Why are my mosfets failing?

[Szewczykm]

Clearly I'm doing something wrong because I've built this basic circuit and my mosfets are failing and locking ON.  I'm not sure why thought.

Application:  I want to be able to control a flipper on a pinball machine using a microcontroller.  This circuit is just a concept for the final design.

The pinball machine has DC power supplies for 5v, 12v, 24v, 48v.

The flipper coil circuit is very simple.  One side of the coil is connected to 48v, a button is pressed which closes a switch giving the current a path to ground and that actuates the flipper.  You can take the "ground side" of the coil connect a test lead to it and touch it to ground and it will fire.

Attached is a picture of what I'm trying.  I am connecting the Drain of an NPN mosfet (http://www.mouser.com/ds/2/149/RFP12N10L-100355.pdf) to the "ground side" of the coil.  The Source is connected to ground.  I have a wire going to the 5v power supply, through a switch, connected to the gate and a 10k pull-down resistor connected between the gate and ground.  Here is a sketched schematic.

I'm finding that with this setup, when I press the button to send 5v to the mosfet, the coil will fire and lock on.  After the first firing, the mosfet is damaged and the connection between Source and Drain is a short.

It seems like everything on the mosfet is within spec, but I'm damaging it and it locks on.  Is it that I need a resistor limiting the current to the gate?  Is that where I'm causing the damage?

[ElektronikLabor]

Which diode are you using parallel to the solenoid?
You need an fast Schottky diode for that, because normal pn-diodes could be too slow.

[jlmoon]

Diode blocking performance, Med... I think you hit the nail on the hammer.

[c4757p]

Which diode are you using parallel to the solenoid?
You need an fast Schottky diode for that, because normal pn-diodes could be too slow.

Normal PN diodes will turn off slowly. They'll still turn on nice and fast...

[free_electron]

there is a ground wire missing from the 5 volt supply to the bottom of the coil.... i hope it is there and not somewhere else ...

[Szewczykm]

Which diode are you using parallel to the solenoid?
You need an fast Schottky diode for that, because normal pn-diodes could be too slow.

There are regular PN diodes, which is typical on pinball coils.  I don't think I've seen a schottky diode on any modern or classic pinball machine they're always regular PN diodes.  For years they used TIP122 (Or TIP102) darlington transistors to control the coils, are those less sensitive than the mosfet I'm using?

So there's no problem with me not limiting current to the Gate?

there is a ground wire missing from the 5 volt supply to the bottom of the coil.... i hope it is there and not somewhere else ...

It's there, I should have scribbled it in.  I get continuity to the same ground on the 50v power supply and on the 5V.

[Szewczykm]

I am considering using relays but the mosfets would be easier, cheaper, and faster.

[bktemp]

Try adding a 6.8V Z-diode across gate-source. It is only rated for +/-10V and there is only a high impedance connection from gate to ground.
A spike on the drain can couple easily into the gate and force the voltage above +/-10V.

[ElektronikLabor]

Normal PN diodes will turn off slowly. They'll still turn on nice and fast...

You're right, my fault; I mixed it up 

[Seekonk]

There is no reason this has to be fast.  Consider putting a .1uF in parallel with the gate resistor.  The failure may be due to long leads picking up a pulse and destroying the gate.  As said before a zener would also do this but a cap is cheaper.  In theory what you have should work.  I suspect that it is how it is wired that is causing the problem.  Also wonder how clean that 5V power supply is.  Since these seem to blow so fast, add the capacitor across the gate resistor and a 1000 ohm between the gate and the switched source as spike filtering.  This will at least give a clue to whether this is a gate problem.

[Szewczykm]

I put the cap on like you suggested and it's still blowing the mosfet.  I am learning things.  Like how to capture something on my scope and save it, etc.  Neat stuff, but I blew through my mosfets.  Here is a capture, clearly it's telling me something but I don't know what. 

Some things I did find out.  The part I'm trying to use (RFP12N10L) did fail without a current limiting resistor going to the gate, even with no load.  I put a resistor on the gate and it stopped failing.

I tried a different part number and FQP20N06L and that also failed immediately.  Then I moved the ground connection from the ground braid in the cabinet, to a ground wire going directly to the 5v power supply.  This stopped the new part from failing immediately, but it did fail after about 7 on/off cycles.  I imagine that's from running 72 volts through the 60v mosfet.

I was encouraged so I went back to the RFO12N10L because it's rated for 100v.  But that part failed immediately.  I did set up my scope and get a capture.

The blue is on the drain of the mosfet.  10v per div.
The yellow is on the gate.  2v per div.

I think that scamble in the middle is the part failing.  I'm out of ideas but I'm learning...

If you've got any more suggestions I'd appreciate it.

I've looked at similar circuits (solenoid driven by mosfet) and the circuits are identical to mine.  It might be wiring, because of the ground coming from different power supplies.  Or I might be using a part out of spec, but not from what I can tell.  Also, the capture isn't showing any ridiculous spikes happening at the gate.... still struggling.  Thanks for any further input.

Mike

[T3sl4co1l]

I haven't seen pictures of layout but this thing smacks of excessive lead length... put 1/4 wave resonators on gate and drain and you're sure to get a power oscillator somewhere.

The diode across the solenoid has to be referenced to the MOSFET.  The coil doesn't give a damn what's applied to it.  The diode is there to protect the MOSFET.  It needs to be wired at the MOSFET drain, and if the power supply is more than a few inches from that, you need a bypass capacitor from source to +V to return the diode to.

An R+C damper may come in handy to keep things stable.  10 ohm + 0.01uF may be a good starting point.  Drain to source.

Likewise, the gate needs a series resistor to charge it slowly, and faster turn-off probably wouldn't hurt either.

Tim

[Seekonk]

One thing I noticed was your 100V FET has a maximum voltage of 10V.  I had never seen that before, every FET I had looked at was 20V max.  I was going to suggest using 330 ohm gate resistors (only 15ma) to shunt out unwanted capacitive lead pickup.   Then I saw your scope image.

Reminded me of an amplifier kit someone had me fix in college that was really distorted.  Didn't have a signal generator so I used a FM tuner.  Noticed that the signal strength meter was getting pegged, The amp was a transmitter.  Looked at the circuit boards outside ground path, It was perfect to be a tuned circuit.  Moved the ground connection three inches and the amp was fine.

Your wiring is doing the same thing.  The transistors before did about the same thing but they took more drive and it dampened quickly.  The series resistor dampened it a bunch,  but it still oscillated abd eventually overheated the FET.    You need a better current path in the ground. The noise is feeding back through the 5V.  That is something you probably can't change.   Increase the size of the series resistor to the gate and the capacitor on the FET gate.  This will filter out the feedback and protect the gate from picking up spikes from the wiring.

As suggested before, the D to S RC network will also dampen the oscillation.  However, highly suggest using  a .1 uF cap.   

If either these suggestions work, I think I would try just the RC network and the 330 ohm gate resistor for a minimum part count.

Good job on capturing that image!  Keep them coming.  Remember, if you see more than a momentary oscillation it will eventually bite you.

[johansen]

the problem is the loop formed between the mosfet, diode, and a capacitor located god knows where.. has to be very low. if the only capacitors are located inside the power supply then it better be really close..how big the capacitor actually needs to be to locally bypass the mosfet and diode, well you got an oscope, use it..
measure the voltage drain to source, and if it exceeds 20% more than the power supply.. reduce inductance or find a bigger bypass cap.

[diyaudio]

I haven't seen pictures of layout but this thing smacks of excessive lead length... put 1/4 wave resonators on gate and drain and you're sure to get a power oscillator somewhere.

The diode across the solenoid has to be referenced to the MOSFET.  The coil doesn't give a damn what's applied to it.  The diode is there to protect the MOSFET.  It needs to be wired at the MOSFET drain, and if the power supply is more than a few inches from that, you need a bypass capacitor from source to +V to return the diode to.

An R+C damper may come in handy to keep things stable.  10 ohm + 0.01uF may be a good starting point.  Drain to source.

Likewise, the gate needs a series resistor to charge it slowly, and faster turn-off probably wouldn't hurt either.

Tim

Yep on the money.

@Szewczykm

Probe the gate VGS, then probe the VDS, show us the waveforms, what ever is happening at the source terminal feeds at into the gate channel, note if its excessive transients, your MCU will damage as well because at high frequencies the parasitic capacitors are near short and feeds all that crap into the driver circuit (that's how most un-buffered mosfet drive circuit fails)   

Designs use a diode (free wheel diode) to assist the FET`s internal (Diode Reverse Recovery) yours is rated at SD = 4A, dISD/dt = 50A/µs, Max 150ns, some will use a RC between the Drain and Source as well, but note this forms resonator used to critically damp oscillations arising from layout,so you need to choose a R and C based on measured ringing you experiencing.

Also note when probing don't use the ground lead and clamp. use direct ground to avoid noise and inductance pickup.

see the difference between a probably switched FET on a good pcb layout and a design using prototype board. image is from a respected member from diyaudio.com showing her design.


Also Note
Gate to Threshold Voltage VGS(TH)  2 V, make sure you add a voltage clamp to keep the gate channel within its limits, this will greatly assist with the thin gate channel from being punctured.

[johansen]

there really is no mechanism for the capacitance between the drain and the gate to overpower the gate and destroy the gate driver.

[Andy Watson]

You need a resistor, 50-500R, in series with the gate lead and as close as possible. The problem is that the MOSFET can operate at frequencies at which the length of the gate lead wire forms a significant inductance. This inductance, together with parasitic capacitance to the drain forms and L-C oscillator - you need to spoil the Q of the L-C with a series resistor. Look up "grid stopper resistor", it's the same problem.

Also, some of the other measures, suggested above, to check the voltage at the drain would not go amiss.

[diyaudio]

there really is no mechanism for the capacitance between the drain and the gate to overpower the gate and destroy the gate driver.

you looking at it from a way too simple approach.  see attachment for clarification.

[Seekonk]

From the responses here it makes you wonder how anything works in this world.  Does it really take 20 components for a FET to drive a relay or massive ground plane?

[T3sl4co1l]

With a bad enough circuit, anything is possible!

If you think 20 components is bad, just imagine trying to represent a *real* component in discretes!  You'll have thousands of resistors, capacitors and inductors before you even get to the die, and that's for a mere percent-level model over most frequencies!  (Say, DC to 100GHz or so.  Hey, just because the transistor won't do much of anything at that frequency, doesn't mean the leads don't.  And that's to say nothing of their shiny finish at optical frequencies!)

Tim

[nuno]

You sure need at least a gate series resistor. I would start at 560 ohm and go up. If you switch slowly, which I would say you can do in your application, the need for bypass caps may probably be droped. If the gate wire is long, there must be other "cares", eventually reaching the point where it's just best to galvanically isolate micro and FET (for example with an opto-coupler).

[nuno]

Also note when probing don't use the ground lead and clamp. use direct ground to avoid noise and inductance pickup.

Oh, and if you didn't do like diyaudio says here, and put the probe directly on the FET's leads (both tip and GND**), your scope shots are garbage.


** but watch where you put our scope's GND lead, don't blow the scope

[Little Brutus]

Quote from: Szewczykm on 2015-01-14, 16:12:40

....
Application:  I want to be able to control a flipper on a pinball machine using a microcontroller.  This circuit is just a concept for the final design.

....
I'm finding that with this setup, when I press the button to send 5v to the mosfet, the coil will fire and lock on.  After the first firing, the mosfet is damaged and the connection between Source and Drain is a short.

It seems like everything on the mosfet is within spec, but I'm damaging it and it locks on.  Is it that I need a resistor limiting the current to the gate?  Is that where I'm causing the damage?

Since I do some work on Pinball machines...

Are you using serial wound or parallel wound coils? Which coil are you using? Normally the there are 2 coils in there, one high power (to make it flip) and one low power (to hold it in the up position).

Where is your end of stroke (EOS) switch for the coil? Is it properly adjusted? If it is missing or not adjusted, you are continuously drawing a significant amount of current through the circuit. The high power side should only be used when the flipper is moved to the up position, then disconnected (normally closed switch is opened mechanically by the flipper). The low side would stay in circuit to hold the flipper

As an example an FL-11630 coil high side, is only ~5 ohms, the low power side is ~500 ohms. At 50 volts, you are continuously drawing 10A.

There is an attached example of a parallel wound coil (like the FL-11630)

The Flipper button is equivalent to your sw-2. The mosfet would be between flipper button and the coil. The Left hand coil is the high power one.

Image came from http://techniek.flipperwinkel.nl/wms11/s11flip0.gif

Also, if you don't have an EOS switch and you solve the mosfet issue (say with a larger fet), you would probably end up burning out the coil.

As a safety, add an inline slow blow fuse in the circuit. 2A would probably suffice since the high power side should only be on for a very short time. (my machine powers 3 flippers with a single 5A SB, but I will also be adding individual ones for extra security)

[BloodyCactus]

Im building pinball controller system and my own DIY pinball. Lot of pinball machines use logic level mosfets for switching, eg the IRL520. Not to be confused with IRF520! I am using IRF640's with a mosfet driver (MCP1407) for all my solenoids at 50V and PWM'ing them. Later I changed my IRF640's to nicer TK42E12N1's.

I know IRL520's work great if your microcontroller is 5v but less so for 3.3v

Im using williams flipper coils but not using the 3rd lug, so its very much the same as Stern's whitestar/sam system.