Maker Pro
Maker Pro

Q: cascode FETS for linear control

R

Robert Baer

Jan 1, 1970
0
Just assume two identical FETs in cascode configuration, with equal
value resistors in divider from top FET drain to gate, then to bottom
FET gate which is driven by an adjustable voltage.
Assume this is used as an adjustable shunt regulator in the zero to
15mA drain current region and the open circuit voltage from the supply
is more than twice the avalanche rating of the FETs.
Question: will any step change in the gate drive (bottom FET) cause a
large Vgs change in the top FET - sufficent to zap its gate?
If so, isn't it sufficent to mitigate the problem by using a G-S
capacitor (instead of a zener)?
Does the use of a FQD2N100 give more problems than use of a IRFBG20?
**
Useage: with opamp and load current monitor for forcing a settable
current thru load (the resulting voltage is read using a 10E9 resistor).
Very nice to measure points on high voltage zeners.
 
W

Winfield Hill

Jan 1, 1970
0
Robert Baer wrote...
Just assume two identical FETs in cascode configuration, with equal
value resistors in divider from top FET drain to gate, then to bottom
FET gate which is driven by an adjustable voltage.

Connecting the bottom resistor to the bottom FET's source is better.
Assume this is used as an adjustable shunt regulator in the zero to
15mA drain current region and the open circuit voltage from the supply
is more than twice the avalanche rating of the FETs.
Question: will any step change in the gate drive (bottom FET) cause
a large Vgs change in the top FET - sufficent to zap its gate?

Answer: No.
If so, isn't it sufficent to mitigate the problem by using a G-S
capacitor (instead of a zener)?

The reason it's safe to do this is because a high-voltage MOSFET's
gate capacitance is very high, compared to its other capacitances.
Possibly a gate-source zener isn't necessary,** but it gives you a
bullet-proofing that's cheap at the price and doesn't compromise
any function. Yes, zener capacitance is high, but not compared to
power MOSFET gates. I also protect the bottom-FET's gate.

The main thing to think about will be the value of the gate-divider
resistors. Too small and their current and power dissipation is a
big pain. Too large and they'll fail to drag the top FET's voltage
quickly to the correct place during a step-current change. I'm using
6M resistors (three 2M in series) in my 2.5kV high-voltage amplifier,
and with 2kV 1.3ms ramps (this is the fastest my low-power amplifier
will slew) the MOSFET gates track perfectly without pause.

The most difficult task for my amplifier is surviving an instantaneous
output short. I have 6.6k of series output resistance, which helps,
but a forced short still has to be dealt with by the series FETs in a
microsecond timeframe. Given my MOSFET's huge gate capacitance (400pF)
I was concerned if my 6M gate resistors would do the job (note the long
time constant, 6M*400pF = 2400us). But happily they do work fine, in
large part because a +/- full-scale to ground short is only half of the
amplifier's range.

** About my six FET gate voltages: during the short event, the FET's
Ciss rules, so that no gate has more than a 4.5V excursion, to -0.5V
in that case. This was measured in spice tests using my bench-tested
accurate sub-threshold FET model, without any zener diodes attached).
So even in this most difficult case, the zener diodes aren't used.

If I was using higher supply voltages with more than 3 FETs in series,
one FET would surely avalanche. But this would be at low currents and
last only a few us, so the FET would scarcely notice! However, with an
avalanche event underway, I'd prefer to see the gate zeners in place!
Does the use of a FQD2N100 give more problems than use of a IRFBG20?
**
Useage: with opamp and load current monitor for forcing a settable
current thru load (the resulting voltage is read using a 10E9 resistor).
Very nice to measure points on high voltage zeners.

Both are fine 1kV transistors and I have lots of experience with both.
In my experience their breakdown voltages have never been below 1050
to 1100V, and they haven't shown a strongly-increasing leakage below
that, as some other HV FETs I've used have shown. Recommended.
 
R

Robert Baer

Jan 1, 1970
0
Winfield said:
Robert Baer wrote...



Connecting the bottom resistor to the bottom FET's source is better.
That would be "ground".
Why would thiss be better?
Answer: No.




The reason it's safe to do this is because a high-voltage MOSFET's
gate capacitance is very high, compared to its other capacitances.
Possibly a gate-source zener isn't necessary,** but it gives you a
bullet-proofing that's cheap at the price and doesn't compromise
any function. Yes, zener capacitance is high, but not compared to
power MOSFET gates. I also protect the bottom-FET's gate.

The main thing to think about will be the value of the gate-divider
resistors. Too small and their current and power dissipation is a
big pain. Too large and they'll fail to drag the top FET's voltage
quickly to the correct place during a step-current change. I'm using
6M resistors (three 2M in series) in my 2.5kV high-voltage amplifier,
and with 2kV 1.3ms ramps (this is the fastest my low-power amplifier
will slew) the MOSFET gates track perfectly without pause.

The most difficult task for my amplifier is surviving an instantaneous
output short. I have 6.6k of series output resistance, which helps,
but a forced short still has to be dealt with by the series FETs in a
microsecond timeframe. Given my MOSFET's huge gate capacitance (400pF)
I was concerned if my 6M gate resistors would do the job (note the long
time constant, 6M*400pF = 2400us). But happily they do work fine, in
large part because a +/- full-scale to ground short is only half of the
amplifier's range.

** About my six FET gate voltages: during the short event, the FET's
Ciss rules, so that no gate has more than a 4.5V excursion, to -0.5V
in that case. This was measured in spice tests using my bench-tested
accurate sub-threshold FET model, without any zener diodes attached).
So even in this most difficult case, the zener diodes aren't used.

If I was using higher supply voltages with more than 3 FETs in series,
one FET would surely avalanche. But this would be at low currents and
last only a few us, so the FET would scarcely notice! However, with an
avalanche event underway, I'd prefer to see the gate zeners in place!




Both are fine 1kV transistors and I have lots of experience with both.
In my experience their breakdown voltages have never been below 1050
to 1100V, and they haven't shown a strongly-increasing leakage below
that, as some other HV FETs I've used have shown. Recommended.
Thanks.
Why i mentioned using a capacitor instead of a (back-to-back) zener,
is the fact that a zener would be very leaky at 185C, and the resistor
current is 15uA for a different application.
 
W

Winfield Hill

Jan 1, 1970
0
Robert Baer wrote...
Why i mentioned using a capacitor instead of a (back-to-back) zener,
is the fact that a zener would be very leaky at 185C, and the resistor
current is 15uA for a different application.

Well, as I said, MOSFETs have huge built-in gate capacitors, which
don't need to be supplemented with more! My calculations and spice
measurements show that Ciss dominates and prevents the gates from
seeing dangerous voltages, without further protection, provided that
certain conditions are met, such as adding drain and output resistors
to limit fault currents, and avoiding the temptation to add parallel
capacitors across the gate-resistor stack. I'm assuming low-speed
modest-current applications.

I haven't found high-voltage zeners (10V and up) to be leaky, even
at the low currents I've been working at. OTOH, without any current
going through them, they don't heat up at all, so my measurements
(in the sub-nA region) are at ambient temp.

Still... Adding gate-source zener diodes (not back-to-back, BTW) is
comfort food for the designer, and helps him sleep at night. So I
continue to recommend it on that basis. It'll also help you get
through any design reviews.

Ahem. What's this about a 185C temperature? How's that happen?
You aren't designing down-hole instrumentation, are you?
 
R

Robert Baer

Jan 1, 1970
0
Winfield said:
Robert Baer wrote...



Well, as I said, MOSFETs have huge built-in gate capacitors, which
don't need to be supplemented with more! My calculations and spice
measurements show that Ciss dominates and prevents the gates from
seeing dangerous voltages, without further protection, provided that
certain conditions are met, such as adding drain and output resistors
to limit fault currents, and avoiding the temptation to add parallel
capacitors across the gate-resistor stack. I'm assuming low-speed
modest-current applications.

I haven't found high-voltage zeners (10V and up) to be leaky, even
at the low currents I've been working at. OTOH, without any current
going through them, they don't heat up at all, so my measurements
(in the sub-nA region) are at ambient temp.

Still... Adding gate-source zener diodes (not back-to-back, BTW) is
comfort food for the designer, and helps him sleep at night. So I
continue to recommend it on that basis. It'll also help you get
through any design reviews.

Ahem. What's this about a 185C temperature? How's that happen?
You aren't designing down-hole instrumentation, are you?
Ayup! Downhole tiz.
Measured 1nA leakage at 9V on a 10V back-to-back zener.
At 185C, that would appear to translate to too many microamps (at 9V).
The leakage appeared to be resistive, as compared to semi-constant
current above some small voltage.
 
W

Winfield Hill

Jan 1, 1970
0
Robert Baer wrote...
Ayup! Downhole tiz.

OK, Robert, spill the beans. More detail, please.
Measured 1nA leakage at 9V on a 10V back-to-back zener.
At 185C, that would appear to translate to too many microamps (at 9V).
The leakage appeared to be resistive, as compared to semi-constant
current above some small voltage.

High-voltage MOSFET drain-source leakage current also looks like
a resistor, about 1.5G in the case of my FQD2n100 1kV FETs.
 
J

Jim Thompson

Jan 1, 1970
0
Winfield Hill wrote: [snip]
Ahem. What's this about a 185C temperature? How's that happen?
You aren't designing down-hole instrumentation, are you?
Ayup! Downhole tiz.
Measured 1nA leakage at 9V on a 10V back-to-back zener.
At 185C, that would appear to translate to too many microamps (at 9V).
The leakage appeared to be resistive, as compared to semi-constant
current above some small voltage.

In the very early '70's Schlumberger was one of my best customers.

I was building them a hybrid in a gold-plated metal dual-in-line with
double-stacked-alumina boards populated with a variety of transistors
and OpAmps for, IIRC, measurement of alpha radiation down-hole.

Schlumberger was so concerned about failures that they would throw
away a unit after ONE USE down hole.

So I asked for return of the units for examination of possible failure
mechanisms.

Not one single device had failed. The only observable deficiency was
discolored gold plating.

But I let them keep throwing them away... better business than toilet
paper ;-)

...Jim Thompson
 
R

Robert Baer

Jan 1, 1970
0
Winfield said:
Robert Baer wrote...



OK, Robert, spill the beans. More detail, please.




High-voltage MOSFET drain-source leakage current also looks like
a resistor, about 1.5G in the case of my FQD2n100 1kV FETs.
I already have a working design for a shunt regulator that replaces
the Victoreen Corotron.
Presently, i limit each one to a max of 900V (using a 1000V FET) and
stack them in series to get the required PMT regulation at a shunt
current of 600-100uA (industry standard).
Was thinking of a way to reduce parts and size via the cascode scheme.
**
At what current levels did you see that resistive component?
 
W

Winfield Hill

Jan 1, 1970
0
Robert Baer wrote...
I already have a working design for a shunt regulator that
replaces the Victoreen Corotron.

Or the AnaLog Services' Codatron? Ahem, have you seen their patent?
Presently, i limit each one to a max of 900V (using a 1000V FET)
and stack them in series to get the required PMT regulation at a
shunt current of 600-100uA (industry standard).
Was thinking of a way to reduce parts and size via the cascode
scheme.
At what current levels did you see that resistive component?

The MOSFET leakage resistance is modeled in parallel with the
textbook FET channel, hence its value has no relationship to
the operating current. As I mentioned, I measured 1.5G for the
FQD2n100, but I've also measured the IRFBG20, getting about 7G.
It would seem one can safely ignore it at room temp. As for high
temperatures, 150C or 185C, you'd better take some measurements!
 
R

Robert Baer

Jan 1, 1970
0
Winfield said:
Robert Baer wrote...



Or the AnaLog Services' Codatron? Ahem, have you seen their patent?




The MOSFET leakage resistance is modeled in parallel with the
textbook FET channel, hence its value has no relationship to
the operating current. As I mentioned, I measured 1.5G for the
FQD2n100, but I've also measured the IRFBG20, getting about 7G.
It would seem one can safely ignore it at room temp. As for high
temperatures, 150C or 185C, you'd better take some measurements!
I might have seen the patent; i wrote it.
I guess that you got the pun on the Titan regulator (our "TitanTwo"
shunt regulator).
Put the patent write-up on the TitanTwo into the public domain -
where it belongs.
 
W

Winfield Hill

Jan 1, 1970
0
Robert Baer wrote...
I might have seen the patent; i wrote it.
I guess that you got the pun on the Titan regulator
(our "TitanTwo" shunt regulator).

Cute. What's the patent number?
Put the patent write-up on the TitanTwo into the
public domain - where it belongs.

Good.
 
R

Robert Baer

Jan 1, 1970
0
Winfield said:
Robert Baer wrote...



Cute. What's the patent number?
In process aka pending.
I might move the Codatron to PD as well.
What do you think?
Have you torn apart Titan's "regulator"?
I have "regulation" voltage VS input voltage step and temperature step (wide
pulse in both cases); the output is worse than the input and there is severe
temperature hystersis.
That is why i made a better design; do not need to "tweak" a divider on FETs
either.
Makes the regulators very repeatable and allows automated assembly.
 
R

Robert Baer

Jan 1, 1970
0
Robert said:
In process aka pending.
I might move the Codatron to PD as well.
What do you think?

Have you torn apart Titan's "regulator"?
I have "regulation" voltage VS input voltage step and temperature step
(wide pulse in both cases); the output is worse than the input and there
is severe temperature hystersis.
That is why i made a better design; do not need to "tweak" a divider
on FETs either.
Makes the regulators very repeatable and allows automated assembly.
Look at the Oil 4 Less LLC site; the patent apps are now officially
in the public domain, and i refer to the Titan regulator patent as well.
 
W

Winfield Hill

Jan 1, 1970
0
Robert Baer wrote...
Look at the Oil 4 Less LLC site; the patent apps are now officially
in the public domain, and i refer to the Titan regulator patent as
well.

How are your corrected HV shunt regulators used, what are they for?
 
Top