Why not place a zener diode across the mosfet? It will significantly clamp any reverse voltage spikes and will clamp any other voltage spikes to the zener voltage. Just make sure the zener voltage is above the normal operational voltage.

I want to hit this thing with a sledge hammer and do everything that could even possibly help. Tired of wiring up these mosfets.
Would you put the zener diode from the mosfets source to drain? And would this be in addition to diodes from the drain to positive terminal correct?

 

The zener diode goes across the mosfet reverse biased (in reverse). You will want to makes sure it can handle large peak currents. And make sure the supply voltage is less than the zener voltage, or you will destroy your zener diode.

 

There is zero chance your calc for the wire's inductance is correct. The values might be in nanohenries. I would expect something in the range of a microhenry per metre.

To minimize the inductance of the power cables to the head, keep them as close together as possible. Normal practice is to twist them, but that is probably impractical. Spiral wrap or an abundance of cable ties are probably realistic. Better is to use multiple twisted pairs connected together, but that is also likely impractical and/or undesirable, though it could facilitate balancing inductance seen by multiple groups of FET. Keeping the "out" and "return" conductors close together allows the magnetic fields due to the currents to at least partially cancel, which reduces the inductance. Instead of the total inductance being the sum of the two lengths it becomes some fraction of that you would get from a single length.

The peak current that a diode to discharge the inductance would see is equal to the current at the instant the FETs turn off, but the inductance will be small and the total energy not huge. Diodes rated for about a tenth of the actual weld current would likely be adequate, though if their duty cycle becomes high due to rapid re-firing of the FETs, a more conservative rating might be in order. Ideally, these diodes should have very short "reverse recovery time" so that they don't actually greatly increase the current the FETs have to switch when they turn on.

The FETs being used have body diodes with well-specified avalanche rating. In the absence of the diodes previously mentioned, the body diodes would avalanche to discharge the inductance. This is perfectly acceptable if the ratings are observed. If you use the external discharge diode there are issues of placement to minimize inductance in the diode's path. You might get into a situation where both that diode and the FETs' body diodes conduct, which should likely be OK.

I strongly suspect the gate drive is the biggest problem.

The cable to the gates is probably long enough to be acting as a transmission line, assuming it is a twisted pair. This means that when the gate driver puts a rising edge on the near end, it "sees" the characteristic impedance of the cable rather than the load at the other end. It really depends on the slew rate of the driver output - if it is low enough the cable won't be seen as a transmission line. If it is seen as a transmission line, a "voltage front" will arrive at the FET end and the very low impedance of the FETs' gates (capacitance) will cause the voltage to drop (conservation of energy). It will take several cable delays before the voltage at the FET gates settles, and by then they have probably melted. If the gate drive connection isn't a twisted pair, there is even less chance of reasonable performance.

As has been mentioned, the drivers should be very close to the FETs. Though the FETs chosen have highly advantageous low total gate charge and the driver can deliver a lot of current, it is still probably inadequate for more than about 6 FETs. This makes for some difficulties because you now require multiple drivers or an additional power stage for the driver. It is critical that all FETs turn on and off at the same time time, otherwise you will get poor sharing of the current. I think the refs Dana posted cover this.

The decoupling capacitor for the supply for the gate driver is grossly inadequate. The datasheet recommends appropriate values. It is extremely critical that the caps are as close as possible to the driver to minimize inductance between them. Use X7R types for the ceramics. Other types may be cheaper but perform much worse (C0G is better, but the improvement won't be worth the much greater cost).

My suspicion is that 12 well-driven FETs will perform much better than 24 badly driven ones. You might even be able to use fewer.

 

On avalanche of the body diodes or external zeners:

The instantaneous dissipation will be equal to the product of the current and the zener/avalanche diode voltage. For example, if you used a 50 V zener and the current at FET turn-off was 1000 amperes, the instantaneous dissipation would be 50 kilowatts, which is rather a lot. The inductance must be known to calculate the energy.

 

On avalanche of the body diodes or external zeners:

The instantaneous dissipation will be equal to the product of the current and the zener/avalanche diode voltage. For example, if you used a 50 V zener and the current at FET turn-off was 1000 amperes, the instantaneous dissipation would be 50 kilowatts, which is rather a lot. The inductance must be known to calculate the energy.

Woops, meant 10,000 nH.
That's just from a calculator on this site though, so probably not much better than guess...

 

But lots to read through, thanks for the tips!

 

Woops, meant 10,000 nH.
That's just from a calculator on this site though, so probably not much better than guess...

Wouldn't nearby objects with high permeability increase the inductance significantly? Did you account for that?

 

Hi,
So Im working on what is basically a fancy capacitor discharge welder, and I keep frying the mosfets so they get stuck on.
Everything works perfectly at low voltages (~12V) but only for a few times then the mosfets are stuck on and no longer function as switches. Ive gone through so many of these things its time to ask for help.
Attached is a wiring diagram of the power circuit. the lower part is the driver/mosfet/capacitor bank I'm having trouble with, I included the top circuit (which is just a way to sense head contact for the controller) just in case it's causing something funny to happen.
Basically the weld head comes down, both electrodes make contact with the metal, then after a short dwell period the caps fire for about 300us.
The resistance of the weld head/ and wires is very very low so I made a .05 Ohm "resistor" out of resistance wire used in heaters which I confirmed to be ~.054 Ohms. I originally had 12 mosfets in parallel but after a few failures doubled down and put 24 but they still fail closed at the same voltage level so I don't think it's a current problem.
I'm thinking maybe there's a voltage spike which is killing the mosfets but have no way of measuring it and they die when the capacitors are only at 15v so it would have to be a big voltage spike... I also might just be doing something obvious and dumb lol
Any ideas or pointers would be greatly appreciated!
Thanks guys!!

A firm I worked for made CD stud welders - they used a dirty great SCR.

Originally they used an SCR that stood around 4" or so tall and had a big square base with 4 bolt holes.

When those became unavailable - they had loads of fun trying to clamp a pill type device with (allegedly) the same ratings.

They used an old Mullard module with a pulse transformer output to drive the gate. Apparently inductive artifacts and/or dirty workpiece can disrupt the holding current, so it needs a burst of pulses long enough to ensure dumping the capacitor bank.

Decently rated SCRs are getting harder to come by and I'm considering how to adapt some applications to MOSFETs - but you'll need a seriously big bank of them for that one.

 

There are other forms of mosfets, ones made to do industrial type power transfer. Not the hobby normal electronics TO-220 case style but the Isotop industrial ones made to work in welders, and big DC motor drives. The EDM machine I'm slowly building uses them to dump the caps into the electrode.

Here is a picture of what I mean by Isotop, that is a STMicro brand name, can't remember the JEDEC number.

 

Not all caps are equal, esr curves for examination -



Read datasheets carefully, not all Cs of same value from different manufactuers
perform equally.


Regards, Dana.