[SOLVED] Alternative to SPICE algorithms?

[tariavo]

spice algorithms

Hello, every one,

do you know alternatives to SPICE algorithms for digital circuit simulation?
links or names of articles?

great thanks!

 

[_TopGun_]

spice algorithm equation solve

Hello, tariavo

What exactly do you mean saying "SPICE algorithm"?
You can use nodal analysys as SPICE but other methods for linearization and solving equations (as I remember, SPICE uses trapezoidal method for intergration, Newton-Raphson method for linearization and Guass-Zeidel for solving system of equation).
You can also use another method for forming system of equation, not nodal. For example branch current method, but it has some difficulties.

 

[tariavo]

spice algorithm nodal

ok, agree, the question looks strange.

well, by "SPICE algorithm" i mean the following process:

iterative loop
{
linearization and creation of matrix
solving equation
converged?
}

Are there any methods for DC analysis without "linearization" phase?
May be not so accurate, but more fast.

That is, i need fast algorithms for DC analysis, may be not so accurate as SPICE, but the method should exhibit main behavior features of digital circuits (yes, there is a switch level simulation of digital circuits but it is too rough, the precision of ~5% is desirable).

 

[_TopGun_]

fast spice algorithm

Are there any methods for DC analysis without "linearization" phase?

Well, as I know all the simulators uses linearization phase.

There re some methods that allows you to reduce time spended on creation of a counductance matrix. Especially this time difference will be significant while you simulating schemes, containing a lot of linear elements (I mean R, L, C elements on each linearization step). This method frequently used when it's needed to simulate schemes containing a lot of RC nets, for example if a scheme was obtained after RC-extraction from layout.

By the way, I saw your previous posts. Do you still need a book "The Spice Book" ? I've got one.

 

[tariavo]

fast spice algorithms

Thanks.
well, in fact, i have already an electronic version of "Inside Spice" by Ron M. Kielkowski (nevertheless, if it is not difficult, can you share "the spice book" or send to airat.halimov [at gmail.com]?).

 

[_TopGun_]

spice algorithm

By the way, about your question about skipping linearization procedure.

How do you going to solve system of nonlinear equations?
Looks like it's very diffucult to do even analitically.

I have to notice, that at the moment there are no models of semiconductor devices, that are not using nonlinear elements.

 

[tariavo]

How are you going to solve system of nonlinear equations?
Looks like it's very diffucult to do even analitically.

yes, i think so too, especially for a general case.

about the particular case of finding leakage through two closed sequential transistors there is an article on nangate site.
but in this article transistors are represented by quite rough analytical models.

thanks for the book