[keebler303]

In my free time, I have been playing around with a dead simple EDM power supply design. I had some IRF610's (200V, 3.3A) laying around so I paralled 6 of them for my intended 15A max current. All the drains and sources are connected directly. Each gate is joined to a common point through a 100 ohm gate resistor. The drains are connected to the "workpiece", source runs through a .1 ohm shunt resistor and then to ground. The "electrode" is connected to +63V on my linear power supply.

I am using a comparator to disable my gate driver at the desired current threshold. I am driving the gate driver with a function generator, set at 1kHz 50% duty square wave. I am monitoring the gate drive signal with a scope and it seems to work properly. The current limiting is running at something like 10-20 kHz while my square wave is high.

I tested the circuit using an LED and resistor between electrode and workpiece and it worked properly (turning on and off). Then I decided to raise the bets and try a real spark. I used a 23 ohm resistor in series to protect the circuit in case my current limiting didn't work. After I "sparked" the electrode with the workpiece (just wires at this point in the game) it appears that I have killed my MOSFET's. They will not turn off now.

I think this may be dv/dt related? I think it cannot be due to current because the series resistor limits the current to under 3 amps, less than the rating of a single MOSFET, let alone the 6 in parallel I have.

Anyone have any great ideas?

Thanks
Matt

[farmerboy1967]

The IRF610 is only rated at 43 watts at 25C. 63V * 15A is 945 watts, not counting the resistance. The mosfets rated at 258 watts.

You really need to find some bigger mosfets.

[keebler303]

43 watts is the power dissipation, not how much power it can switch.With Rds on of 1.5 ohms, each mosfet can carry a current of 5.3A without exceeding the 43 watt power dissipation. Of course thats assuming that it stayed at 25C, which it obviously would not.

But I don't think that's the problem anyway. As I stated earlier, I used a 23 ohm series resistor to limit the current. Thus making the peak current the MOSFETS could see be:

I=V/R = 63/23 = 2.74A

The 6 MOSFETs in parallel exhibit a .25ohm equivalent resistance. Which would dissipate less than 2 watts total in the MOSFETs. The MOSFETs are heatsinked quite well and there is a fan blowing on the heatsink.

I don't think that current or power dissipation is to blame here. I have some 200V 83A MOSFETs sitting here but those are a little pricey for the trial and error approach.

Matt

[neilw20]

Consider the connections between the drains have some inductance. Small as it is.
You might have made a monster pushpull oscillator.
Go with a single 83A device.
With your current limit resistors, it will survive.
Even though the wires you have may be short, they WILL have inductance.
Big high voltage spikes, that will be a challenge to measure may result.

Just bite the bullet on a BIG MOSFET.

Even 100A devices can be affordable.
You need at least 50% margin on voltage.

63v * 2 = 126v. 200v device should be OK.

Make sure the driver is capable of sinking and sourcing at least 1A at 0-10v. Use one of the IR family of current limiting drivers.
All the work is done with one chip. IR2112 or similar.

Use a gate resistor - maybe even 200 ohms.
The slower you switch it off the less chance of creating spikes, at the expense of a slight switching loss.

Put a 15v 1W zener from gate to source. Good insurance.

The reflected gate capacitance rises as the drain current increases.

[keebler303]

Neil

Thanks for the reply. So you think voltage spikes have killed my MOSFETs? Why would the 200V rating on an 83A MOSFET be any more rugged than a 200V rating on my IRF610s? Larger parasitic capacitances to absorb the spikes? My current limiting resistor was wirewound, probably contributing to the inductance issue. I was using 100 ohm gate resistors. My gate driver is from TI, capable of 4A peak current. My Vgs is 15 volts.

The 83A MOSFETs wer eonly $5.80 each so its not like they are made of gold, but I'd rather not immediately blow one due a terrible design.

Here is a schematic to clarify what I am working with.

Cheers
Matt

[wildwestpat]

Hi Matt

There are several problems with your proposal to operate the semiconductors in parallel.

Firstly the devices will differ in their conduction time and dynamics of the switching resistance from off to on. Even devices out of the same dice (batch) of silicon will exhibit different characteristics.

Secondly the current will not share equally amongst the devices as they switch on due to the variations in the inductance and capacitance of the leads to the common point on the load / power supply.

Your gate drive balancing resistors being inductive does not help but the real problem is that it is impractical to parallel devices as you are proposing as they can not be matched with sufficient accuracy as semiconductors nor in the external component wiring.

As others are telling you get devices that will take the current and voltage as well as staying out of the critical disipation during the switching cycle. If you consult the data sheets the rate of rise of current and voltage across the device has to be taken out of the chip into the heat sink before the next pulse. This is called secondary breakdown and is limited by the thermal dynamics inside the device. This paper will give you a taster of what is involver if you want to persist with parallel devices:- http://www.semikron.com/internet/web..._3-3_7_1_2.pdf

Sorry just get the rating of the devices that match your application.

Pat

[keebler303]

OK, the concensus seems to be to use a bigger MOSFET, I'll give it a try.

Thanks All
Matt

[neilw20]

Cnczone went offline on me so this is belated answer.
Make sure your gate drive is relative to the source, as it is possible to have a 63v spike across the current sense, making the gate absolute voltage spec a very likely reason for you killing your devices.
Your marked logic ground should be the negative of the supply and -ve current sense signal relative to logic ground which needs moving to the source pin.
There are many ways to fix this, but don't exceed the gate voltage specified.

[bharbour]

I agree with Neil's comment about checking to make sure that the gate drive is referenced to the FET source leads.

I have done a few of the current sense circuits like you are describing and wire-wound sense resistors just don't work. The inductive spike at switch transitions makes your voltage sense comparator circuit not work predictably. I have had good luck with the low inductance sense resistors that Digikey sells.

I suspect that the FET problem may also be related to the voltage spiking, especially if the gate drive is ground referenced instead of tied directly to the source. That loop between the sources, the gate driver and the gates needs to be as small and tight as possible. The switching currents are pretty large and they can be very fast.

I think that you are fine with the paralleled MOSFETs. Bipolar Junction Transistors don't share current properly because the temperature coefficent is positive, causing the transistor that is carrying more current (and hotter) to carry more current still. The temperature coefficient on MOSFETS is negative, the hotter transistor will have a higher Rds causing it to carry less current, so they will share current properly. With multiple MOSFETs, the inductance in the individual parts is paralleled, causing it to go down. Probably the biggest problem with paralleling multiple MOSFETs is that it requires more gate drive current. That said, most of the big MOSFET systems I have seen use multiple paralleled MOSFETs.

I would make sure that the inductance of the system wiring is as low as possible. Maybe use flat strips of copper to parallel the MOSFETs and put a good sized electrolytic cap between the ground side of your sense resistor and the +63V to the supply. Some small (0.1 uF or so) ceramics would probably be good as well. Keep the wiring as short as possible, especially the source to driver chip and driver chip to gate wiring. I think that the easiest way to blow MOSFETs is to exceed the maximum gate to source voltage spec.

I am not sure on the gate resistor. I would put individual resistors from the gate driver to the gates, probably on the order of 30 ohms each. Other peope have offered contradictry advice. My thinking is that switching the MOSFET quickly will result in less power dissipation while you are transitioning from on to off and vice versa. I doubt that your problems are thermal right now though, because it did not sound like you were pushing much current or running it long. The individual resistors will isolate the MOSFET gates slightly, to better accomodate different turn on thresholds and such.

To dissipate the 43 watts mentioned, you will need real heat sinking for each MOSFET. Think about a 40 watt lightbulb.

Good Luck,
Bob

[Larken]

I think you need to add an inductor (on the high side) to give the circuit some charge time. The way it is now the current rises so fast that the current sense circuit isn't fast enough.
Do you have a (very fast) sense circuit that shuts the gate drive off when the current reaches a set peak ? Without it, it will definitly blow.

Also you should lower the 100 ohm res to about 10 ohms and make sure your drive circuit can output a clean 15 volt square wave with lots of current.

Those mosfets arn't designed for repetative high current abuse, get some newer ones that are.

I see you are grounding the work with the live output of the fets, Is you 63volt power supply isolated ?

[Blaugen]

dear friend: I have the same problem. see my posts on Mosfet driver signal generator - Page 2

[neilw20]

I have not tried this circuit yet, but it certainly has promise.
At first the circuit is a little confusing, but if you join the names with lines it starts to make sense.
You must closely observe proper connection techniques for this circuit as voltage drops in the wires, inductive and resistive, will fry the MOSFETs with great abandon.
With out referring to the IRF data sheets for the driver chips or design knowledge in this are will most likely create the odd bit of $smoke.
Easy to get out, hard to get back in again.

[wildwestpat]

There are number of physical conditions that need to be taken into account if you are going to operate power MOSFETs in parallel. Use separate resistors for each device i.e no common sharing of components.

Ideally the devices should be matched at the current used. Yes there are a load of commercial designs that use parallel operated power devices but these will almost certainly be from the same die.

There are useful papers to be found by Googling operating power MOSFETS in parallel from which this is a general summary.

http://www.ixyspower.com/images/tech...s/IXAN0058.pdf

And

http://www.fairchildsemi.com/an/AN/AN-558.pdf

are relevant to this thread.

Hope this helps. Regards - Pat

[Blaugen]

I have a solution for a part of the circuit, which is the current limit resistor: I bought a threaded ceramic core and wind around a first winding on a sense of the double of ohms and another in the other sense (in antiparallel) over the first. Thus I created an antiinductive charge. It is a little, but it is something.

[Blaugen]

If you have circuits, especially with pulse generators, please upload! I finish my first EDM 5 years ago (its building takes me 3 years), it is a Lazarenko type (with RC oscillator) and runs well, but it’s just a metal disintegrator, now I wanna make a better machine. In my last post, when I said charge I meant load. sorry my grammar, I'm not american. a hug.