FET models

[Robert Baer]

I would like to have a way of altering a FET model so that it follows
the Vgs VS log(Is) instead of dropping rapidly near 1mA like:
/
/
/ log(Is)
/
/
|
|
|
------------------
Vgs

Any ideas?

Also, the model for the Fairchild FQD2N100 totally sucks.
Is there a way to make it work?

..SUBCKT FQD2N100 d g s
Rg g 1 0.04
M1 2 1 3 3 DMOS L=1u W=1u
..MODEL DMOS NMOS(VTO=4.66 KP=1.9 LEVEL=3)
Cgs 1 3 380p
Rd d 4 3.5
Dds 3 4 DDS
..MODEL DDS D(BV=1050 M=0.42 CJO=35p VJ=0.12)
Dbody 3 d DBODY
..MODEL DBODY D(IS=2.8E-13 N=1.00 RS=0.005 EG=1.10 TT=520n)
Ra 4 2 3.5
Rs 3 5 0.024
Ls 5 s 2.6n
M2 1 8 6 6 INTER
E2 8 6 4 1 2
..MODEL INTER NMOS(VTO=0 KP=10 LEVEL=1)
CGDMAX 7 4 380p
RCGD 7 4 1E7
DGD 6 4 DGD
RDGD 4 6 1E7
..MODEL DGD D(M=0.52 CJO=380p VJ=0.12)
M3 7 9 1 1 INTER
E3 9 1 4 1 -2
..ENDS

 

[colin]

I would like to have a way of altering a FET model so that it follows
the Vgs VS log(Is) instead of dropping rapidly near 1mA like:
/
/
/ log(Is)
/
/
|
|
|
------------------
Vgs

Any ideas?

Also, the model for the Fairchild FQD2N100 totally sucks.
Is there a way to make it work?

theres been an improvement discused in sci.electronics.design for sub
threshold operation of FET models.

Colin =^.^=

 

[Robert Baer]

theres been an improvement discused in sci.electronics.design for sub
threshold operation of FET models.

Colin =^.^=

I remember Winfield Hill's note of 6/24 concerning altering the model
for the Fairchild FQD2N100 1KV FET.
BUT the original model does not work properly, and i am a bit
unfamiliar with SPICE and the mod that was made - so it is not possible
to evaluate the result (there was no observable difference).
Do you have a decently working model set (original & modified)?

 

[colin]

I remember Winfield Hill's note of 6/24 concerning altering the model
for the Fairchild FQD2N100 1KV FET.
BUT the original model does not work properly, and i am a bit
unfamiliar with SPICE and the mod that was made - so it is not possible
to evaluate the result (there was no observable difference).
Do you have a decently working model set (original & modified)?

yes thats the one i was thinking of, there were several threads, one gave
modified/unmodified models although i cldnt put my finger on it straight
away, one also mentioned a method of combining the log characteristic of a
diode with the input. Im afraid i didnt have any more to offer than pointing
you there. good luck.

Colin =^.^=

 

[Jim Thompson]

I would like to have a way of altering a FET model so that it follows
the Vgs VS log(Is) instead of dropping rapidly near 1mA like:
/
/
/ log(Is)
/
/
|
|
|

[snip]

Do you have any accurate data for this region?

I'd like to try modeling it with a higher level model, rather than
patching onto a Level=1 or 3 model.

...Jim Thompson

 

[Robert Baer]

I would like to have a way of altering a FET model so that it follows
the Vgs VS log(Is) instead of dropping rapidly near 1mA like:
/
/
/ log(Is)
/
/
|
|
|

[snip]

Do you have any accurate data for this region?

I'd like to try modeling it with a higher level model, rather than
patching onto a Level=1 or 3 model.

...Jim Thompson

Most non-logic power FETs seem to follow (Vgs1-Vgs2) =
0.25*log10(Id1/Id2); logic FETs have a different slope.
What really bugs me, is if i "slide" the straight line (to the right)
along a log (vert) / linear (H) scale to cover 6-7 decades instead of
the original 2 decades, the calculated factor changes from about 0.25 to
about 0.17 and i cannot figure out what is wrong.

 

[Jim Thompson]

I would like to have a way of altering a FET model so that it follows
the Vgs VS log(Is) instead of dropping rapidly near 1mA like:
/
/
/ log(Is)
/
/
|
|
|

[snip]

See attachment at...

Newsgroups: alt.binaries.schematics.electronic
Subject: Power MOS, Log(ID) versus Gate-Source Voltage -
LogIDvsVGS.pdf
Message-ID: <[email protected]>

All the discussions here, S.E.D and in A.B.S.E about cobbling up a
"patched" power MOS Spice model to cover low-voltage-low-current
characteristics have always struck me as dangerous, fraught with
possibility for error in other parameters.

The attachment shows how Level=7 seems to naturally cover the region.

So I think two things are in order...

(1) Harass Power MOS manufacturers to provide a correct Spice model

or

(2) Roll you own, but use Level=7 parameters rather than patching a
lower level model.


...Jim Thompson

 

[Mike Engelhardt]

Jim,

All the discussions here, S.E.D and in A.B.S.E about
cobbling up a "patched" power MOS Spice model to cover
low-voltage-low-current characteristics have always
struck me as dangerous, fraught with possibility for
error in other parameters.

The attachment shows how Level=7 seems to naturally
cover the region.

So I think two things are in order...

(1) Harass Power MOS manufacturers to provide a
correct Spice model

or

(2) Roll you own, but use Level=7 parameters rather
than patching a lower level model.

Level=7 will handle sub-threshold, but for many applications
there's an even greater discrepancy between the model and
the part -- that being the way it stores charge, particularly
the Miller capacitance. The only way I've seen that handled
properly in a native SPICE device is in LTspice's VDMOS device,
though the DC curves of that is just a Level 1 MOS with no
sub-threshold modeled.

The deal with the charge is that the vertical double-diffused
MOSFET qualitatively behavior entirely differently than a
monolithic device, particularly w.r.t. the Gate-Drain
capacitance.

--Mike

 

[Jim Thompson]

Jim,


Level=7 will handle sub-threshold, but for many applications
there's an even greater discrepancy between the model and
the part -- that being the way it stores charge, particularly
the Miller capacitance. The only way I've seen that handled
properly in a native SPICE device is in LTspice's VDMOS device,
though the DC curves of that is just a Level 1 MOS with no
sub-threshold modeled.

The deal with the charge is that the vertical double-diffused
MOSFET qualitatively behavior entirely differently than a
monolithic device, particularly w.r.t. the Gate-Drain
capacitance.

--Mike

Just amusing myself, reading up on the subject, the EKV model may be
best.

...Jim Thompson

 

[Jim Thompson]

Jim,


Level=7 will handle sub-threshold, but for many applications
there's an even greater discrepancy between the model and
the part -- that being the way it stores charge, particularly
the Miller capacitance. The only way I've seen that handled
properly in a native SPICE device is in LTspice's VDMOS device,
though the DC curves of that is just a Level 1 MOS with no
sub-threshold modeled.

The deal with the charge is that the vertical double-diffused
MOSFET qualitatively behavior entirely differently than a
monolithic device, particularly w.r.t. the Gate-Drain
capacitance.

--Mike

Mike, Is that an LTspice-specific model or is it a subcircuit?

Does it solve Win's low-current fretting issues?

...Jim Thompson

 

[Mike Engelhardt]

Jim,

Just amusing myself, reading up on the subject, the EKV model
may be best.

EKV's forte is really about D.C.(and some think that it a break-
through for a compact description of the monolithic FET). But I
haven't seen the EKV charge model solve the Gate-Drain capacitance
modeling problem of a VDMOS power transistor. The VDMOS transistor
typically used as descrete power transistors just fundamentally
stores charge differently than the monolithic MOSFET.

--Mike

 

[Mike Engelhardt]

Jim,

Mike, Is that an LTspice-specific model or is it a subcircuit?

It's a LTspice specific device. An intrinsic new SPICE device,
not a subcircuit and without internal nodes. It's documented
toward the end of the section of the help LTspice=>Circuit
Elements=>M. MOSFET

Does it solve Win's low-current fretting issues?

No, not at all. It has no sub-threshold. It just gets the
charge storage right. It's forte is that it switches correctly
because it has the Miller capacitance correct. It does not
handle the Win's problem with sub-threshold. That's a real
problem, too. If you wanted to fix both, an approach would
be to use a subcircuit with a Level 7 for DC(with no charge
storage) in parallel with an LTspice VDMOS with no conduction,
but supplying the charge storage. I haven't added the power
MOSFET's sub-threshold to LTspice's VDMOS device because it
only comes up, well, in sub-threshold, not when you're actually
turning the thing on and off or even in proper linear range.
Besides, people can cobble some sub-threshold behavior in when
they need it.

--Mike

 

[Kevin Aylward]

Jim,


EKV's forte is really about D.C.(and some think that it a break-
through for a compact description of the monolithic FET). But I
haven't seen the EKV charge model solve the Gate-Drain capacitance
modeling problem of a VDMOS power transistor. The VDMOS transistor
typically used as descrete power transistors just fundamentally
stores charge differently than the monolithic MOSFET.

Oh...you mean it uses an inductor?

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Mike Engelhardt]

Kevin,

Oh...you mean it uses an inductor?

Inductors don't store charge, they store flux.
The VDMOS Miller capacitance behaves fundamentally
differently than that of a monolithic device. See
LTspice help documentation for the VDMOS charge model.

--Mike

 

[Kevin Aylward]

Kevin,


Inductors don't store charge, they store flux.

Ahmmm...indeed they do. According to relativistic electromagnetics, all
magnetic and electric phenomena are one and the same, but observed from
different reference frames.

What we identify as magnetic or electric is simply one of convieniance,
not physical reality.

Hint: why should a magnetic field suddenly appear just because an
*observer* moves past a charge?

Hint: just what do you propose "magnetic flux" *actually* is?
Hint: just what do you propose "electric flux" *actually* is?
Hint: photon exchange.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Mike Engelhardt]

Kevin,

Ahmmm...indeed they do. According to relativistic electromagnetics,
all magnetic and electric phenomena are one and the same, but
observed from different reference frames.

Completely irrelevant for lumped-constant reactances and
doesn't change the fact that a power MOSFET's Gate-Drain
capacitance is fundamentally different then that of a
monolithic MOSFET, no matter what reference frame is used.

Your penance shall be to hand write "Inductors don't store
charge, they store flux." 100 times and tape it to your monitor
for 2 days.

--Mike

 

[Jim Thompson]

Kevin,


Completely irrelevant for lumped-constant reactances and
doesn't change the fact that a power MOSFET's Gate-Drain
capacitance is fundamentally different then that of a
monolithic MOSFET, no matter what reference frame is used.

Your penance shall be to hand write "Inductors don't store
charge, they store flux." 100 times and tape it to your monitor
for 2 days.

--Mike

Sno-o-o-o-ort ;-)

...Jim Thompson

 

[Kevin Aylward]

Kevin,


Completely irrelevant for lumped-constant reactances and
doesn't change the fact that a power MOSFET's Gate-Drain
capacitance is fundamentally different then that of a
monolithic MOSFET, no matter what reference frame is used.

ROTHFLMAO.

You certainly walked right into that one with your eyes closed Micky
laddie.

So, fundamentally different, so, you cliam that gate capacitance is due
to moon cheese then?

Your penance shall be to hand write "Inductors don't store
charge, they store flux." 100 times and tape it to your monitor
for 2 days.

As Inspector Clouseau would say, "you fooool"

First, it was obvious to any with any sense that I was trolling.

Second, regarding your last comment, as a point of fact, you should
indeed stick to what you know, like spice matrices.

I will tell you once more. *All* electromagnetic phenomena is due to
charge. Period. End of story. All electromagnetic energy storage is by
way of the effective charge position and charge motion. Charge is a
number that identifies aspects of the momentum flux of photons.
"Magnetic stuff" does not exist. Its all photons mate. Flux is no more
than a distribution of photons... oh dear...never mind...stick to
software...

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Mike Engelhardt]

Kevin,

So, fundamentally different, so, you cliam that gate capacitance
is due to moon cheese then?

See the reference(LTspice help pages). It's because of the drain
being on the back of the die. The functional dependence of the
gate-drain charge on the terminal voltages is fundamentally
different then that of monolithic devices. As this is the
Miller capacitance, it fundamentally changes the the switching
behavior of the grounding source device. Fundamental in that
the behavior is qualitatively distinct. It's why MOSFET
vendors usually have to use a subcircuit to model their
MOSFET's, but LTspice has an intrinsic device with an appropriate
charge model.

I will tell you once more. *All* electromagnetic phenomena
is due to charge. Period. End of story.

I didn't said it was or wasn't. But you're wrong, you need
space and time too, though. Charge by itself won't cut it.
And moving charge has behavior to be reckoned with. And the
laws of physics aren't invariant in non-inertial frames.
There's no inertial frame which converts all of the general
inductor's B-field to an E-field. The formulation of lumped
constant reactances has no concept of space, you just have to
trade things like E-field times distance with electromotive
force. Inductance and flux are as real as they are useful,
no matter how much you would like to dismiss it. Of course
the joke is that I'm the one here that's the physicist in
real life and you're not.

You were supposed to put your penance work on the FRONT of your
monitor for two days!

--Mike

"Capacitance stores charge, inductance stores flux."
Mike Engelhardt 2005

 

[Evgenii Rudnyi]

Hello,

I have a small comment to the discussion. It is not too helpful at the
moment, but it shows some perspective that may become real in the
future.

The problem to make a compact transistor model happens to be close to
the goal of Model Order Reduction. We have contrasted compact modeling
with model order reduction in

MST MEMS model order reduction: Requirements and Benchmarks
http://dx.doi.org/10.1016/j.laa.2005.04.002

preprint is at

http://www.imtek.uni-freiburg.de/simulation/mor4ansys/pdf/lienemann03LLA.pdf

Well, at present model reduction cannot be used in the case of
transistor, yet it is already working extremely well for linear
systems. You can try it for ANSYS models with our free software

http://www.imtek.uni-freiburg.de/simulation/mor4ansys/

I hope that one day mathematicians will extend it to nonlinear problems
as well.

Best wishes,

Evgenii Rudnyi