Any ideas to stop a huge negative voltage spike during mosfet shut off?

Thread Starter

Cfez202

Joined Apr 25, 2018
37
Hi everyone,
Hopefully you have some thought on this weird issue I've been having.

Basically I have a mosfet acting as a switch. The end goal is to have this circuit act as a capacitive discharge welder. When on, it is essentially short circuiting a capacitor bank through a ~1 ohm resistor to limit amperage (the final design will have a much lower resistance and parallel mosfets to handle the amperage).

My problem is when the mosfet turns off, there is a HUGE negative voltage spike which kills the fet. Charging the capacitors to 24v produces an almost 300V negative spike.

I assumed this was due to inductances and tried adding diodes to kill the spike but no matter what I try nothing effects the magnitude of the spike.

My testing layout is attached to paint a picture of what I'm doing. The Fet needs to turn on and off very quickly (will eventually run PWM at 62kHz) so options like relays or SCRs wont work.

Before anyone asks, I'm pretty much positive the routing/ layout quality of my mosfet and driver have nothing to do with this particular issue. The problem is something more substantial based on the "change one thing at a time" bounding tests I've done (below if you're curious)
  • 3 completely different driver circuits with 3 different PCB layouts (1 good, one ok, one not so good, but all with short, fat traces). I also remade those 3 circuits on breadboards with real lazy wiring to see if there was any difference (there wasn't).
  • Mosfet pcb mounted, mosfet through hole mounted on breadboard, mosfet floating at the end of ~4" wires (zero difference)
  • Flyback diodes across every combination of points in the test layout (I tried multiple diodes, most recent were BYC30W-600PQ and C3D16065A-ND)
  • 12 feet of 8ga wire between the two ring connectors
  • 5 inches of 8ga wire between the two ring connectors
  • Moving R1 to before the source pin as well as to after the drain pin
  • Diode in series with drain wire
  • Changing gate resistor values. Tried 2.2, 3.3, 10, 22, 100, 220, 470 and 1k ohm. Currently have a resistor/ diode to make the shut off gate resistance 220 while keeping turn on resistance 10.
Not a single one of these changes effected the negative voltage spike at all. I was sure I just needed a diode to kill the spike but nothing helps. The circuit is so simple I just can't think of anything else I could be doing wrong. Anyone have good ideas to try??

Edit: I forgot to mention... at the suggestion of another forum I added an RC circuit across the drain/source. Originally I had a -300V spike and then lots of ringing, the added RC circuit eliminated that but did not effect the magnitude of the initial spike.
 

Attachments

Bordodynov

Joined May 20, 2015
3,177
Your resistor is inductance at the same time. More correctly do it using bifilar winding. But if you conceived this way (i.e. you need an inductance), then emissions are fought using suppressors. At the same time, it is soldered practically to the protected electrodes. I recommend to protect (parallel to the protected electrodes of the transistor) gate-source and drain-source.
 

AlbertHall

Joined Jun 4, 2014
12,345
To catch the spike you need a diode between C and D with the anode connected to D. The diode will carry the full discharge current briefly when the FET switches off so check the diode surge rating.

The problem is caused by the inductor you have in the circuit. If your circuit needs the resistance but doesn't need the inductance then as @Bordodynov says wind the wire in a non-inductive manner like this:
 

jpanhalt

Joined Jan 18, 2008
11,087
Why are you trying to solve problems with a circuit that you won't be using in the final design? As you point out, that resistor is wrong. There are lots of other wrong things you could also test, like 12 feet of 8 awg wire. That is way too long/small for your electrode connections.

I would go with your best besign and try to work through its problems. Here is one recent example: https://www.keenlab.de/index.php/portfolio-item/kweld/ His design development is documented on another site linked to therein.
 

Thread Starter

Cfez202

Joined Apr 25, 2018
37
Your resistor is inductance at the same time. More correctly do it using bifilar winding. But if you conceived this way (i.e. you need an inductance), then emissions are fought using suppressors. At the same time, it is soldered practically to the protected electrodes. I recommend to protect (parallel to the protected electrodes of the transistor) gate-source and drain-source.
I thought so too, but regardless of what value I make that resistor and what length of wire I use the spike magnitude does not change. Also adding diodes across the inductive loads do not effect the value. What else could be causing this? Is there something going on with the large capacitors perhaps?
 

Thread Starter

Cfez202

Joined Apr 25, 2018
37
To catch the spike you need a diode between C and D with the anode connected to D. The diode will carry the full discharge current briefly when the FET switches off so check the diode surge rating.

The problem is caused by the inductor you have in the circuit. If your circuit needs the resistance but doesn't need the inductance then as @Bordodynov says wind the wire in a non-inductive manner like this:
That was my thought also, but unfortunately adding a diode across any point in that image, including C-D doesn't effect the negative voltage spike. The inductance is already very low and regardless of whether I make that resistor 1 ohm (2 ft of resistance wire) or .1 ohms (2 inches of resistance wire) the voltage is the same. To give you a sense of wire lengths, that image is pretty much to scale. There is not a lot going on in the way of inductive loads. The only thing i can think of is maybe the capacitors are causing the odd behavior but I dont know that the actual problem is, never mind the solution lol
 

Thread Starter

Cfez202

Joined Apr 25, 2018
37
Why are you trying to solve problems with a circuit that you won't be using in the final design? As you point out, that resistor is wrong. There are lots of other wrong things you could also test, like 12 feet of 8 awg wire. That is way too long/small for your electrode connections.

I would go with your best besign and try to work through its problems. Here is one recent example: https://www.keenlab.de/index.php/portfolio-item/kweld/ His design development is documented on another site linked to therein.
HI,
So the actual final circuit, assembly and control of this is much more complicated than the image I attached. I'm just trying to diagnose this problem. I originally started by doing complete systems (hence the 3 different PCBs) but since i found bread-boarding doesn't change the result I'm doing that until I find a cure.
I've been removing piece by piece until now I'm left with nothing but the main driver circuit, a single mosfet and my capacitors but the voltage spike has not changed at all. What gets me is I'm basically laying out exactly what that kweld and other cap discharge examples are doing but mine is somehow making wild mosfet murdering voltage spikes.
Just hoping some of you guys have some ideas.
 

ian field

Joined Oct 27, 2012
6,536
Hi everyone,
Hopefully you have some thought on this weird issue I've been having.

Basically I have a mosfet acting as a switch. The end goal is to have this circuit act as a capacitive discharge welder. When on, it is essentially short circuiting a capacitor bank through a ~1 ohm resistor to limit amperage (the final design will have a much lower resistance and parallel mosfets to handle the amperage).

My problem is when the mosfet turns off, there is a HUGE negative voltage spike which kills the fet. Charging the capacitors to 24v produces an almost 300V negative spike.

I assumed this was due to inductances and tried adding diodes to kill the spike but no matter what I try nothing effects the magnitude of the spike.

My testing layout is attached to paint a picture of what I'm doing. The Fet needs to turn on and off very quickly (will eventually run PWM at 62kHz) so options like relays or SCRs wont work.

Before anyone asks, I'm pretty much positive the routing/ layout quality of my mosfet and driver have nothing to do with this particular issue. The problem is something more substantial based on the "change one thing at a time" bounding tests I've done (below if you're curious)
  • 3 completely different driver circuits with 3 different PCB layouts (1 good, one ok, one not so good, but all with short, fat traces). I also remade those 3 circuits on breadboards with real lazy wiring to see if there was any difference (there wasn't).
  • Mosfet pcb mounted, mosfet through hole mounted on breadboard, mosfet floating at the end of ~4" wires (zero difference)
  • Flyback diodes across every combination of points in the test layout (I tried multiple diodes, most recent were BYC30W-600PQ and C3D16065A-ND)
  • 12 feet of 8ga wire between the two ring connectors
  • 5 inches of 8ga wire between the two ring connectors
  • Moving R1 to before the source pin as well as to after the drain pin
  • Diode in series with drain wire
  • Changing gate resistor values. Tried 2.2, 3.3, 10, 22, 100, 220, 470 and 1k ohm. Currently have a resistor/ diode to make the shut off gate resistance 220 while keeping turn on resistance 10.
Not a single one of these changes effected the negative voltage spike at all. I was sure I just needed a diode to kill the spike but nothing helps. The circuit is so simple I just can't think of anything else I could be doing wrong. Anyone have good ideas to try??

Edit: I forgot to mention... at the suggestion of another forum I added an RC circuit across the drain/source. Originally I had a -300V spike and then lots of ringing, the added RC circuit eliminated that but did not effect the magnitude of the initial spike.
A back emf spike can be handled by any of various clamps/surge arresetors - a reverse spike should be clamped by the internal "body diode" anyway.
 

Thread Starter

Cfez202

Joined Apr 25, 2018
37
Given that large electrolytic caps consist of coiled sheets of foil, how much inductance does each cap have?
Thats a good question... I dont know. The spec sheet has no info on ESL, ESR is 26 mOhms...
I have put a diode across them though with no change... Maybe the diode was too slow or oriented incorrectly? Ill try this again.
 

Thread Starter

Cfez202

Joined Apr 25, 2018
37
A back emf spike can be handled by any of various clamps/surge arresetors - a reverse spike should be clamped by the internal "body diode" anyway.
If its killing mosfets is that because the negative spike is higher than the mosfets voltage capability? So maybe I can just find fets with 400-600V ratings and call it a day?
What kind of circuit would you use to take that load away from the mosfets body diode? I've tried various clamp/ surge circuits found online with no luck.
 

jpanhalt

Joined Jan 18, 2008
11,087
Thats a good question... I dont know. The spec sheet has no info on ESL, ESR is 26 mOhms...
I have put a diode across them though with no change... Maybe the diode was too slow or oriented incorrectly? Ill try this again.
That data (inductance) for those caps is a little hard to find. From other sources, the capacitor per se is about 2 nH or less. The connections are a greater contributor, and the total is about 20 to 50 nH.

Source: http://www.cde.com/resources/catalogs/AEappGUIDE.pdf
 

MisterBill2

Joined Jan 23, 2018
18,176
Have you tried a high current rated diode, probably 500 volts or more, from t eh source bact to the drain, to short circuit that spike? Also, try disconnecting 3 of the capacitors to see if the spike changes. Then you may have another clue, if it does or does not change. That would not be a fix but it may tell us something.
And one additional thought is that if the wires in the circuit form a loop with any real space in between, then that loop has inductance, and so it would make sense to make that loop as small as possible.
 

Thread Starter

Cfez202

Joined Apr 25, 2018
37
Have you tried a high current rated diode, probably 500 volts or more, from t eh source bact to the drain, to short circuit that spike? Also, try disconnecting 3 of the capacitors to see if the spike changes. Then you may have another clue, if it does or does not change. That would not be a fix but it may tell us something.
And one additional thought is that if the wires in the circuit form a loop with any real space in between, then that loop has inductance, and so it would make sense to make that loop as small as possible.
Ah good idea with the # of caps... Ill play around with that tonight.

I have tried a Source to drain diode (with both diodes in the og post, 600v+) and it either has no effect or holds the circuit closed once fired.
 

Sensacell

Joined Jun 19, 2012
3,432
I think it would be very helpful to post a photo of your setup.

At the speed and current levels this circuit operates, every bit of wiring becomes a significant circuit element.
 
Top