List Of Simulation Model Errors

shadders

Member
2010-02-13 2:16 pm
London
Hi,
After some issues with a simulation model being in error with a power transistor, and latterly another medium power transistor model also in error – I have started this thread to warn people of the model errors.

On Semiconductor MJL4302 both spice3 models have VTF=10000 which is in error. I wrote to On Semi in July 2019, they stated that they will correct, and it is still not corrected.

On Semiconductor BD140 spice3 model has BF=650.842. This is unrealistic, and their data sheet states 250 maximum. The On Semi BD139 BF seems to be appropriate. I used the ST Microelectronics BD140 spice model which seems to present much closer in circuit measured values.

It would be helpful if others added to this, so we have a list of models and their errors.

Regards,
Shadders.
 

shadders

Member
2010-02-13 2:16 pm
London
BF=650.842 is the ideal coefficient. There are several other spice-model parameters that reduce this ideal coefficient to a real BETA value. That is, BF is not a BETA! And you shouldn't be outraged. Make the BETA gain depend on the collector (emitter) current for this model.
Hi,
I compared it to the ST Microelectronics model - BD139 and BD140 - they have similar BF.

Are you stating that the BF=650 is not an issue ?
Thanks and regards,
Shadders.
 

Elvee

Member
2006-09-08 2:04 pm
Models do have errors and inaccuracies; some even manage to violate thermodynamic laws:
Overunity transistors

However, as has been pointed out, many parameters are modulated by the operating conditions and other parameters.

In the example given here: A constant-Ic β-tester: β-Master 🌈
the BF of the 2N3055 is stepped: 100, 200, 400, 600, 800, but the "measured" result is ~10% lower for these operating conditions as you can see on the graph.

That said, a BF of 650 must stretch the various factors to their limit, if the actual beta is the region of 100
 

Mark Johnson

Member
Paid Member
2011-05-27 3:27 pm
Silicon Valley
Bob Cordell's book has a very nice section that teaches you how to build your own SPICE model parameter sets from data that you measure yourself. And the Peak Atlas DCA Pro is the most full-featured curve tracer on the market, in the under-$150 price range.

Creating your own model parameter sets is a lot of fun, and doing it will entice you to dive deep into the statistics of Linear Regression, which is a great thing. Because once you understand the mathematics of Linear Regression, you'll understand how to USE it, as a multi-featured package of calculations that is already built-in to every copy of Microsoft Excel. That's right: SPICE model parameter fitting inside Excel. Cowabunga.

_
 
Last edited:

shadders

Member
2010-02-13 2:16 pm
London
Hi All,
Thanks for the replies. I ran the Ic/Ib for both and one B was 160 (on semi) and the other was 220 (ST). I was not outraged - but concerned.

I have the DCA Pro - and have tested all of my power transistors, but not the BD140/139.

It was because the simulated results for the Vbe multiplier were so different from circuit, that it caused me to examine the models.

I won't go as far as examining the models in Excel ( :) ), but i did have a quick look at the model used by spice and the factors surrounding beta use.

Thanks and regards,
Shadders.
 
Last edited:

shadders

Member
2010-02-13 2:16 pm
London
Vbe Multiplier Modelled

Hi,
I implemented the Vbe multiplier using the same circuit, where one uses the On Semi model, and the other uses the ST Microelectronic model, both for the BD139 and BD140.

The results are not as expected. The difference in the Vbe curves for a change in resistance for setting the Vbe is much greater than expected.

I attach my simulation file and out for the two Vbe's. Have i made an obvious error ?
Thanks and regards,
Shadders.
 

Attachments

  • 2020 02 09 Vbe Multiplier No Library.asc
    7.6 KB · Views: 16
  • 2020 02 09 Vbe Multiplier Modelled.png
    2020 02 09 Vbe Multiplier Modelled.png
    21.4 KB · Views: 86