Why MOSFET keeps dying???

[3dgeo]

EDIT: SOLVED

Hi,

as some of You might know I'm remaking my sit/stand table control PCB and i'm keep getting trouble making it work – mosfet keeps dying.
I ordered 100V 57A 23mOhm IRF3710 http://www.irf.com/product-info/datasheets/data/irf3710.pdf
I didn't do a lot of testing with it due to being a noob and picked too high resistance fets for this purpose, but it seems they work, getting very hot in a few sec when only one actuator is powered (need to power two), so unless I want to burn my house down (heatsinks inbound) I needed to get less resistance fets, and I did.
Today I received 100V 195A 2mOhms (more than 11 times less resistance) https://www.infineon.com/dgdl/irfp4468pbf.pdf?fileId=5546d462533600a40153562c73472019
Aaaaaaand I killed 2 of them already....  I got back to old PCB and compared old and new circuits and found out that I missed 1 extra diode near mosfet and 2 small SMD (ceramic?) capacitors, picture below – marked them in yellow.

Mosfet purpose is to PWM 50V to ~24V.

In Picture on the left is old PCB, on the right is new one (I removed 2 extra mosfets because all they do is on/off power to relays, no PWM)
All mosfest I burned were without those 3 components, is this it? Is this the reason why fets dies? Before killing 3rd mosfet I thought I give it a try on this forum 
I know that mosfets dies probably from voltage spike from DC motors, but should diodes next to relays take care of it?
I'm assuming that those small caps "eats" spike and protects fets, please enlighten me, I'm getting a bit frustrating here 

BTW, I killed original mosfets same way as well (I know, I know, I'm a genus...)
Original mosfet markings:
ON5247
PEm 0519 D6
5291



OLD PCB (don't mind other parts, its for logic):

[SiliconWizard]

What do you drive the gate of your MOSFET with?

Those MOSFETs are monsters and thus have a pretty high input capacitance. If you drive them with say a digital output with limited current, they will take a long time to get in saturation, and during this time, they will dissipate A LOT of power. They may just die from overheating due to this.

Getting a lower RDSon transistor won't help in that matter - it may even make things worse, because it will actually need a lot more current to be switched on. An input capacitance of almost 20 nF is quite a lot.
You should use some kind of gate driver to drive the gate of those big MOSFETs.

See: https://en.wikipedia.org/wiki/Gate_driver for a very quick reference.

There are many such gate drivers available, for instance: http://www.ti.com/power-management/mosfet-igbt-gate-drivers/products.html

[3dgeo]

I used mosfet driver: TC4420 http://ww1.microchip.com/downloads/en/DeviceDoc/21419D.pdf
But probably TC4420 didn't get enough current cos I was feeding 10V with voltage divider from 50V.
And I doubt mosfets dies due to overheating cos they were dead cold (newones and original ones).
The only one  mosfet that got crazy hot (IRF3710) did survived, but heat was too hight to keep it on for a long time, and it did PWM it to 24V.

Edit: I just realized that TC4420 drive capability: 10nF, is it still suitable to drive that mosfet at low frequency, like 100hz? (original mosfets were driven only at 100hz)
Hm, I still can't believe they got killed due to this, they were cold, not even a bit warm, and original mosfets were driven from 5V with no driver and they died as well...

[Zero999]

No, those capacitors will make things worse if anything, especially the massive 2200μF one! They will cause a huge surge, every time the MOSFET is turned on, which is probably killing it,

Those capacitor shouldn't be in parallel with the load but the power supply, before the motor.

[3dgeo]

No, those capacitors will make things worse if anything, especially the massive 2200μF one! They will cause a huge surge, every time the MOSFET is turned on, which is probably killing it,

Those capacitor shouldn't be in parallel with the load but the power supply, before the motor.

Can You please explain more? Than why original board was made this way?
You mean I should put capacitor like this:

[james_s]

That capacitor looks like it's just a bulk filter on the main supply line, I'm not sure why it's drawn the way it is.

Do you have an oscilloscope? If you're driving an inductive load (motor) with a PWM signal you end up with a boost converter. You could easily have spikes of 2-3 times the supply voltage or more, you should use mosfets with a high enough voltage rating to tolerate this, as well as you might include a snubber.

[Mechatrommer]

as some of You might know..

no, i dont know. where is your load in the schematics and how its connected?

[David Hess]

We need to see the correct schematic.  As shown, the capacitors produce an AC short across the MOSFET.  If it was not failing, then something would be wrong.

Usually I would expect an inductor somewhere.

[Zero999]

No, those capacitors will make things worse if anything, especially the massive 2200μF one! They will cause a huge surge, every time the MOSFET is turned on, which is probably killing it,

Those capacitor shouldn't be in parallel with the load but the power supply, before the motor.

Can You please explain more? Than why original board was made this way?

Did the original board do PWM with MOSFETs?

If not, then there's your answer: the capacitors just smooth the power supply voltage.

If the answer is yes: it did do PWM with MOSFETs and big capacitors, connected across the motor, then it was poorly designed to start with.

You mean I should put capacitor like this:

Yes, that's the correct way to connect the capacitor.

If the motor is being controlled with PWM, the only capacitors allowed across the motor are tiny ≤1nF RFI filter capacitors, from the motor's terminals to its case and there should be some inductors, before the capacitors too. When a capacitor is suddenly connected to the supply voltage, it draws a huge current spike, which is most likely destroying the MOSFETs.

[3dgeo]

Did the original board do PWM with MOSFETs?

Yes, but only bottom one, it was at 100hz (5V), I measured it, it's solid info. The circuitry in a first picture on the left is exact copy what was made in original circuit, tho I didn't add PWM circuit cos I think it's irrelevant in this situation.

Now I'm making 3 IRF3710 fets in parallel and will screw them to aluminum plate, hope to get good thermals so I can inspect whats going on in the circuit with my crappy oscilloscope...

[3dgeo]

We need to see the correct schematic.  As shown, the capacitors produce an AC short across the MOSFET.  If it was not failing, then something would be wrong.

Usually I would expect an inductor somewhere.

What do you mean correct schematic? Old one on the left is correct, and it did work.
I updated main image with CD motors included. Motors are the inductors...

My frankenstein with 3 parallel mosfets is complete, lets see whats happens...

[SiliconWizard]

I used mosfet driver: TC4420 http://ww1.microchip.com/downloads/en/DeviceDoc/21419D.pdf
But probably TC4420 didn't get enough current cos I was feeding 10V with voltage divider from 50V.
And I doubt mosfets dies due to overheating cos they were dead cold (newones and original ones).
The only one  mosfet that got crazy hot (IRF3710) did survived, but heat was too hight to keep it on for a long time, and it did PWM it to 24V.

Edit: I just realized that TC4420 drive capability: 10nF, is it still suitable to drive that mosfet at low frequency, like 100hz? (original mosfets were driven only at 100hz)
Hm, I still can't believe they got killed due to this, they were cold, not even a bit warm, and original mosfets were driven from 5V with no driver and they died as well...

The TC4420 wouldn't meet the requirements for driving the biggest MOSFET you chose which has a ~20 nF input capacitance, but for the former MOSFET you tried (which I think has ~3 nF IC), it would be fine.

You do have to power it properly, since it has to provide the required current for charging the gate at each switching edge. Powering it from a simple voltage divider (which you don't say what resistor values you used, but if too high, that may actually even be worse than with no gate driver at all.) So first thing, I would get this gate driver thing sorted before going further.

Another point, what are the specs for D9?
And what's the actual max current draw of your motors?

[3dgeo]

The TC4420 wouldn't meet the requirements for driving the biggest MOSFET you chose which has a ~20 nF input capacitance, but for the former MOSFET you tried (which I think has ~3 nF IC), it would be fine.

You do have to power it properly, since it has to provide the required current for charging the gate at each switching edge. Powering it from a simple voltage divider (which you don't say what resistor values you used, but if too high, that may actually even be worse than with no gate driver at all.) So first thing, I would get this gate driver thing sorted before going further.

Another point, what are the specs for D9?
And what's the actual max current draw of your motors?

I really appreciate You want to help, but it's not a mosfet driving issue...
D9 is 100V 3A SR2100, in original circuit D12 and D19 are 60V 3A B360B

BTW, I killed those 3 parallel IRF3710 mosfets, looks like while just turning off and on they stay alive, but when PWM they die... Tho they didn't get hot this time.
I was driving them with proper signal from mosfet driver which this time got enough power due to me using dedicated high current power supply at 11V.
Back to the drawing board...
It's seems like those 2 extra mosfets in old circuit was doing something I'm missing now, it's kinda strange why you need them when you have relays, but maybe their purpose was to reduce/isolate voltage spikes between 2 motors. Tho I'm assuming to kill 3 mosfets in parallel you need very hight voltage (as I understand 3 fets should handle 3x more voltage and current?). I'm deep in the woods at this point, tho I have plenty mosfets left 

[james_s]

I'm almost positive you're killing the mosfets due to high voltage spikes. If you put a scope across the motor you'll likely see a very large spike with some ringing. Have a look at this for some relevant information.

http://www.masinaelectrica.com/dc-motor-speed-controller-pwm-0-100-400hz-3khz-freq/

[3dgeo]

I'm almost positive you're killing the mosfets due to high voltage spikes. If you put a scope across the motor you'll likely see a very large spike with some ringing. Have a look at this for some relevant information.

http://www.masinaelectrica.com/dc-motor-speed-controller-pwm-0-100-400hz-3khz-freq/

I don't have good scope, I only have DS0138 and I have practically no idea how to use it. Link looks promising, thank you.
P.S. I updated 1st post with old PCB traces, hope that will help to help me

[Zero999]

That should do the job.

Double check that the 2200µF capacitor is wired, as per your previous schematic and if it is, try removing it. Adding a filter capacitor, with no inductor, to a PWM circuit totally defeats the object of PWM and causes excess power dissipation in the switch.

[Mechatrommer]

you may try adding another diode (big schottky?) close to mosfet's drain to gnd (parallel to body diode). if that works meaning your mosfet was taking bad negative spike from relay switching. and i think 2200uF capacitor should be connected to gnd, not to mosfet's drain, your way the mosfet will be taking high current transient when turned on as others mentioned. the ceramic caps usually found near the motor terminals, not near mosfet, ymmv.

[Zero999]

you may try adding another diode (big schottky?) close to mosfet's drain to gnd (parallel to body diode). if that works meaning your mosfet was taking bad negative spike from relay switching.

How will that help? The MOSFET's body diode never conducts, when it's used as a low side switch.

[Mechatrommer]

you may try adding another diode (big schottky?) close to mosfet's drain to gnd (parallel to body diode). if that works meaning your mosfet was taking bad negative spike from relay switching.

How will that help? The MOSFET's body diode never conducts, when it's used as a low side switch.

negative spikes... i added my previous msg, 2200uF should not be connected to drain (normally), usually its to the gnd...

[Mechatrommer]

try to look at it, i dont see any supply path when negative spike occurs during relay switching other than the body diode, anyway if that doesnt help, he can simply remove the "auxiliary" body diode forever, just want to give idea on another troubleshooting method. fwiw..

[Zero999]

try to look at it, i dont see any supply path when negative spike occurs during relay switching other than the body diode, anyway if that doesnt help, he can simply remove the "auxiliary" body diode forever, just want to give idea on another troubleshooting method. fwiw..

There is no negative spike. The motor is connected to the positive supply and when it is turned off, it produces a positive spike which is shorted to +V, via the free-wheeling diode. The MOSFET's body diode doesn't conduct.

[james_s]

You don't even need a *good* scope, nearly any scope would do for this task and it would be an excellent opportunity to learn how to use it. Without a scope you're literally flying blind.

[Zero999]

Attached is a simulation of a coil being driven from a MOSFET with a 60V diode in parallel. Note how the voltage when the MOSFET is turned off is positive, not negative and if the body diode conducted, it would be a much lower voltage.

[3dgeo]

Wow, thank You all for participating!

I had enough dead mosfets, so I thought I just solder two wires and short them in place of a mosfet to test what happens, sadly I couldn't measure voltages while shorting mosfets wire due to my oscilloscope or my knowledge of using it limitations, but I managed to measure what happens when relays gets on while mosfet is on (wires shorted) and it seems like voltage spikes from motors themselves isn't an issue (at least while mosfet is not PWMed), picture below.
I'm also adding image drain to source wires while relays is on.
And I should point out that while shorting out "mosfet wires" I did get sparks, so few people that already said that CAP is the issue were right.
I didn't understood why before, but thanks to You all I developed some new neurons paths and now I'm "beginning to believe"  

[3dgeo]

Well, I resoldered that CAP to ground and it seems to work just fine now, I didn't get any significant voltages spikes in any direction while driving at 32khz PWM, and voltages at 24-26V seems ok, is there are any components I should add to improve this circuit? Should I order proper driver and use one IRFP4468 or go with two IRF3710 (I think one will get too hot)?
If PWM at 120hz motors whistle, but they didn't with old PCB (on 100hz), old PCB is still a mystery, I think mosfets they use were somewhat special, I can't find any information about those parts...

[Zero999]

Well, I resoldered that CAP to ground and it seems to work just fine now, I didn't get any significant voltages spikes in any direction while driving at 32khz PWM, and voltages at 24-26V seems ok, is there are any components I should add to improve this circuit? Should I order proper driver and use one IRFP4468 or go with two IRF3710 (I think one will get too hot)?
If PWM at 120hz motors whistle, but they didn't with old PCB (on 100hz), old PCB is still a mystery, I think mosfets they use were somewhat special, I can't find any information about those parts...

Good, I'm glad you got it working. If it's not overheating and is working properly then why change it?  Using lower resistance MOSFETs and faster drivers won't necessarily gain you anything, if they're not needed.