I don't think there are many people who have done this from scratch.  Maybe you can find the old Spice2 documentation from Berkeley?  There's an awful lot of different pieces that go into a simulator; is there something specific you're looking for?

There's also:

Steady-State Methods for Simulating Analog and Microwave Circuits (The Springer International Series in Engineering and Computer Science)
by Kenneth S. Kundert (Author), Jacob K. White (Author), Alberto L. Sangiovanni-Vincentelli (Author)

ISBN 978-0792390695

Regards,

Andrew.

Geoffrey_Coram wrote on Apr 15^th, 2009, 9:48am:


Thanks for all the tips. However, I think this book you mention is more on how to use the simulator rather than how to create one. I have a copy of it and read it once long ago (very useful). The other texts recommended by Andrew might be of more interest. I also found references to those elsewhere.

I was hoping to get hold of some sort of class notes for a course on the subject. It seems though that UCB notes are only accessible to UCB students (quite the opposite of MIT OCW). Anyway, I will look further.

Thanks,

Vivek

The most inspiring book at the beginning of the 80's was this:

L. O. Chua and P. M. Lin, Computer-aided Analysis of Electronic Circuits, Prentice-Hall, 1975

So it gives deep understanding of the solution techniques starting from graph theory going forward to the nodal, modified nodal, hybrid analysis than over to integration methods. It have also very interesting chapters for special solution methods. So homotopy and shooting is already 1975 a textbook documented  technique!

I think there is still space for improvements for circuit simulation.

The point is that in circuit simulation there is not enough money to found solution for the most depemanding problem of mapping sparse matrix methods and sparse integration with multi-timesteps to parallel machines.

To be honest It takes me some years to understand most of the clues of today circuit simulation techniques. Now I also understand what is missing, or what could be improved. But to write your own is a long way. I guess decades for a focused person. That is needed to understand the hole code base with all the math background where you then could add a smaller code feature.

Thanks all for the responses:

Geoffrey: I agree that there is a lot of stuff like netlist handling, stamping etc. There used to be a course at the university where I went dealing with all that stuff but at that time, I was not interested in simulation and so never took it. There are also no online notes.

Reiner: Indeed, there must be a lot that can be optimized and improved, and something that one could spend a lifetime on. However, I do not even imagine trying something like this. For me, it is of interest to get the basic knowledge of various practical aspects which go into making a basic simulator. The only aspect which seems to be well covered in books is the theory of numerical computation. The rest of the practical aspects are not even mentioned anywhere, but I want to look at this book by Leon Chua (thanks for the tip).

AG: Thanks for the links. I will go through this and hopefully this is something.

Best regards,

Vivek