If the bank/rose method is used in the devsim?

1

I use the default option in the function named equation, but the speed of the calculation is still not very fast. The maximum step only 1 V. I want to ask whether the bank/rose method is used in the devsim. Thanks for your attention.

2

Can you please explain what the Bank Rose method is, or provide a reference to the paper? They have several publications from that group.

3

I have sent the pdfs to your e-mail. Thank you!

4

Hi,

I have not received your email. Please send them to [email protected]

Thanks,

Juan

5

If possible, I can help you to add the bank/rose method in the devsim.

6

Thanks for sending the papers. They use a lot of mathematical sophistication, which I find difficult to understand.

The default mode in DEVSIM is to be a full Newton solver, with a complete Jacobian matrix with all of the derivatives. There is an iterative mode, but it uses a very simple preconditioner with the GMRES method, and would not be robust enough for most problems.

For transient mode, I have implemented the TRBDF method mentioned here:

Transient Simulation of Silicon Devices and Circuits

and expressed in equations 31-33. I do not use the same LTE estimation or any of the other methods in the paper.

Please let me know what aspects of their methods you are interested in.
It is possible to get the matrix and rhs entries directly, and to perform your own transient method, but this would require much work.

Please let me know what type of problem you are trying to solve. You can reach me via email or in this forum.

7

Hello @fendou1997

Your contribution to the project would be most appreciated. I am able to help you if you wish to implement your own solver. It should be possible to do this in either C++ or Python. I am also able to expose more capability to interfaces to support your algorithm.