MOSFET burning up but similar circuit on another board is fine

[zacaj]

Is the original circuit made on a compact printed circuit board but you made yours on a solderless breadboard with messy wires all over the place?

I made it on a PCB and it didn't work. I then also made it on a breadboard for easy swapping of fets and it was no different.

What is the voltage across the MOSFET when it is turned on (and in the process of "burning up". I would attach a heatsink to the MOSFET for this test.

across the mosfet? not sure what you mean. Where should I put the DMM leads?

Is the solenoid actuating?

Yes. It works properly.

What is the actual Vgs when the MOSFET is turned on (no need for a load for this test)

4.94V

What is the actual Vgs when the MOSFET is turned off.

Bounces around a bit but <100mV

Does the MOSFET get hot when when the solenoid is supposed to be turned off?

No. Or at least very slowly if it does. I let it sit for about a minute and didn't feel any heat.

What frequency is the MOSFET being switched at? Or is the MOSFET turned on and off for long periods (perhaps for periods not less than 100ms)?

I'm turning it on for 50ms and then off for 1000ms repeatedly right now

 

[Harald Kapp]

across the mosfet? not sure what you mean. Where should I put the DMM leads?

Drain-Source

Before burning up further MOSFETS you could use a lighter load (e.g. 1 kΩ) to test the voltages around the MOSFET as asked. You may also, if you have a scopa at hand, take readings of the voltage waveforms (Vgs, Vds) at turn-on, turn-off and while the MOSFET is on or off respectively. DC measusrements are one thing, AC measurements may show (or may not) ringing, voltage spikes etc. which may cause damage.

 

[Alec_t]

If the circuit layout differs in any way from the ok one, then perhaps the FET goes into spurious oscillation. That could account for excessive heating.

 

[zacaj]

You may also, if you have a scopa at hand, take readings of the voltage waveforms (Vgs, Vds) at turn-on, turn-off and while the MOSFET is on or off respectively. DC measusrements are one thing, AC measurements may show (or may not) ringing, voltage spikes etc. which may cause damage.

I only have an old analog one available. I hooked it up and took some readings but I don't have any real way to capture them. Everything seemed to be fairly digital, no notable spikes or waves or anything on Vgs, Vds, etc.

 

[(*steve*)]

OK, what about the voltage across the mosfet (drain to source) when the load is switched on?

 

[zacaj]

OK, what about the voltage across the mosfet (drain to source) when the load is switched on?

Sorry I forgot about that one... 54.7VDC with a resistor attached from my ~50V rail to the drain, or about 20mV with nothing attached to the drain besides my DMM.

 

[(*steve*)]

Sorry I forgot about that one... 54.7VDC with a resistor attached from my ~50V rail to the drain, or about 20mV with nothing attached to the drain besides my DMM.

OK, maybe you don't understand.

Firstly the voltage will be between zero and your power supply voltage (so is your power supply 54.7V?). Under load this may drop, but I'm not too concerned about that)

Secondly, this must be done under load, so connect up the solenoid. If you don't already have a heatsink on the mosfet, add one -- it will keep the mosfet running a little longer before failing.

You also need to tell me what current is drawn by the solenoid. By your figures, it draws more than 20A!

From the datasheet, you should expect about 2V Vgs at Id of 2A. This points to a dissipation of 40W which will require a fairly large heatsink to dissipate (you do have a heatsink on it, right?).

 

[zacaj]

Firstly the voltage will be between zero and your power supply voltage (so is your power supply 54.7V?). Under load this may drop, but I'm not too concerned about that)

The power supply on the solenoid jumps around a bit depending on load, capacitors, etc. It's generally around 50V.

Secondly, this must be done under load, so connect up the solenoid. If you don't already have a heatsink on the mosfet, add one -- it will keep the mosfet running a little longer before failing.

Even with a heatsink, the mosfet will fry within 10-20 seconds if I just lock it on, and I'll probably damage the coil as well. I don't really trust my meter do even out at a voltage before something starts burning. Is the resistor not enough?

(you do have a heatsink on it, right?).

I don't have one on it. The other boards I'm basing mine off of don't have any heatsink either, and they're just fine.

OK, maybe you don't understand.

Those readings seem weird to me too, although my understanding of mosfets isn't the best I'll admit, however that's the voltage between the drain and source according to my DMM. I would have assumed the voltage across the mosfet would be low, since when it's on it should have a very low resistance?

 

[(*steve*)]

Even with a heatsink, the mosfet will fry within 10-20 seconds if I just lock it on, and I'll probably damage the coil as well. I don't really trust my meter do even out at a voltage before something starts burning. Is the resistor not enough?

OK, if the coils could get damaged then they sound like they're either not rated for this voltage, or not rated for continuous use -- which is it?

Do the SAME coils work with the different driver?

You need to measure the voltage across the mosfet with the real load to determine the actual voltage drop. This, along with the actual current drawn by the load allows you to determine the power dissipated by the mosfet.

This particular mosfet, without a substantial heatsink, is not capable of switching 20A for any significant period of time.

If the coils really are 2Ω, then you can replace the the coil with a 2Ω 2000W (yes, two thousand watt) resistor. (OK, for a brief test, you could probably get away with this 600W resistor).

You REALLY need to confirm details about these solenoids. If they really are consuming 2000W of energy when energised, then that's some big solenoid.

 

[zacaj]

OK, if the coils could get damaged then they sound like they're either not rated for this voltage, or not rated for continuous use -- which is it?

They're not for continuous use, but it's hard to get a voltage reading without having them on for longer than they're designed for.

Do the SAME coils work with the different driver?

Yes

You REALLY need to confirm details about these solenoids. If they really are consuming 2000W of energy when energised, then that's some big solenoid.

They're not that big. maybe an inch in diameter and and inch and a half long. The amperage/wattage numbers you get from the normal calculations just seem crazy to me, but there's no question that it measures 2 ohms and that I've got 50V hooked up to it. With my elementary knowledge of mosfets it doesn't seem like the IRL540N should be able to drive them that well either (especially without a heatsink), and yet it does it just fine on these other boards, hence my confusion

 

[(*steve*)]

Even switched for 100ms every couple of seconds, I'd be employing a gate driver to reduce the switching and to ensure that I gate the gate as close as possible to +5V.

Your circuit will probably switch quite slowly, and the peak power will be very high -- this may not result in failure due to heating, but definitely isn't good for the mosfet.

if switched for 100ms every second, the energy dissipated would be...

0.1 * 2 * 20 + (2.2 E-9 / 0.011) * 50 * 20 = 4J (the extra stuff confirms the switching speed is not an issue)

so we're looking at an average dissipation of 4W.

With no heatsink, the junction temperature (Tj) will reach ta + Tja * 4 degrees C = 25 + 62 * 4 = 25 + 248 = 273C.

The absolute maximum Tj is 175C, so that's a reason why it's smoking. If it smokes even with a heatsink then something else is wrong. Even a 10 degC/W heatsink would keep Tj under 70C, which would be fine (to keep it below 150C, even a 30 degC/W heatsink would suffice (and that's tiny).

To stop it smoking, either a heatsink, or an average dissipation of 2W (or less) is required. This could be achieved (for a 10% duty cycle) with a load drawing less than 15A, assuming you can drive the gate to achieve the rated Rgs of 0.077Ω.

I suspect the other mosfets have an Rds(on) of less than 0.077Ω, or the duty cycle is lower.

A device with Rds(on) of 0.007Ω will have an average dissipation of 280mW -- so it would hardly be warm.

 

[zacaj]

Your circuit will probably switch quite slowly

What would you say is the cause of that? Or what could I do about it? When digital stuff turns analog is where I start to lose my basis for stuff...

I suspect the other mosfets have an Rds(on) of less than 0.077Ω, or the duty cycle is lower.

Other mosfets?

A device with Rds(on) of 0.007Ω will have an average dissipation of 280mW -- so it would hardly be warm.

Where did the 0.007 come from here?

 

[(*steve*)]

What would you say is the cause of that? Or what could I do about it? When digital stuff turns analog is where I start to lose my basis for stuff...

Don't worry too much about that. My calculations suggest this won't contribute too much.

Other mosfets?

Those on working boards

Where did the 0.007 come from here?

It's an Rds(on) of some mosfets which have a low Rds(on).

Reducing the duty cycle (say turning the solenoids on for 100ms every 3 seconds) would also help. (you want the duty cycle to be at 5% or lower, assuming nothing is wrong with your circuit and the load is as you suggest).

 

[zacaj]

Those on working boards

But they use the same mosfets?

Reducing the duty cycle (say turning the solenoids on for 100ms every 3 seconds) would also help. (you want the duty cycle to be at 5% or lower, assuming nothing is wrong with your circuit and the load is as you suggest).

The problem is, with pinball I can't really rely on a specific duty cycle. A bumper could get fired 20 times in five seconds, but then won't fire at all for 30 seconds. Really I can't guarantee more than probably a 25% duty cycle during the short bursts (assuming the amount of time for a ball to bounce back and forth quickly between two bumpers is around 200ms with 50ms pulse). The fact that I'm anywhere close to burning them up with this light of a load concerns me. If I was having occasional problems then I would just put a heatsink on and live with it but the fact that it blows so easily means that I need to do something drastic. Since it seems to act so much worse than equivalent other boards, I feel like there must be some specific problem to fix, which will give a drastic improvement once I can find it. These other boards can be run for hundreds of thousands of hours without issue, so trying to edge mine 'into spec' won't really cut it.

 

[(*steve*)]

But they use the same mosfets?

Then there is something different or your specifications of the coil are inaccurate.

Please check your solenoid here. Is it there? What type is it?

 

[zacaj]

Then there is something different or your specifications of the coil are inaccurate.

Please check your solenoid here. Is it there? What type is it?

A-22-550. Even if the specifications were wrong though, I measured the resistance of it directly, so I don't see why that would cause a problem

 

[(*steve*)]

As far as I can tell, that coil is used for things used rarely (ball ejectors), not flippers.

But hey, the answers I've given you are based on the specs of the mosfet.

It is also possible that your coil is partially shorted. This would increase the current through it, and for DC use, may not affect the pull too much -- but it, and the driver, will run hot.

 

[Audioguru]

Did you measure the resistance of the solenoid coil plus the resistance of your meter's leads?

 

[BobK]

From the info you gave us:

Coil resistance 2 Ohms.
Voltage 55

We can calculate the current

55 / 2 = 27.5A

The Rdson is 0.077V so we can calculate the drain to source voltage:

Vds = 0.077 * 27.5 = 2.12V

The power when fully on is:

2.12 * 27.5 = 58.23 W

And with on time of 100ms each second, the duty cycle is 10%

So the average power is 5.8W

This is too much power for the device to survive with no heat sink.


So the mystery is:

Why is the commercial board not overheating and destroying the transistors.

My theory: It limits the on time to something less than 100ms,

Bob

 

[Audioguru]

The replacement IRL540L has a maximum on-resistance of 0.077 ohms when its Vgs is 5V. But the original STP22NE10l is worse at 0.1 ohms. But the original circuit might have used Mosfets selected (or survived) to be the best ones with a much lower on resistance.