Power Driver circuit frying

[Pauven]

I'm new around here, just found this community, and it feels like the right (only?) place to ask for help.  Sorry if this reads long, it's complicated.

I've developed 3 different power drivers for use in pinball machines.  My first two worked pretty well, so I'm not a complete newbie, but this new design is behaving badly, and I can't figure out what I did wrong.

The power driver concept:  Use the 5v signal from a PC's USB bus to control a pinball solenoid at up to 50V @ 5A.  My power driver is the size of a business card, and has 8 inputs/outputs.

The high level solution:  Implement high-side switching of solenoid power.  PNP Darlington controlling an N-Channel MOSFET controlling a P-Channel MOSFET.  The idea is that the first Darlington isolates the USB bus from the solenoid's high voltage. 

The components: 

PNP Darlington: ROHM DTA123JE http://www.rohm.com/web/global/datasheet/DTA123JE/dta123jetl-e 
This has built in biasing resistors, making it easy to use.  The datasheet indicates it can handle up to -50V, and I'm thinking in my design it never sees more than -5V.  It seems like the gate opens at -1.1V, so with -5V from the USB bus, this thing should be fully on, sending the full 5V to the N-Channel MOSFET.

N-Channel MOSFET:  Fairchild FDC3601N https://www.fairchildsemi.com/datasheets/FD/FDC3601N.pdf
This MOSFET is rated for 100V.  The datasheet indicates the Gate Threshold Voltage is typically 2.6V, with Min of 2V and Max of 4V.  I think this might be one of my mistakes, as 5V from the PNP Darlington may not be enough to reach saturation of this MOSFET, but I can't tell.  My goal is to always reach full saturation, basically using these transistors like light switches, minimizing resistance and heat and prolonging life.  I'm using a couple biasing resistors, 4.7k and 47k.

P-Channel MOSFET: Infineon IRF5210PbF http://www.infineon.com/dgdl/irf5210pbf.pdf?fileId=5546d462533600a4015355e3576b198b
I successfully used this MOSFET in my previous power driver design, and it has operated pinball machine solenoids for many years without fail.  Now that I'm troubleshooting and paying closer attention, I see that it has the typical MOSFET Gate-to-Source Voltage of +/-20V, and a Gate Threshold Voltage of -2V to -4V.  Come to think about it, my previous design is successfully using a Gate-to-Source Voltage of -50V, and somehow it works!  Makes me think I'm not reading this correctly. 

That's it as far as the design.  By the way, I modeled all this in 5Spice, and all the voltages and currents appear to be in spec, increasing my confusion as to why I'm having problems.

The symptoms:

Today I set up a test with a 38V power supply to feed an electromagnet (used on some pinball games to "catch" the ball).  Part of the test was to get a Max Current reading on the magnet (I saw a peak around 2 Amps, but it should have been closer to 7A, as the measured resistance is 5.1 Ohm).  For a good while, the test was going great, and the circuit was operating correctly.  Simulating the game, I would only trigger the gate for about 9 seconds at a time, though to be honest I wasn't waiting that long between tests, perhaps just 5-10 seconds.  I was emboldened by the low amp readings I was getting on the electromagnet, thinking my power driver could handle a couple amps easily.  I did this on/off test for a couple minutes.

Then the little PNP Darlington sizzled and smoked, and the gate was shorted open.  This happened at about 7 seconds into a 9 count, though my finger might have flinched cutting power, not sure.  If I hadn't released the trigger yet it shouldn't be a TVS, but if I did flinch it could be a TVS.  With the gate shorted open, I had to cut power to the board.  It was then that I noticed that the electromagnet was super hot (and connected to a freshly restored pinball playfield, I felt lucky I didn't burn the wood or melt the paint).  2A @ 38V is 76W, but with an electromagnet the size of my palm, I didn't expect it to get quite that hot that quick, which makes me think it might be pulling more amps than my meter is reporting.  The electromagnet is also 37 years old, and enjoyed a rough life in the 80's, so it might not be in game spec anymore.

I don't know if the frying Darlington damaged other components at the same time, but I found I had to scrape off the remains (nothing left but crumbs) in order for the power driver board to resume somewhat normal operation.  I knew 2 outputs were blown, but I still had 6 left on this power driver and wanted to keep testing.

I scaled back my test, to instead try the lowest power solenoid on the playfield, one with a 270 Ohm resistance that should only pull about 150 mA.  Immediately the next Darlington popped - hot blue sparks and smoke.

So then, to troubleshoot some more, I completely disconnected the solenoids from the outputs.  I still had the USB +5V and power supply +38V on the board, though.  I triggered another output, and instantly the Darlington popped (easy to tell, as I get hot blue sparks and smoke).  I tried the rest of the output and got the same result, instant death.

Now, to be fair, I can't tell if the PNP Darlington is failing first, or if the N-Channel MOSFET is failing first and in turn causing the Darlington to put on a fireworks show.  Being that these are tiny SMT components, I'm not really set up to test these individually to see if the N-Channel is failing too.

I'm also thinking that the very first failure, when powering the electromagnet, somehow fried the other PNP chips (shared power plane).

I suppose it is also possible that my problem is a Transient Voltage Spike.  In my previous design, I had to put TVS diodes in between the IRF5210PbF and the control chips (which were only rated for 50V) to handle up to 180V spikes. 

Somehow, I convinced myself that in my new design I wouldn't need them, but the more I think about it, the more I think I still need them.  This electromagnet seems to hold a charge for a while, putting off a mean TVS light show after power is cut even on the original pinball wiring (not using my power driver).  The TVS and the heat makes me scared to use it.  I haven't bothered to measure it, but the TVS might be higher than the 180V spikes I recorded a few years ago on simple solenoids.

I guess the first TVS could fry the N-Channel, and in turn fry the PNP Darlington, and since they are all on a shared power plane, fry all of them at the same time.

Hmmm, seems I might have answered my own questions.  Writing it all out and thinking about it helps. 

Though I would certainly appreciate a second set of eyes on my design.

-Paul

[drussell]

When you cut off the power to an energized magnetic coil it will try to drive the voltage across the now open circuit upwards towards infinity until it finds somewhere for it to go.    You DO need to take that into consideration in your design.

Without seeing your circuit schematic, I can't say why your unloaded outputs were blowing up, though, if they weren't actually connected to a solenoid...

[Pauven]

When you cut off the power to an energized magnetic coil it will try to drive the voltage across the now open circuit upwards towards infinity until it finds somewhere for it to go.    You DO need to take that into consideration in your design.

Without seeing your circuit schematic, I can't say why your unloaded outputs were blowing up, though, if they weren't actually connected to a solenoid...

You're absolutely right.  I never gave the voltage an escape path.  Seems an obvious oversight at this point.  Doubly painful is the fact that I just spent half a day troubleshooting the exact same issue I solved 4 years ago.  TVS diodes were the right solution then, and should work again. 

Are there any other realistic alternatives? 

Also, are there any general guidelines on sizing TVS diodes?  Last time, I bought what I thought would work, rated at 62V & 1.5KW (Reverse around 52V, Breakdown in the mid 50V's, and Clamping around 87V), and found that my 50V power supply triggered it even at idle and would burn it out.  So I tried the next size up, 68V & 1.5KW, and my power supply no longer triggered it, but it would die after a while from normal use.  I then wired up 4 of the same in parallel, and it has lasted years.  My understanding is that by wiring 4 in parallel, I increased the watt capacity from 1.5KW to 6KW.  How would I have known that I need so much capacity?

Also, I see some 5KW TVS Diodes, and I'm thinking that is a better solution, but what if 5KW isn't enough?  A handy formula would be helpful, just don't know if one exists.  Is it as simple as P=V^2/R?  Using the measured Transient Voltage Spike (180V) and the resistance of the coil (i.e. 7 Ohm) I get 4.6KW with that formula.  I also get 6.5KW with the 5.1 Ohm resistance of the electromagnet (and I see a new pinball magnet rated at 4.6 Ohm, so over 7KW).  Is my math and logic right?

On my previous machine I used a 50V power supply, but on my new build I'm using 38V.  Does a lower supply voltage result in lower TVS?  Or is a TVS voltage always max out at the full potential of the mechanical assembly (i.e. I manually operate the solenoid by hand, and release it, will it generate a 180V spike in an unpowered circuit)?


Pinball machines typically use low side switching, and I think that isolates the circuit from the voltage spike (plus the standard diode across the solenoid itself.  With my use of high side switching, I'm thinking the change allowed the spike to reach the circuitry, so I created a problem to solve that "real" pinball circuits don't have.  Am I thinking about this correctly?

Best guess on the unloaded circuits blowing up is that they got fried by the initial TVS event.  Before that I believe I was able to trigger them while unloaded without any issues.

Thanks again,
Paul

[james_s]

The classic solution for this is a simple diode across the coil to shunt the reverse spike that occurs when the switching element opens. I would try that first and see what happens. An oscilloscope would be useful here too, particularly a DSO that can capture a single shot event as you open the switch. Then you'll see exactly what's happening with the voltage.

[Pauven]

The classic solution for this is a simple diode across the coil to shunt the reverse spike that occurs when the switching element opens. I would try that first and see what happens. An oscilloscope would be useful here too, particularly a DSO that can capture a single shot event as you open the switch. Then you'll see exactly what's happening with the voltage.

You still get the TVS even with the diode.  I always make sure I have the diodes in place.

I use a Tenma 72-8400, which is a handheld oscilloscope/multimeter, which is how I got a snapshot of the 180V spike.  I have what I think is an original DSO, but whatever it is, it maxes out at 80V.  No need to fry the test equipment too.  You're right, testing with the oscilloscope proved very useful, but also very time consuming.  I'm on a very tight timeline, and need to order some diodes asap, and was hoping for some general wisdom on matters I know little about before I take a stab in the dark and order some diodes.  I took one EE class in college 25 years ago, and they didn't teach what I'm building...

I was also thinking about the electromagnet, which doesn't store any kinetic energy, yet still has a TVS.  I'm thinking a TVS must be influenced by the supply voltage.  It also seems reasonable to think that a 5V supply voltage would not produce a 180V TVS, and a 500V supply voltage would likely produce a TVS much larger.  Perhaps there's a formula, i.e. 3.5*Vs=TVS, which would calculate that a 5V supply voltage could create an 18V TVS, and a 500V supply voltage could create an 1800V TVS.  I doubt the formula is that simple, but it would be very helpful from an engineering standpoint.  I want to design for worst case scenarios, not just tested scenarios.

Thanks,
-Paul

[james_s]

DSO = Digital Storage Oscilloscope, it's a generic term covering a whole class of instruments. Your handheld scope/meter is a DSO, likely a fairly crude one but it sounds like it's capable enough to do the job.

I'm a little confused by what you're saying but I'll try to cover the important bits of what is a rather complex subject. Any time you pass current through a wire you will get a magnetic flux. When the circuit opens that magnetic flux collapses and induces a voltage in the wire (or coil of wire) proportional to the rate of change in the flux. This means that even from a low supply voltage of a few volts you can get spikes of hundreds of volts from a coil, this is the basic principal behind a boost converter.

Now as far as selecting a diode or TVS to use as a snubber, you need one that will not conduct at the highest voltage the power supply will produce otherwise it will shunt the power supply and whatever is the weakest link will burn out. The TVS must conduct at a voltage lower than that which will cause failure of the transistor that is driving the coil, otherwise the transistor will be damaged. So to pull some numbers out of the air for an example, let's say you have a 50V power supply, it's probably reasonable to use a 100V transistor to control the coil so a TVS rated to conduct at around 75V is a sensible choice to protect the transistor.

Now in reality since higher voltage mosfets are readily available at the current levels you need, you might consider using a 400-600V mosfet instead, perhaps even go with an avalanche rated part as they're much more tolerant of large spikes. You still will want a snubber across the coil, as I mentioned a simple diode may work just fine but a TVS is another option. There are also more advanced snubber circuits that use a capacitor to absorb the spike and dissipate it in a resistor. No need to make things too complex though.

You might also just look at schematics for some of the more modern pinball machines and see how they drive the coils, many of them are fairly well engineered and robust. I've replaced a few coil driver transistors in 90s pins but not too many.

Oh and forget about the DC resistance of the coil, that's not really relevant to the spike situation. The resistance applies to purely resistive loads which a coil is not, it's a very reactive load.

[Pauven]

DSO = Digital Storage Oscilloscope, it's a generic term covering a whole class of instruments. Your handheld scope/meter is a DSO, likely a fairly crude one but it sounds like it's capable enough to do the job.

I'm a little confused by what you're saying but I'll try to cover the important bits of what is a rather complex subject. Any time you pass current through a wire you will get a magnetic flux. When the circuit opens that magnetic flux collapses and induces a voltage in the wire (or coil of wire) proportional to the rate of change in the flux. This means that even from a low supply voltage of a few volts you can get spikes of hundreds of volts from a coil, this is the basic principal behind a boost converter.

Now as far as selecting a diode or TVS to use as a snubber, you need one that will not conduct at the highest voltage the power supply will produce otherwise it will shunt the power supply and whatever is the weakest link will burn out. The TVS must conduct at a voltage lower than that which will cause failure of the transistor that is driving the coil, otherwise the transistor will be damaged. So to pull some numbers out of the air for an example, let's say you have a 50V power supply, it's probably reasonable to use a 100V transistor to control the coil so a TVS rated to conduct at around 75V is a sensible choice to protect the transistor.

Now in reality since higher voltage mosfets are readily available at the current levels you need, you might consider using a 400-600V mosfet instead, perhaps even go with an avalanche rated part as they're much more tolerant of large spikes. You still will want a snubber across the coil, as I mentioned a simple diode may work just fine but a TVS is another option. There are also more advanced snubber circuits that use a capacitor to absorb the spike and dissipate it in a resistor. No need to make things too complex though.

You might also just look at schematics for some of the more modern pinball machines and see how they drive the coils, many of them are fairly well engineered and robust. I've replaced a few coil driver transistors in 90s pins but not too many.

Oh and forget about the DC resistance of the coil, that's not really relevant to the spike situation. The resistance applies to purely resistive loads which a coil is not, it's a very reactive load.

Ahhh, I gotcha now.  I have a Seeed Studio "DSO Nano".  I didn't realize DSO was a common term.  Thanks for the definition!

Your explanation on TVS diode sizing matches my understanding.  Thanks for confirming.

I'm too close to "go-live" to have time to re-engineer the power driver to use different MOSFETs, though that is a good suggestion to use higher rated components and I'll keep it in mind for the next design revision.  The challenge here is that at some point in the transistor chain I have to use logic level components for 5v control, which typically aren't rated to hundreds of volts.  And even if they were, wouldn't the voltage spike keep working its way up the line, and eventually cross over to my controller board and fry that instead?  The PNP Darlington appears to have sacrificed itself to save the upstream controller components (which are much more $$$).  The other challenges here are size and cost.  Often more capable components will be both bigger and more expensive.  Two things I'm trying to avoid.

I've done a lot of reverse engineering of pinball power driver schematics.  Part of the problem is me, in that I like to look at a solution, understand the solution, throw away the solution, and try and come up with my own (hopefully better) solution.  The big difference in pinball is that they use low side switching, meaning that there is live power on the playfield at all times and the ground is switched.  Sure, there is an interlock switch that cuts the power when you open the coin door.  But in general having live power running to the playfield, and the potential for a wire to break loose (pinball is violent, stuff happens) and connect to a metal structure that extends to the top side of the playfield, waiting for you to get shocked... yeah all that scares the electrons out of me. 

Of course, with low side switching, the TVS is running the other direction, away from the circuitry.  With my high side switching, I put my circuitry in the line of fire, creating a situation where there is a tvs that needs to go somewhere, and all it can find is my transistors.

Interesting that the coil resistance is not relevant to the tvs.  Assuming it was allowed me to do some beneficial math, that now I need to throw away.

I did find a 15KW TVS diode.  I figured one of these should more than do the job of the four 1.5KW diodes I'm using now, with the other specs being pretty much the same.


Thanks for all the helpful feedback!

Paul

[Audioguru]

The threshold voltage of a Mosfet is when it is almost turned OFF, but you want it completely turned ON, and the datasheet shows it completely turned on when its gate-source voltage is 10V that you do not have. You need a "logic level" Mosfet that is completely turned on with 4.5V or more.