowledge about the model that you did not give. Asking for help like you did is a bit of an art, and to maximize the chance of getting a meaningful answer, you have to give enough information so that people who are not intimately familiar with what you are doing can help you. That means you have to give more information. At a minimum, you should have given enough of the model so that we could see the relationship between Ids and V(g).

-Ken