Has anyone built simulation models from scratch for mosfets and diodes using their datasheets? If yes, can someone share example(s) on how these things can be done/made, how to do this?
I’ll be surprised if anyone claims this is possible, but stranger things have happened.
Why be so strange and not possible? For example the case when any simulation model of the respective mosfet/diode not exist.
My general approach is to draft test fixtures that will extract the plots in the datasheet and twiddle model parameters to match. Since I’m familiar with the device equations, I’m aware which part of the plot is most sensitive to which model parameter, so I adjust that one to match and then look at a different part of the plot to and twiddle the model parameter that has the strongest impact on that section of the plot. I call it Manhattan iteration, since I find it reminiscent of the navigating a city on a rectangular grid.
To get familiar with the device equations, I like the 2nd Edition of Semiconductor Device Modeling with SPICE by Giuseppe Massobrio and Paolo Antognetti, McGraw Hill, 1993
I’m not expecting the TS’s first one any time soon. Doesn’t sound like anything I’d care to attempt, but to each their own.
Does that mean you have a set of standard test fixtures to examine various properties of transistor models so you don’t have recreate them from scratch every other Tuesday? Just wondering.
Well, sort of. I have a tool that digitizes a datasheet and does the Manhattan optimization for the DC parameters for diodes, JFETs, and MOSFETs. I kept its existence a secret since I used to use it for contract modeling for some semiconductor mfgs.
But I’m the only person who ever used it, so it’s quite rough. The only people that have ever seen it besides myself are a modeling group at Dresden TU. But if I ever polish it point where it could stand up to the general public, I’ll probably donate the tool to the QSPICE cause.
I’m sure Cornel would appreciate the donation.
My question was more about creating test simulations to exhibit device parameter behavior AFTER you have created the model in order to determine what done means. Thanks for the peak behind the curtain though.
There is VDMOStool (and the diode tool from the same author) which will generate MOSFET models based on some datapoints (there are other tools and IIRC Pspice had some built in modeling utilities). This will be a basic model but you can start with it and tweak based on what guides you can find. One thing I have done is decompile datasheet PDFs and use the chart datapoints directly rather than digitizing the plots.
That said, some parameters cannot be set correctly using only what is in the datasheet, and using half measured and half datasheet information to create a model is not advised, but sometimes the only way. The DCA Pro curve tracer can give usable data files for fitting models at low power levels.
You can find more information about VDMOStool from this post
Howto .model without the .model? - QSPICE - Qorvo Tech Forum
Here is a device guideline demonstrate some critical MOSFET parameters and their effect by referring to Qspice help and also Giuseppe and Paolo textbook
Qspice/Guideline/Qspice - Device Reference Guide by KSKelvin.pdf at main · KSKelvin-Github/Qspice
Bob Cordell’s book has a section for transistor modeling that is a fairly approachable way to start: