It is interesting to note that the Gummel-Poon model for BJT transistors has been superseded (sort of, but not quite, see below) by the VBIC model in 1995, but that it is used by almost no-one outside of research labs.
The VBIC model more accurately simulates bipolar junction transistors and their parasitic behaviour. It was an industry wide effort, hence the name (note the ‘inter-company’): Vertical Bipolar Inter-Company model.
I know that Gummel-Poon is sufficiently accurate in many cases, but not quite accurate in situations where life is getting interesting. Early-effect, breakdown and avalanche effects, parasitic capacitance, noise simulations, temperature effects, etc. are examples of where simulations would benefit from VBIC.
The VBIC model has been implemented in ngspice, so in theory we could all be using it, but the availability of transistor models is virtually zero.
I have not found any manufacturer supplying VBIC models for its products, but perhaps I have overlooked something. For an engineer without access to the internal information of manufacturers it is hardly possible to measure and extract VBIC parameters for a particular transistor, because the parameters depend on an intimate knowledge of the structure of the transistor.
We are now more than 20 years after the introduction of the industry wide VBIC model, but in practical terms, we are still not able to use it for our designs.
Have I overlooked something?
Are there resources for some BJT’s available?
Is VBIC dead in the water?
Cool. Never heard of VBIC before. I think the best way to find more info on what you want is to search for the phrase VBIC parameter extraction or VBIC95 parameter extraction.
That last one has a 2N3904 model, provided by our good friend @holger. Maybe he has some more insight into this or at least his own opinions?
Anyway, I think the reason why you aren’t finding models for VBIC is a simple supply/demand issue. There is not much demand for super accurate BJT models, since like you said, Gummel-Poon is sufficient in most applications. Hell, even Ebers-Moll works great for many things. So why would manufacturers waste time making sophisticated models, when the target audience is so small? I think most resources have been pushed into MOSFET models and subcircuits, since switch-mode-power-supply stuff has really dominated the SPICE modeling arena. I dunno, just my guess.
Thanks for the @holger link. I had not seen that.
I had found a couple of those doc’s myself, but I also noticed a surprising lack of interest by manufacturers. After all, why go through the obvious effort to develop a model and then largely ignore it.
I was hoping I had not yet looked in the right places.