Idea: Using SPICE models?

[fritzw]

Using SPICE models provided by manufacturers seems like a very powerful way to extend the number of components that can be used in circuitjs. I've done a little bit of research and found that there are basically two types of SPICE models and 3 types of files:

1. Compact models or parameter models. These consist only of parameters for the C/C++ model that is compiled into the simulator.
2. Subcircuit models consist of a netlist of other components, just like CompositeElm in circuitjs.
3. There are also files for circuit symbols, which are used to represent the models on screen.

For parameter models one would need a compatible Java implementation of the model. Old versions of SPICE are open source, so perhaps this could be a basis for porting the models to Java and for verification. However, I'm not sure how compatible the old SPICE versions are with the model files you can download from manufactureres. At least for transistors, new models have been developed over time (see http://bsim.berkeley.edu/models/ and https://en.wikipedia.org/wiki/SPICE#Device_models).

Once basic models are supported, subcircuit models should be easy.

A way to draw custom symbols would be nice anyway, so if one were to be implemented, looking at the SPICE symbol file format would probably be worthwhile to ensure future compatibility.

I think the first step towards this goal would have to be more research on the variety of parameter models and their mutual compatibility.

What do you think of these ideas?

[pfalstad]

I spent some time working on a feature that would automatically import spice subcircuits. You can look at the branch "spiceimport". I don't feel like it's ready to be added to the main branch, and maybe it never will be, because a lot of testing would be required. But I used it to create some models like the TL431.

Re parameter models, it's not really the goal of the simulator to reimplement SPICE. It's intended as a learning tool. But in order to add more components (so people can use them and learn about them), I've found it useful to make the simulator act more like SPICE. The current transistor model is right out of SPICE.

[fritzw]

You are right, reimplementing SPICE should probably not be a goal. But even for a learning tool, using accurate simulation models (if available) seems to be a good idea. After all, when learning about electronics it can be pretty unhelpful if the real MOSFET on your breadboard behaves radically different from the simulation you did before. If that turns the simulator into "nearly SPICE" (but with a better user interface) that's not a bad thing, I would say. Unless the simulation becomes extremely slow.

So, for the bipolar transistor model, it should be possible to apply the parameters from a SPICE model file (e.g. https://www.cordellaudio.com/book/Cordell-Models.txt) and get (nearly) the same behavior as when using the same model in SPICE? That's really nice. Maybe I'll do some work to provide an interface for importing such model in the future.

Just in theory, would you accept contributions that port SPICE model code to Java, for example for the MOSFET model?

[pfalstad]

Yes, the one thing that's included from the SPICE transistor model is capacitance. I decided to leave that out, but maybe I shouldn't have.

Yes, I think I would accept contributions. I'm not sure I want to port all the MOSFET models over.. There's 10 of them right? What I really want is support for subthreshold conduction, but last time I looked, none of the SPICE models supported that, at least in the free code I was looking at.