Maker Pro
Maker Pro

High side FET drive question

C

Chris Carlen

Jan 1, 1970
0
Greetings:

I am considering to build an H-bridge around a solenoid coil with the
capability of delivering 150V at 30A for up to 500us pulses, followed by
about 40V at 20A (PWMed) for up to about 4.5 ms, with up to about 5%
duty cycle of solenoid pulses (58.5W average power, 4500W peak power).
The H-bridge would be part of a PWM constant current drive, controlling
solenoid current to be proportional to some command voltage. This is a
deisel fuel injector driver. I want to achieve about a 500us current
rise/fall time from 0-30A into a 2ohm 2mH load.

I am looking at a employing four of IRFP260N 200V 50A 0.04ohm 300W
devices. They seem beefy enough to not need to parallel devices.

The question is how to accomplish the high side drive with a 150V B+
rail? I am considering using a DC-DC converter to provide isolated 12V,
and powering a FET driver such as TC4426 with that. The input of the
FET driver would be driven by the output of a fast optocoupler powered
by the DC-DC converter as well. The input to the optocoupler is then
accessible at the ground level, and can turn my FET on and off. The
problem with this is that a bunch of circuitry (the DC-DC output
circuitry, and everything to the right of the optocoupler LED) must
swing with the output voltage. I guess there is nothing inherently
wrong with this, but it must radiate a lot of EMI as well as presenting
capacitance to the output.

Is this the right approach, or is there a better way?

I think it is impossible to use a pulse transformer in this situation,
because there are times when the solenoid must be switched on for a
solid 500us. But I haven't evaluated this option in depth yet, as I
have only recently settled upon the H-bridge approach after realizing
the factors that make a 1/2 bridge impractical, as well as a zener catch
diode approach.

But my intuition as well as a few cursory calculations suggest that any
pulse transformer capable of holding the FET on for 500us would have too
high of an inductance, and thus too high of a leakage inductance, to be
able to turn the FET on and off quickly enough.


Comments appreciated.


Good day!




--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
C

Chris Carlen

Jan 1, 1970
0
Chris said:
Greetings:

I am considering to build an H-bridge around a solenoid coil with the
capability of delivering 150V at 30A for up to 500us pulses, followed by
about 40V at 20A (PWMed) for up to about 4.5 ms, with up to about 5%
duty cycle of solenoid pulses (58.5W average power, 4500W peak power).
The H-bridge would be part of a PWM constant current drive, controlling
solenoid current to be proportional to some command voltage. This is a
deisel fuel injector driver. I want to achieve about a 500us current
rise/fall time from 0-30A into a 2ohm 2mH load.

I am looking at a employing four of IRFP260N 200V 50A 0.04ohm 300W
devices. They seem beefy enough to not need to parallel devices.

Whoops, I think that since the coil current is always flowing in the
same direction, I really need only two NMOS devices, in the upper left
and lower right corners of the H-bridge, with a freewheeling diode in
each of the lower left and upper right corners.

Hmm, that makes things a little easier! But I still need one high side
driver circuit.


Good day!



--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
W

Winfield Hill

Jan 1, 1970
0
Chris Carlen wrote...
But my intuition as well as a few cursory calculations suggest that any
pulse transformer capable of holding the FET on for 500us would have too
high of an inductance, and thus too high of a leakage inductance, to be
able to turn the FET on and off quickly enough.

Listen up, grasshopper, and learn. Your goal is to turn on the FET,
which requires speed and power and then leave it on, followed later
by turning it off, which also requires power. Here's one easy way
to achieve these tasks without requiring a low-frequency transformer.
In fact, the technique requires a high-frequency pulse transformer.

.. D1
.. ,---+--|>|---+-- gate
.. | | |
.. |( Rx Q1 |/ source
.. |( | ,--| |
.. | | | |\V |
.. | | | | |
.. '-------+----------'
.. |________|

Positive transitions turn on the FET, afterwhich it remains on due to
its gate capacitance. The negative transition at the end of the 500us
pulse turns off the FET by means of Rx and the NPN. Actually, I use
a 4-layer construction for Q1 to insure rapid and complete turnoff.

Thanks,
- Win
 
J

Jim Thompson

Jan 1, 1970
0
Chris Carlen wrote...

Listen up, grasshopper, and learn. Your goal is to turn on the FET,
which requires speed and power and then leave it on, followed later
by turning it off, which also requires power. Here's one easy way
to achieve these tasks without requiring a low-frequency transformer.
In fact, the technique requires a high-frequency pulse transformer.

. D1
. ,---+--|>|---+-- gate
. | | |
. |( Rx Q1 |/ source
. |( | ,--| |
. | | | |\V |
. | | | | |
. '-------+----------'
. |________|

Positive transitions turn on the FET, afterwhich it remains on due to
its gate capacitance. The negative transition at the end of the 500us
pulse turns off the FET by means of Rx and the NPN. Actually, I use
a 4-layer construction for Q1 to insure rapid and complete turnoff.

Thanks,
- Win

Very similar to what I used for a satellite spinner/launcher about 15
years ago, except I used a PNP in place of D1 and I had a zener in
there to set VGmax.

IIRC I also put it in the Hubble Telescope controls.

...Jim Thompson
 
W

Winfield Hill

Jan 1, 1970
0
Jim Thompson wrote...
Very similar to what I used for a satellite spinner/launcher about 15
years ago, except I used a PNP in place of D1 and I had a zener in
there to set VGmax. I also put it in the Hubble Telescope controls.

How long did you rely on the FET's gate voltage remaining charged
for your space app?

Thanks,
- Win
 
J

James Meyer

Jan 1, 1970
0
Chris Carlen wrote:

Whoops, I think that since the coil current is always flowing in the
same direction, I really need only two NMOS devices, in the upper left
and lower right corners of the H-bridge, with a freewheeling diode in
each of the lower left and upper right corners.

Hmm, that makes things a little easier! But I still need one high side
driver circuit.
I'm probably missing something vital, but if the current never reverses,
why do you need two switch elements?

Jim
 
W

Winfield Hill

Jan 1, 1970
0
Chris Carlen wrote...
I assume this means that you would favor this method over
what I had suggested, right?

Yes. I've used your method (required for really-long ON time)
but it's not only more complicated, it's much much slower.

Thanks,
- Win
 
C

Chris Carlen

Jan 1, 1970
0
James said:
I'm probably missing something vital, but if the current never reverses,
why do you need two switch elements?

Jim


In order to get the current fall time to be as fast as the rise time, it
is necessary to be able to reverse the polarity of voltage applied to
the coil. I had considered a single switch with zener, but that
requires a physically unrealizable zener which if it has a voltage of
150V and the current is 30A, must dissipate 4500W peak power:

http://home.earthlink.net/~crobc/misc/images/sol-zen.png

Then I considered a 1/2 bridge, but that leads to problems with the DC
nature of the coil current. The midpoint of the caps and resistor
voltage divider can't get back to 1/2 the supply volts until an
inordinate amount of time, unless I am willing to use heating elements
for the resistors. Bad idea:

http://home.earthlink.net/~crobc/misc/images/sol-halfb.png

A full bridge gets me the desired results:

http://home.earthlink.net/~crobc/misc/images/sol-fullb.png


Now if *I'm* missing something vital, please let me know.

Thanks for the response.


Good day!
 
F

Fritz Schlunder

Jan 1, 1970
0
Chris Carlen said:
Greetings:

I am considering to build an H-bridge around a solenoid coil with the
capability of delivering 150V at 30A for up to 500us pulses, followed by
about 40V at 20A (PWMed) for up to about 4.5 ms, with up to about 5%
duty cycle of solenoid pulses (58.5W average power, 4500W peak power).
The H-bridge would be part of a PWM constant current drive, controlling
solenoid current to be proportional to some command voltage. This is a
deisel fuel injector driver. I want to achieve about a 500us current
rise/fall time from 0-30A into a 2ohm 2mH load.

I am looking at a employing four of IRFP260N 200V 50A 0.04ohm 300W
devices. They seem beefy enough to not need to parallel devices.

The question is how to accomplish the high side drive with a 150V B+
rail? I am considering using a DC-DC converter to provide isolated 12V,
and powering a FET driver such as TC4426 with that. The input of the
FET driver would be driven by the output of a fast optocoupler powered
by the DC-DC converter as well. The input to the optocoupler is then
accessible at the ground level, and can turn my FET on and off. The
problem with this is that a bunch of circuitry (the DC-DC output
circuitry, and everything to the right of the optocoupler LED) must
swing with the output voltage. I guess there is nothing inherently
wrong with this, but it must radiate a lot of EMI as well as presenting
capacitance to the output.

Is this the right approach, or is there a better way?

I think it is impossible to use a pulse transformer in this situation,
because there are times when the solenoid must be switched on for a
solid 500us. But I haven't evaluated this option in depth yet, as I
have only recently settled upon the H-bridge approach after realizing
the factors that make a 1/2 bridge impractical, as well as a zener catch
diode approach.

But my intuition as well as a few cursory calculations suggest that any
pulse transformer capable of holding the FET on for 500us would have too
high of an inductance, and thus too high of a leakage inductance, to be
able to turn the FET on and off quickly enough.


Comments appreciated.


Good day!



If you haven't already read Texas Instruments "Design and Application Guide
for High Speed MOSFET Gate Drive Circuits" by Laszlo Balogh I would highly
recommend it:

http://focus.ti.com/lit/misc/slup169/slup169.pdf

It isn't quite comprehensive but it makes an extremely good effort to be
nevertheless. It is certainly the best MOSFET gate drive resource I've ever
seen compiled into a single document.
 
W

Winfield Hill

Jan 1, 1970
0
Chris Carlen wrote...
In order to get the current fall time to be as fast as the rise
time, it is necessary to be able to reverse the polarity of
voltage applied to the coil. I had considered a single switch
with zener, but that requires a physically unrealizable zener
which if it has a voltage of 150V and the current is 30A, must
dissipate 4500W peak power:

http://home.earthlink.net/~crobc/misc/images/sol-zen.png
Now if *I'm* missing something vital, please let me know.

Indeed you are.

The inductance flyback nicely reverses the voltage as far
as you want, there's no need to apply a reversed voltage
source. And small zener diodes can handle surprisingly-
large peak currents, or for truly impressive peak power
capability use the silicon Transient Voltage Suppressor
(TVS) type of zener diodes. These have massive blocks of
copper on both sides of the silicon die to absorb large
amounts of energy.

Calculate the energy in the coil's inductance E = 1/2 LI^2
and compare it to the capability of parts like the 5kp100A
(5kp110C), which can repeatedly handle 5kW peak exponential
pulses lasting 1ms (about 2.5J of energy). These can also
handle 10kW lasting 250us, etc. The breakdown voltage at
30A will be under 162V. Should you have a need to dissipate
more energy, use multiple low-voltage TVS parts in series.

Remember the inductor's discharge will be very fast if you
let the flyback go to 150V. What's the inductance value?

Thanks,
- Win
 
C

Chris Carlen

Jan 1, 1970
0
Fritz said:
If you haven't already read Texas Instruments "Design and Application Guide
for High Speed MOSFET Gate Drive Circuits" by Laszlo Balogh I would highly
recommend it:

http://focus.ti.com/lit/misc/slup169/slup169.pdf

It isn't quite comprehensive but it makes an extremely good effort to be
nevertheless. It is certainly the best MOSFET gate drive resource I've ever
seen compiled into a single document.



Thanks for the link, Fritz. I'll definitely check that out and keep it
in my collection.


Good day!


--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
C

Chris Carlen

Jan 1, 1970
0
Pieter said:
Is this going to be a testing device for diesel injectors? I am
interested in what you are building here because I have had a qusetion
for the same thing.

It's not to test injectors, but to drive them. What is under test is
engine combustion. We have specially designed and modified production
deisel engines with optical access to the combustion chambers. We
perform laser induced incandescence and laser induced fluorescence
imaging spectroscopy to study the processes of pollutant formation, as
well as some investigations into alternative fuels and the effects of
oxygenates on soot formation.

We want to be able to perform up to 4 or 5 injections per engine cycle.
Our existing "ECM" production engine Electronic Control Modules
aren't able to issue more than 1-2 inject pulses per cycle.

You only want to control a one-direction current through the coil?
The could could use ONE transister or fet.

Perhaps I can use one FET with an amplified zener damper diode for the
current ramp down as another respondent indicated.

A 1/2 H-bridge of course gives the coil the possibility to go up to
the (negative) voltage of the bridge at turn-off, increasing speed.

Yes, but it requires two power supplies, no? One sources while the
other sinks. I tried it in simulation, like this:

http://home.earthlink.net/~crobc/misc/images/sol-halfb.png

but realized there are serious problems with keeping the voltage divider
midpoint at V/2 without using very stiff resistors and wasting a lot of
power.

To accelerate thing at turn-on, you could discharge a pre-charged
capacitor over the coil.

That's OK for a single injection. That's what the ECM boxes do.

At turn-off you can dissipate above certain voltages with zeners
(combined with a tor/fet for the power) etc.

Yes. I will work on this approach.


Thanks for the input!


Good day.


--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
C

Chris Carlen

Jan 1, 1970
0
Jim said:
Your "turn off" power supply also has to handle 4500 peak Watts.
How do you propose to keep its output voltage in bounds at 30 Amps?

In the H-bridge, energy is removed from the capacitor when the current
ramps up, and much of that energy is returned to the capacitor when the
current ramps down. The loss of energy would be the integral of
I^2*Rcoil, and the average power over time would be the energy loss per
pulse divided by the repetition period. The power supply needn't sink
any current. The duty cycle is low at about 5-10% so the average power
is low, at about 90W max for the heaviest operating conditions.
A transistor amplified zener could handle the power. Then you
could get back to a single low-side active switch and diode steering
of the currents involved.

Yes, I will look into this. Maybe my first circuit was right after all,
just missing some assistance to the zener.


Good day!



--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
C

Chris Carlen

Jan 1, 1970
0
Winfield said:
Chris Carlen wrote...



Indeed you are.

The inductance flyback nicely reverses the voltage as far
as you want, there's no need to apply a reversed voltage
source. And small zener diodes can handle surprisingly-
large peak currents, or for truly impressive peak power
capability use the silicon Transient Voltage Suppressor
(TVS) type of zener diodes. These have massive blocks of
copper on both sides of the silicon die to absorb large
amounts of energy.

Calculate the energy in the coil's inductance E = 1/2 LI^2
and compare it to the capability of parts like the 5kp100A
(5kp110C), which can repeatedly handle 5kW peak exponential
pulses lasting 1ms (about 2.5J of energy). These can also
handle 10kW lasting 250us, etc. The breakdown voltage at
30A will be under 162V. Should you have a need to dissipate
more energy, use multiple low-voltage TVS parts in series.

Remember the inductor's discharge will be very fast if you
let the flyback go to 150V. What's the inductance value?

Thanks,
- Win


I expect about 2mH plus 2 ohms of coil R, maybe a little less
inductance, so about 0.9J of energy at max current. The problem is that
I have to sustain 20A of coil current for up to 5-10ms perhaps. It's
not just a matter of dumping the coil energy once, then waiting a while
to cool down. Thus, I will be PWMing the supply voltage, with a control
loop implementing a constant current source. But the TVS would then
have to disspate about 1100W for several ms during the 20A level.

What do you think of a transistor assisted zener?

A TVS (if the average power can be worked out) or transistor assisted
zener isn't such a bad idea for a simpler design for one injector.

But I may have to drive 4 injectors, so power supply consumption becomes
an issue.

The H-bridge results in less total power consumption from the power
supply, since energy is returned to the supply's output cap during
flyback, instead of being dissipated as with a zener.

For instance, with a 1200RPM engine condition, with 5ms injects per
cycle consisting of 1ms of 30A peak and 4ms of 20A hold current
(assuming square current waveforms here) we have to supply only the 2
ohm coil resistance power dissipation with a full bridge, or 100W
average power per injector. (Actually 99W from the simulator.)

But with a zener flyback damper, the power dissipation from the
simulator is about 209W.


Maybe I will be building that H-bridge after all.

Thanks for the input.


Good day!




--
_______________________________________________________________________
Christopher R. Carlen
Principal Laser/Optical Technologist
Sandia National Laboratories CA USA
[email protected] -- NOTE: Remove "BOGUS" from email address to reply.
 
J

John Larkin

Jan 1, 1970
0
It was PWM'd at 20KHz, so the structure between the transformer and
the gate was only for DC-restoration.

...Jim Thompson

Sure: transformer coupling followed by DC restoration is a nice drive
trick, if the time constants don't get'cha.

John
 
R

R Adsett

Jan 1, 1970
0
I am looking at a employing four of IRFP260N 200V 50A 0.04ohm 300W
devices. They seem beefy enough to not need to parallel devices.

The question is how to accomplish the high side drive with a 150V B+
rail? I am considering using a DC-DC converter to provide isolated 12V,
and powering a FET driver such as TC4426 with that. The input of the
FET driver would be driven by the output of a fast optocoupler powered
by the DC-DC converter as well. The input to the optocoupler is then
accessible at the ground level, and can turn my FET on and off. The
problem with this is that a bunch of circuitry (the DC-DC output
circuitry, and everything to the right of the optocoupler LED) must
swing with the output voltage. I guess there is nothing inherently
wrong with this, but it must radiate a lot of EMI as well as presenting
capacitance to the output.

Is this the right approach, or is there a better way?

I think it is impossible to use a pulse transformer in this situation,
because there are times when the solenoid must be switched on for a
solid 500us. But I haven't evaluated this option in depth yet, as I
have only recently settled upon the H-bridge approach after realizing
the factors that make a 1/2 bridge impractical, as well as a zener catch
diode approach.
Why not use a high side gate drive like the IR 2110? If the high side on
time is too long to sustain the drive you can always add a charge pump
circuit. IR has an app note on doing just that. Two 2110's would give
you all the gate drives needed for a full H.

Robert
 
Top