High current MOSFET on solenoid driver shorts when power is applied

[canadaboy25]

I have designed a driver circuit for a piece of agriculture equipment that has a high current electromagnetic solenoid.  The driver utilizes an ESP8266 MCU and a small app on a smartphone which lets the user set the period and interval of the solenoid (ie. 1 second on, 30 seconds off).  The driver is powered from a 12V lead acid battery in a tractor.

The driver works great and has operated trouble free for many hours on end.  The problem I am having is when applying power, there is a chance that the IRFZ44Z MOSFET goes short circuit and the solenoid slams open.  Replacing the MOSFET repairs the driver.  If the MOSFET does not short circuit immediately upon applying power, the driver works trouble free.

After shorting several MOSFETs, I think I have narrowed down the cause to the order of connecting the positive and negative connections.  The driver is connected via a heavy duty power connector (see picture).  If I kind of "twist" the power connector while plugging it in to "favor" the negative terminal and have it make contact first, it seems the MOSFET does not short.  This may be a complete fluke as I cannot see how this should make any difference, as the rest of the circuit is isolated from any grounds on the machine.

Upon drawing out the schematic, I realized that I have connected my flyback diode incorrectly.     Obviously the anode of the diode should be connected to the drain of the MOSFET.  However, I do not believe this is causing my MOSFET to short as the driver works perfectly fine indefinitely as long as it survives the initial power-up.  The MOSFET is also not being overloaded and does not even get warm after hours of normal operation.

Any ideas what is causing the MOSFET to intermittently short circuit only upon applying power?

[inse]

The solenoid needs a freewheeling diode or a clamping circuit to protect the MOSFET from inductive kickback at turn-off.
I also would reverse the control logic to a pull-down resistor at the gate and the driver transistor to the high side. You will need another transistor for level shifting. Also include gate protection to limit gate voltage to the specified value.

[rteodor]

Upon drawing out the schematic, I realized that I have connected my flyback diode incorrectly.     Obviously the anode of the diode should be connected to the drain of the MOSFET.  However, I do not believe this is causing my MOSFET to short as the driver works perfectly fine indefinitely as long as it survives the initial power-up.

Yup, if it would be the diode the the damage would occur when MOSFET is turned off. If MOSFET gets damaged at turn on then I would look for high inrush current. You also show those Anderson connectors there and that makes me think high current.

The datasheet of IRFZ44 I have, says 49A at 25C with a max pulse of 160A. Is it possible that those currents are exceeded ?

[canadaboy25]

Yup, if it would be the diode the the damage would occur when MOSFET is turned off. If MOSFET gets damaged at turn on then I would look for high inrush current. You also show those Anderson connectors there and that makes me think high current.

The datasheet of IRFZ44 I have, says 49A at 25C with a max pulse of 160A. Is it possible that those currents are exceeded ?

I have not measured the inrush current of the solenoid, however I do not believe that it is exceeding 160A.  I know the on-current is well below 49A.  Perhaps I will measure the currents just to double check.

However, the driver has been used with the parameters of 0.25 seconds on, and 15 seconds off for many tens of hours on-end.  I would think that if the currents were being exceeded, the MOSFET would have shorted at some point during operation.  I have never had a failure during operation, only immediately upon power up.

Also, with the R1 pullup resistor, the default state of Q1 is on, turning Q2 off.  So the solenoid should not even be activating upon power up.  And this is exactly what happens on a successful power-up, the solenoid stays off.

The only thing I can think of is if it takes some small amount of time for the DC-DC converter to start and turn on Q1, that Q2 may try to conduct briefly.  I still don't see why this would cause it to short.

[canadaboy25]

The solenoid needs a freewheeling diode or a clamping circuit to protect the MOSFET from inductive kickback at turn-off.
I also would reverse the control logic to a pull-down resistor at the gate and the driver transistor to the high side. You will need another transistor for level shifting. Also include gate protection to limit gate voltage to the specified value.

Yes, as mentioned, I realized that I soldered up my flyback diode incorrectly upon drawing this schematic.

I agree, I would've preferred a pulldown on the gate, but then as you said, I would need additional circuitry to pull the gate up high enough to turn on the MOSFET with the 3.3V MCU.  I was trying to do this with minimal components, but I may have to go that route.

I would still like to find the underlying cause for the shorting MOSFET in the current configuration

[PCB.Wiz]

I see no caps shown, and hot plugging can cause significant spikes with cable inductance.
You may like to try some serious transient clamping, and pick a newer fet with more headroom, and that is rather feeble gate drive for a high current load.
Use a proper gate driver.

[inse]

There is a small delay between 12V and the 3.3V upon power-on which might (!) have been sufficient to shortly turn on the MOSFET.
Destroying it by inductive kickback when it got turned off.
You found the answer yourselves before I could read it 😁
But you need gate protection for this inductive switching (or go for a more robust IGBT?)

[canadaboy25]

Later edit (didn't wash my eye properly):

After shorting several MOSFETs, I think I have narrowed down the cause to the order of connecting the positive and negative connections.  The driver is connected via a heavy duty power connector (see picture).  If I kind of "twist" the power connector while plugging it in to "favor" the negative terminal and have it make contact first, it seems the MOSFET does not short.  This may be a complete fluke as I cannot see how this should make any difference, as the rest of the circuit is isolated from any grounds on the machine.

What do you mean ? Do you disconnect it and reconnect it while the MOSFET is on ?
From what little I know of tractors, its the positive of the battery that is connected to the chassis.

Sorry, I did not make this very clear as I could not see how it could possibly be related.

The power connection was always made with the solenoid off and the control set to off.

It seems as though if I make an effort to connect the negative terminal of the connector before the positive terminal, the MOSFET does not short.  When just plugging the connector normally (random chance of which pin in the connector makes contact first), I was shorting a MOSFET 50% of the time.  After making an effort to ensure that the negative terminal of the connector makes contact first, I was only getting a shorted MOSFET ~5% of the time.

The battery of a tractor (manufactured later than ~1960) has the negative terminal connected to the chassis.  Since the rest of the circuit is isolated from the chassis, this order of connections should have no affect and is likely just a coincidence.

[canadaboy25]

I see no caps shown, and hot plugging can cause significant spikes with cable inductance.
You may like to try some serious transient clamping, and pick a newer fet with more headroom, and that is rather feeble gate drive for a high current load.
Use a proper gate driver.

I agree with your points.  This was a quick driver thrown together with spare parts I had on hand.  Not even close to a production design.

There is a small delay between 12V and the 3.3V upon power-on which might (!) have been sufficient to shortly turn on the MOSFET.
Destroying it by inductive kickback when it got turned off.
You found the answer yourselves before I could read it 😁
But you need gate protection for this inductive switching (or go for a more robust IGBT?)

This most likely seems to be the culprit.  Fixing my flyback diode *should* save the MOSFET, but I will also refactor the gate drive to not be affected by the 3.3V supply delay, and add a gate protection zener diode.

Thanks to all for the input, and I will report back with the results of the modifications.

[Ian.M]

When you connect power, there is a small inrush, due to the input capacitance of the DC-DC converter.  This may ring with the supply lead inductance and transiently over-voltages the MOSFET gate via its pullup resistor till the increasing current through the load inductance damps it.   Also intermittent contact as you make the connection causes the MOSFET turn on and off rapidly, probably far more frequently than it normally does, possibly exceeding its repetitive avalanche energy rating and/or SOA due to the weak pulsing gate drive keeping it in its linear region too long.

Clamp the gate with an 18V 5% Zener across Q1 and also fix the free-wheeling diode and it should stop killing MOSFETs. 

However I believe Inse has a good point - the MOSFET should be normally off with a gate pulldown, and the MCU should actively turn it on via a high side driver or a proper bidirectional gate driver, so it doesn't pulse on before the 3.3V comes up and the MCU boots.

[johansen]

The inductance of the supply needs to be bypassed by a capacitor.. otherwise it doesnt matter if you have a diode to clamp the inductor.. you still have an inductor in the circuit.

The diode, capacitor, mosfet have to have a small loop area. The capacitor has to be able to handle the current flowing in the inductor at turn off, and the energy stored in the supply inductance.

[SteveThackery]

I would still like to find the underlying cause for the shorting MOSFET in the current configuration

I'm puzzled: you might have already found the problem. The anode of the freewheel diode should be connected to the bottom of the solenoid, not ground. If you fix that you might well fix the problem. Or have you done that already?

[WattsThat]

It’s worth a mention that the IRFZ44Z is a highly counterfeited part. Anything sourced from eBay, Aliexpress or similar is guaranteed to be at best off spec and at worst, remarked e-waste.

[canadaboy25]

When you connect power, there is a small inrush, due to the input capacitance of the DC-DC converter.  This may ring with the supply lead inductance and transiently over-voltages the MOSFET gate via its pullup resistor till the increasing current through the load inductance damps it.   Also intermittent contact as you make the connection causes the MOSFET turn on and off rapidly, probably far more frequently than it normally does, possibly exceeding its repetitive avalanche energy rating and/or SOA due to the weak pulsing gate drive keeping it in its linear region too long.

Clamp the gate with an 18V 5% Zener across Q1 and also fix the free-wheeling diode and it should stop killing MOSFETs. 

However I believe Inse has a good point - the MOSFET should be normally off with a gate pulldown, and the MCU should actively turn it on via a high side driver or a proper bidirectional gate driver, so it doesn't pulse on before the 3.3V comes up and the MCU boots.

Thanks for the detailed analysis and explanation.  The MOSFET being damaged through the gate makes the most sense to me as it would finally explain why the driver is able to switch the solenoid indefinitely once powered up, and only fail upon applying power.

I will add the protection components and also redo the drive design.  This was thrown together quickly with spare parts as I needed it right away, but I now have time to redesign it.

The inductance of the supply needs to be bypassed by a capacitor.. otherwise it doesnt matter if you have a diode to clamp the inductor.. you still have an inductor in the circuit.

The diode, capacitor, mosfet have to have a small loop area. The capacitor has to be able to handle the current flowing in the inductor at turn off, and the energy stored in the supply inductance.

I'm afraid I'm somewhat out of my depth in regards to dealing with the power lead inductance.  Would you be able to elaborate some more?  The circuit is connected by roughly 20ft of 14AWG cable to the 12V battery

I'm puzzled: you might have already found the problem. The anode of the freewheel diode should be connected to the bottom of the solenoid, not ground. If you fix that you might well fix the problem. Or have you done that already?

I realized the flyback diode was connected incorrectly while making this post, but I do not believe it was the main cause of my issues.  I think Ian.M's idea of the MOSFET being damaged through the gate is most probably what was causing the failures.

It’s worth a mention that the IRFZ44Z is a highly counterfeited part. Anything sourced from eBay, Aliexpress or similar is guaranteed to be at best off spec and at worst, remarked e-waste.

IRFZ44Z's were sourced from Digikey.  I refuse to buy the crappy, cheap junk on ebay unless I have no choice.

[johansen]

The inductance of the supply needs to be bypassed by a capacitor.. otherwise it doesnt matter if you have a diode to clamp the inductor.. you still have an inductor in the circuit.

The diode, capacitor, mosfet have to have a small loop area. The capacitor has to be able to handle the current flowing in the inductor at turn off, and the energy stored in the supply inductance.

I'm afraid I'm somewhat out of my depth in regards to dealing with the power lead inductance.  Would you be able to elaborate some more?  The circuit is connected by roughly 20ft of 14AWG cable to the 12V battery

Solder a 1000uf 25v electrolytic cap and a 1 uf film cap as close as possible to the mosfet, flyback diode, As in those three components within 1 inch of each other.

https://electronics.stackexchange.com/questions/167484/electrolytic-capacitor-in-solenoid-circuit

[canadaboy25]

Solder a 1000uf 25v electrolytic cap and a 1 uf film cap as close as possible to the mosfet, flyback diode, As in those three components within 1 inch of each other.

https://electronics.stackexchange.com/questions/167484/electrolytic-capacitor-in-solenoid-circuit

Thank you for the link.  I was greatly overthinking your response but it is very clear now.  I will add that to the list of modifications.

[canadaboy25]

I have reconnected my flyback diode to the correct location, added an 18V zener on the gate of the MOSFET, and added some more bulk capacitance to the board.

The circuit performed flawlessly throughout the entire season with no more shorted transistors.  I am quite certain it was the zener on the gate which did the trick.

Thanks to all for the advice.