Current pulse generator project

[PartialDischarge]

I've been working on a current pulser, short pulses (adjustable 5-100us) in the 100s of Ampere range (adjustable). This one I made should get to 600-700A. I've been researching thyratrons for this reason but this one is IGBT based.

The system is simple, a high voltage adjustable 0-1000V DC-DC converter charges a capacitor (42uF) that is discharged through a low inductance power resistor (Vishay RPS250) by means of a IGBT (CM300DC). The pulse is a one shot generated by a 555 an applied to a IGBT driver. I used thin paralleled copper bars to make the joints and kept distances small. The heatsink acts as ground plane. I got a respectable 210ns rise time.

This is the system:





And some measurements with the Tek A6303 and AM503B amplifier. I didn't go over 100A just yet, since I need something else to go beyond that.

50A pulse, 10us.



50A pulse rise time.



100A pulse and the VGE of the IGBT



100A rise time.

[daqq]

Nice setup!

Minor observations/tips:

It's better to have the gate drive twisted together, this lowers the loop area, lowering the gate series inductance.

A bit lower main circuit inductance could be achieved by connecting E2 from the other side.

A steeper edge could possibly be achieved by using a higher resistance (thus lowering the L/R constant) with higher voltage. At these speeds you could be dealing with a limitation of the transistor.

If you want some (other) resistors that can take some serious pulsed abuse, see:
http://www.hvrint.com/rtrlseries.htm
https://www.hvproducts.de/en/hv-passive-components/bulk-ceramic-non-inductive-resistors/

If you need this for some really serious applications (since you've been looking into thyratrons), you could take a look at the goodies Behlke has:
http://www.behlke.com/productlines.htm

Don't forget to add a snubber if you'll be connecting loads that add inductance to the system.

[Wolfgang]

Hi,

nice pulser ! Some comments though:

- IIRC High power IGBTs as well as MOSFETs have a dI/dt rating. If you go to still higher currents this rating could be exceeded.
- The base circuit must be free of resonances (including the hefty and nonlinear IGBT capacitance). A voltage spike caused by a resonance could kill
  the gates. Damping is a good idea. Use twisted wire and a termination resistance (e.g., 120Ohms).
- Turn-off time of IGBTs can be long. When they need to be fast, you have to "suck out" current from the gate. Your IGBT driver probably does that.

Some years ago I tried to build a can crusher using semiconductors. No survivers so far. The weapon of choice for games like that is probably an IXYS pulse power thyristor. These bastards withstand up to 90kA and a few kV. But the price is astronomic.

Have fun
  Wolfgang

[Insatman]

I had lots of experience designing low impedance drivers.  What we did was use a step-down transformer so you can switch at higher voltages and lower dI/dt.  Then step-down the drive impedance with a pulse transformer.  With proper design you get better rise/fall times.  Of course pulse length is limited by the magnetics.

[David Hess]

What about grounding the gate of the IGBT and driving the emitter with the drain of a power MOSFET which then only has to sustain low voltages?

[PartialDischarge]

Still have some improvements to make but it's not that bad. Specifically the rise/fall times of VCE tell another story and thats what *I believe* is quoted in most datasheets, for example Belhke pulsers.

For the 50A pulse (yellow trace), the 110V rail (2.2ohm resistor) has a 17.5ns fall time which is pretty good for a 1200V device.



The fall time of the current is very fast at 65ns, and the voltage has a 45ns rise with a 60V overshoot which I don't care for now as I still have room up to 1200V.

[PartialDischarge]

What we did was use a step-down transformer so you can switch at higher voltages and lower dI/dt.  Then step-down the drive impedance with a pulse transformer.  With proper design you get better rise/fall times.  Of course pulse length is limited by the magnetics.

I don't think I'm following the idea, could you draw a concept schematic?

[PartialDischarge]

Here are the datasheets of the modules used:

IGBT: https://www.digchip.com/datasheets/download_datasheet.php?id=5967568&part-number=CM300DC-24NFM
IGBT driver: http://www.symmetron.ru/suppliers/power-integrations/igd5xxe.pdf
DC-DC converter: http://www.analogmodules.com/admincenter/datasheets/562.pdf
Resistor: https://www.vishay.com/docs/50007/rps250.pdf

[PartialDischarge]

IIRC High power IGBTs as well as MOSFETs have a dI/dt rating. If you go to still higher currents this rating could be exceeded.

I'm in uncharted territory here, lack of data, and use of the igbt in single pulse mode makes this case special, will have to go up until it blows

Some years ago I tried to build a can crusher using semiconductors. No survivers so far. The weapon of choice for games like that is probably an IXYS pulse power thyristor. These bastards withstand up to 90kA and a few kV. But the price is astronomic.

Yep, for uncontrolled high current shorts, thyristors are the best for their extremely high I2t ratings

What about grounding the gate of the IGBT and driving the emitter with the drain of a power MOSFET which then only has to sustain low voltages?

Not a bad idea, and wouldn't mind having a negative rail, but I had already this drivers which are overkill here since I don't use the isolation or the protections but..

A bit lower main circuit inductance could be achieved by connecting E2 from the other side.

This is one of the pending improvements, I have a couple of 10V zeners from G-E, but there is ringing in the gate, and for short time it crosses the 20V barrier, will add a resistor to damp it also.

[daqq]

This is one of the pending improvements, I have a couple of 10V zeners from G-E, but there is ringing in the gate, and for short time it crosses the 20V barrier, will add a resistor to damp it also.

This could be a measurement issue. When measuring this kind of thing, it's better to do it differentially (hook one probe directly onto the GATE input, a second one on the EMITTER (the terminal close to the gate input) and do a substraction of the two), since when dealing with 100s of amps, grounds tend to bounce about.

[PartialDischarge]

This is one of the pending improvements, I have a couple of 10V zeners from G-E, but there is ringing in the gate, and for short time it crosses the 20V barrier, will add a resistor to damp it also.

This could be a measurement issue. When measuring this kind of thing, it's better to do it differentially (hook one probe directly onto the GATE input, a second one on the EMITTER (the terminal close to the gate input) and do a substraction of the two), since when dealing with 100s of amps, grounds tend to bounce about.

Yes, could be, these fast signals are hard to measure accurately, although I'd rather use my 7A13 plugin since digital scopes are bad for this kind of diff measurement. Anyway the current trace also shows some ringing so I believe its coming from the GE.

[duak]

When I was debugging fast power FET circuits, I found that the 'scope would show a damped ringing during the circuit's transitions even with the probe connected but grounded to itself.  This was due to HF common mode currents from the circuit passing thru the probe and scope.  I could eliminate them by passing the probe cable thru a high mu common mode choke a few times.

Another use for a high mu choke is to use it as a pulse compressor or intensifier.  The theory is to apply a voltage pulse to a high mu choke in series with the load.  The choke's inductance will initially limit the di/dt to allow the switch to fully close.  At some point the current saturates the core of the choke and it essentially shorts out and not longer limits di/dt giving a sharper, higher current pulse.  If another parallel C and series L is added, the pulse can be further compressed.

I found this effect when performing ESD susceptibility tests on a product when a common mode choke saturated and passed an intensified pulse to the power supply which then glitched.  I then found that this effect was used to generate narrow pulses for lasers and radar transmitters, far sharper and narrower than you'd think the switching devices would be able to generate.  I found a few papers on drivers using this effect, but no tutorials on how to roll your own - an exercise for the student, I suppose.  I've never tried to use this effect, so I can't give much direct advice.  If memory serves, a 10 A pulse became a 30 A pulse with about 1/4 the duration.

Intuitively, it might help here because it could allow the IGBT to operate within its safe operating area by reducing Vce more rapidly during turn on, perhaps compensating for the Miller effect.

Cheers,

[David Hess]

What about grounding the gate of the IGBT and driving the emitter with the drain of a power MOSFET which then only has to sustain low voltages?

Not a bad idea, and wouldn't mind having a negative rail, but I had already this drivers which are overkill here since I don't use the isolation or the protections but..

The driver drives the MOSFET instead.  The IGBT gate is then tied to a low impedance positive bias supply high enough to fully turn it on.

Bipolar designs that work this way are more difficult because the positive base bias has to supply high current but this is often done with a separate low voltage high current winding off of a transformer.  They do not literally ground the base.

Higher performance designs yet use special packaging for the transistor which does allow the base/gate to be directly grounded.  In the past, you could get TO-39 transistors with the base instead of the collector tied to the metal case.  Big stripeline RF transistors are still available with the emitter and base leads swapped.

[PartialDischarge]

The driver drives the MOSFET instead.  The IGBT gate is then tied to a low impedance positive bias supply high enough to fully turn it on.

Got it, will try it in a future version.

I found this effect when performing ESD susceptibility tests on a product when a common mode choke saturated and passed an intensified pulse to the power supply which then glitched.  I then found that this effect was used to generate narrow pulses for lasers and radar transmitters, far sharper and narrower than you'd think the switching devices would be able to generate.  I found a few papers on drivers using this effect, but no tutorials on how to roll your own - an exercise for the student, I suppose.  I've never tried to use this effect, so I can't give much direct advice.  If memory serves, a 10 A pulse became a 30 A pulse with about 1/4 the duration.

Interesting, will try to look up something on this

[PartialDischarge]

A steeper edge could possibly be achieved by using a higher resistance (thus lowering the L/R constant) with higher voltage. At these speeds you could be dealing with a limitation of the transistor.

Eventually this worked very well, I added another 3.3 ohms in series (3x10ohm paralleled) with the 2.2ohm and the rise time nearly halved. This limits the current pulse of this setup to about 200A, but with this I can test current probes of >2MHz bandwidth. I shortened the pulse to 5us for the sake of the resistors well-being.



Here are the 50A and 100A screenshots
50A and VCE


50A and VCE


100A pulse rise time (545Volts VCE)


100A pulse (545Volts VCE)

[David Hess]

The driver drives the MOSFET instead.  The IGBT gate is then tied to a low impedance positive bias supply high enough to fully turn it on.

Got it, will try it in a future version.

I should have mentioned that the only reason this works is that IGBTs and MOSFETs have the same gate drive requirements.

A high current low voltage power MOSFET driving the emitter of a bipolar transistor has the advantages of both devices and is sometimes used where either alone would be either prohibitively expensive or not high enough performance.  I have never tried it with a IGBT replacing the bipolar transistor but if I ever have an excuse to, I will.

[daqq]

Eventually this worked very well, I added another 3.3 ohms in series (3x10ohm paralleled) with the 2.2ohm and the rise time nearly halved. This limits the current pulse of this setup to about 200A, but with this I can test current probes of >2MHz bandwidth. I shortened the pulse to 5us for the sake of the resistors well-being.

Yeah, you are getting to the limits imposed by your transistor, but you still have some space there. You can probably further reduce the inductance of the setup by reorganizing things a bit (make the loop as small as possible) and adding a smaller capacitor parallel to the setup (the big one is on somewhat longer leads).

I think to go any faster, you'd have to go into MOSFET territory.

[PartialDischarge]

Yeah, you are getting to the limits imposed by your transistor, but you still have some space there. You can probably further reduce the inductance of the setup by reorganizing things a bit (make the loop as small as possible) and adding a smaller capacitor parallel to the setup (the big one is on somewhat longer leads).

I think to go any faster, you'd have to go into MOSFET territory.

I need a small loop to fit in the A6303 head, anyway I'm happy with the results. My aim was a 1-5MHz current pulse at at least 100A, and I plan to increase currents in future setups not necessarily (decrease) rise times. The fall time of the 100A pulse gives me 4MHz. I have another setup with 1800V 2000uF polypropylene busbar caps that are shorted with a thyristor. I used that to create high temperature arcs using thin wires. Never got to study the currents in that setup, I did estimate currents in excess of 10kA based on the voltage discharge profile (I may need a coaxial shunt) but I can tell that the cables did shake after each shot.