Killing MOSFETS at high bus voltages (h-bridge)

[JohnWondersWhy]

Hello,

I've seem to be able to kill every mosfet I try when I use a bus voltage > 50v.

I attached my circuit, all real basic stuff, just like the data sheets recommend. The circuit works great below 50 volts. But above that, both mosfets will die, even at near zero motor amperage.

I've scoped everything out . . . all the voltage readings at the mosfets are in range.

Everything on the driver is in range except at bus voltages >45, I get a negative voltage spike between Ho and COM at the end of a pulse. At 50v the spike is -10v, at 100v the spike is -20v. Again, that's when I measure relative to COM.

Taking readings at the mosfet between g-d-s, everything looks clean and crisp, no overshoots. Seems like the driver ought to be the component that fails. In desperation, I've been poking around with different caps, resisitors, diodes, big and small mosfets, and adding parts randomly. Something is seriously wrong . . .

This is nuts. Any ideas of what I can try?

 

[jpanhalt]

Which mosfets are you using? John

 

[JohnWondersWhy]

I tried these mosfets;
irf3415 (at lower voltages)
irf13n50apbf
stp16ns25
stp30nf20
irfb4332

 

[dknguyen]

Did you try adding flyback diodes (like schotkeys) in anti-parallel across all the MOSFET's (I do not mean the slow, large Vd parasitic diode inside the MOSFETs that is always there and is only sometimes drawn into the MOSFET symbol)?

You also have a large capacitor across the power terminals to absorb the voltage spike from cutting off power to the source? (battery? rectified mains?)

You're also sure it's not shoot through? (despite you saying that only one side receives PWM at any one time in your drawing)

You could also try snubber circuits maybe, like an RC snubber but I don't know much about those:
https://www.electro-tech-online.com/custompdfs/2007/12/slup100.pdf

If you stick shorten the motor wires down to almost zero and stick a resistive load into there instead of a motor, do you still have problems killing the MOSFETs?

Please keep us updated on this. I'm trying to build something similar.

 

[Krumlink]

Try an IRFP250 With a heatsink on it. Should work good, since it has 200V and 20A inputs.

 

[OutToLunch]

scope out the gates of the FETs to make sure there they aren't shorting the input rail.

 

[kooloosj]

Hi JohnWondersWhy,

Here's a link you should check out for some insights into what you're playing with.
http://www.mcmanis.com/chuck/robotics/projects/esc2/hbridge_spiking.html

How are you prototyping this? Lead Inductance is hugely important in keeping this type of thing alive at elevated powers.
Kind Regards, JK

 

[JohnWondersWhy]

Hello,

I wanted to anwser the responses I received so far. I tried everything to no effect. I'm going to try everything all at once, with ALL new components to see what happens (although, quite honestly, I'm burning up parts every 1/2 hour or so).

Keep in mind, I'm not trying to drive anything at this point. I just hooked everything up for what I thought was going to be a quick test. The motor is drawing about 100mA !!!

- external flyback diodes didn't show any effect in my circuit

- I have a large capacitor on the bus. I couldn't see anything happening when I scoped it. nice and clean

- shoot thru? I'm not exactly sure what that means. I did scope the digital signals to ensure only ONE ir2104 is receiving pwm. however the one ir2104 recieving the pwm does alternately pulse HO and LO; the dead time there looks OK. see the attached picture.

- I tried using a non-wirewound 300 ohm resistor, instead of the motor, exact same problems

- I'm not sure I understood "scope out the gates of the FETs to make sure there they aren't shorting the input rail." see attached picture

- OK, I'll reluctantly admit this is being done on a breadboard. I'd expect some noise and heat; but, killing these things at low voltage, low current, and no significant heating is drepressing me terribly. I tried the snubber without success.


All my scoping looks good except for between HO and COM. see attached picture. the spike occurs around 45~55 volts. even a diode between HO and COM has no effect.

I'd cry if I'd thought it would help.

 

[dknguyen]

Don't breadboards have a limit of about 50-75V? Well that wouldn't explain why the breadboard still works when you replace the components and run it at a shorter time. If there's no heating and it still doesn't work with just a 300 ohm resistor, I would guess there's too much inductance somewhere in your circuit (but wouldn't that also kill it at lower voltages too?)

 

[jmoorsie]

i agree with dknguyen, the breadboard is probably your problem...

 

[kooloosj]

Hi,
Some possibly relevant data.
https://www.cnczone.com/forums/showthread.php?t=23216&page=3
JK

 

[Hero999]

Don't breadboards have a limit of about 50-75V?

That's purely for safety reasons, the breakdown voltage is probably closer to something like 500V to 1kV but I certainly wouldn't go over more than a couple of 100VDC and I wouldn't use it for mains as there could be transients of 1kV which could cause arcing.

 

[JohnWondersWhy]

To follow up . . .

I hardwired a pc board . . . things worked OK until the first mosfet died again. Then pretty much nothing worked as I chanegd parts.

The short solution is that I gave in and made a dedicated printed circuit board with 100% new parts (no breadboard, no hand wired pc board). Everything worked OK. Also, the one time I shorted pins; I then replaced the drivers and the mosfets.

I think my problems were primarly caused by this;

1. my microcontroller may have been skipping a beat every so often

2. some breadboard contacts may have been loose

3. my digital scope was out of range approaching 48v (that caused the seemingly apparent negative spikes).

4. the ir2104 drivers were probably damaged everytime a mosfet shorted out (even though this had never been a problem with past projects). I think old drivers killed new mosfets and old mosfets killed new drivers; along with an occasional microcontroller burp and breadboard mis-connect. Like I said early on, I could capture any problems when scoping around; but, apparently there were problems not being captured.



thanks for everyone's help.

 

[kooloosj]

Hi again JohnWondersWhy, Great effort indeed, I've played with this combination quite a bit in the past and found it a real challenge to create a robust bridge. You've done well.JK

 

[HZYN]

yes.. john something is wrong . when you apply 50 volts pwm ...at the end of the cycle dc motor's electro magnetic field reversing the voltage maybe 250 volts spike ... i guess u cant see on your scope . 180volts.....!!!!???