Better power MOSFET models in LTSpice

Are the equations for the new Ksubthresh model available? I am curious about how the subthreshold behavior was merged with the VDMOS model. A major benefit of Kst is the ease with which it can be added to existing power FET models. The only change Bob Cordell made to his old IRFP240C and IRFP9240C models was the addition of ksubthes=xxx.
 
Are the equations for the new Ksubthresh model available? I am curious about how the subthreshold behavior was merged with the VDMOS model. A major benefit of Kst is the ease with which it can be added to existing power FET models. The only change Bob Cordell made to his old IRFP240C and IRFP9240C models was the addition of ksubthes=xxx.

Hi Ihquam,

Yes, one gets pretty close by just adding ksubthres to a good existing VDMOS model.

These are the ksubthres models that I have been using. ksubthres was added with a value that provided a decent match to my EKV models for the IRFP240 and 9240.

* ksubthres models
*
* IRFP240Ckst VDMOS copyright Cordell Audio December 6, 2010
.model irfp240Ckst VDMOS(nchan Vto=4.0 Kp=4.8 Lambda=0.0032 Rs=0.01 Rd=0.1 Rds=1e7 Cgdmax=2600p Cgdmin=10p a=0.35 Cgs=1250p Cjo=3000p m=0.75 VJ=2.5 IS=4.0E-06 N=2.4 ksubthres=190m)
*
*
*
* IRFP9240Ckst VDMOS copyright Cordell Audio December 6, 2010
.model irfp9240Ckst VDMOS(pchan Vto=-3.76 Kp=9 Lambda=0.004 Rs=0.064 Rd=0.1 Rds=1e7 Cgdmax=1200p Cgdmin=15p a=0.26 Cgs=1130p Cjo=2070p m=0.68 VJ=2.5 IS=4.0E-06 N=2.4 ksubthres=107m)
*


Note that Ian and Keantoken are looking at some modeling approaches that may provide increased accuracy.

Cheers,
Bob
 
I'm playing with a nominally high performance circuit at the moment, using the IRFP pair, and using the ksubthres models I get a difference in simulated performance. The behaviour at crossover has varied, and this translates to a decrease in 2nd and 3rd, and an increase in much higher harmonics in the FFT - about 4dBs worth in the particular instance.
 
I'm playing with a nominally high performance circuit at the moment, using the IRFP pair, and using the ksubthres models I get a difference in simulated performance. The behaviour at crossover has varied, and this translates to a decrease in 2nd and 3rd, and an increase in much higher harmonics in the FFT - about 4dBs worth in the particular instance.

Hi Frank,

What models were you using previously? The conventional VDMOS without ksubthres? Were these my models?

When I changed from conventional VDMOS models to EKV or VDMOSksubthres I did see a slight net increase in THD, but I don't remember a significant increase in high order hermonics.

Cheers,
Bob
 
Bob, yes, I was using your IRFP(9)240C models as is, and I did separate runs without ksubthres, and with, this being the only change between them.

The circuit uses strong local feedback in the output stage, and biasing to prevent the output devices switching off - I used the technique of subtracting the output of the two runs from each other in LTspice to see where the variation occurred, and this was a high frequency wriggle at crossover - 20kHz, near maximum power, into 2 ohms. This is a severe test of the circuit, but it performs well - I suspect the variation will be much less for more moderate situations, something I haven't looked at yet.

Cheers,
 
Bob, yes, I was using your IRFP(9)240C models as is, and I did separate runs without ksubthres, and with, this being the only change between them.

The circuit uses strong local feedback in the output stage, and biasing to prevent the output devices switching off - I used the technique of subtracting the output of the two runs from each other in LTspice to see where the variation occurred, and this was a high frequency wriggle at crossover - 20kHz, near maximum power, into 2 ohms. This is a severe test of the circuit, but it performs well - I suspect the variation will be much less for more moderate situations, something I haven't looked at yet.

Cheers,

Hi fas42,

I have attached my EKV models for the 240 and 9240 in a zipped folder. These models are what I have been using as my baseline reference for evaluating the VDMOS ksubthreshold models. The test benches include Id vs Vgs, log Id vs Vgs, gm vs Id, wingspread, and THD.

These simulations are ready to run.

Wingspread and THD are done for bias currents of 150, 250, 350 and 450mA at 20V peak into 8 ohms. For each value of bias current, output dc offset was adjusted to be less than 1mV.

Source resistors in all cases were 0 ohms.

Cheers,
Bob
 

Attachments

  • EKV 240_9240.zip
    13.3 KB · Views: 402
I tried the SKA amplifier simulation with the EKV models and the crossover looks much more progressive (and realistic) than using the Cordell models. A strange side effect is that when injecting an ac current source into the output, the Cordell models cause a low level oscillation, while this is gone after a few us using the ksubthreshold version
 
I tried the SKA amplifier simulation with the EKV models and the crossover looks much more progressive (and realistic) than using the Cordell models. A strange side effect is that when injecting an ac current source into the output, the Cordell models cause a low level oscillation, while this is gone after a few us using the ksubthreshold version

Hi david,

That's great to hear. I would not have thought that the VDMOS model without ksubthres would have caused an oscillation as long as the idle bias was high enough to provide a reasonable amount of transconductance from one output device or the other.

Cheers,
Bob
 
Hi david,

That's great to hear. I would not have thought that the VDMOS model without ksubthres would have caused an oscillation as long as the idle bias was high enough to provide a reasonable amount of transconductance from one output device or the other.

Cheers,
Bob
The SKA has a very low quiescent for a mosfet design and shows off the abrupt knee of the old simple model.
This plot shows the output device currents at crossover.
When I drive the output with an ac current source the simple model goes unstable
 

Attachments

  • 2015-01-06 22-41-54_LTspice IV - SKA_EKV.png
    2015-01-06 22-41-54_LTspice IV - SKA_EKV.png
    57.4 KB · Views: 920
  • SKA_EKV.asc
    14.1 KB · Views: 171
When I drive the output with an ac current source the simple model goes unstable

When the old model turns off it shoots straight to zero rather than falling naturally into the subthreshold region. This abrupt change creates a spray of harmonics that tends to stimulate any resonances in the circuit. This is always present if you use the old model, even if it is attenuated by high bias, and causes high frequency impulses that can cause strange behavior.

It's easily possible that the model itself isn't causing circuit instability, but that it is simply stimulating a resonance already present in the amplifier. If however you have both MOSFETs biased low enough that both of them fall in the "dead zone" where there should have been subthreshold conduction, the open-loop gain of your amp can fall dramatically and start doing things that are hard to predict. This is a flaw in the old MOSFET model, not in the circuit.
 
When the old model turns off it shoots straight to zero rather than falling naturally into the subthreshold region. This abrupt change creates a spray of harmonics that tends to stimulate any resonances in the circuit. This is always present if you use the old model, even if it is attenuated by high bias, and causes high frequency impulses that can cause strange behavior.

It's easily possible that the model itself isn't causing circuit instability, but that it is simply stimulating a resonance already present in the amplifier. If however you have both MOSFETs biased low enough that both of them fall in the "dead zone" where there should have been subthreshold conduction, the open-loop gain of your amp can fall dramatically and start doing things that are hard to predict. This is a flaw in the old MOSFET model, not in the circuit.
Exactly, class AB amplifiers operate output devices in the region where the simplistic model fails, which is why the improved model is so important for simulating analog use of power mosfets.
 
Exactly, class AB amplifiers operate output devices in the region where the simplistic model fails, which is why the improved model is so important for simulating analog use of power mosfets.

The new models are also important for Class-A amplifiers operating the FETs at currents well below the range the VDMOS models were designed for. The old models are nearly useless.
 
The new models are also important for Class-A amplifiers operating the FETs at currents well below the range the VDMOS models were designed for. The old models are nearly useless.

The VDMOS internal simulator model was updated with subthreshold conduction, which is what this thread is about, so I don't know what you mean here? Many of the old models are VDMOS as well, just missing the new parameters. NMOS/PMOS are the old models which don't model subthreshold or capacitance well.

The EKV models Cordell posted might be better than the VDMOS models as far as harmonic behavior at low frequencies (not sure), but they still don't model capacitance as well as VDMOS.
 
Last edited: