High-side PMOSFET failing (can't turn it off after a few days) 3

[blcpro]

Hey all,

I'm having some trouble with a PMOSFET which I use to supply power to a thermal printhead. The problem is, after several days, or several power cycles, the PMOSFET seems to become a short circuit, and can't be turned off.

In the schematic provided, Q1 controls the turn-on and turn-off of the PMOSFET T1. There is a power sequencer that enables T1 after the system is ready. This seems to work fine for a while, but then the PMOSFET won't turn off, even when its gate voltage is 24V (Q1 off, R84 pulls gate voltage up to +24V).

The specs for the transistor T1: Alpha and Omega Semiconductor AOD4189: 40V, 40A, Vgs(th) ~ 3V @250uA.

http://www.aosmd.com/res/data_sheets/AOD4189.pdf

I know there is a lot of capacitance here, and I added D7 to help conduct when the power +24V is turned off, with those big capacitors fully charged. I thought maybe the body diode was failing, but adding D7 has not helped. What I have noticed, is that the resistance measured between the source and drain is very low after the "failure". I measured some new pmosfets and they seemed more normal (high resistance). Also, did a diode check on the PMOSFET body diode, and also on D7. D7 is still functional, but the Q1 body diode seems like a short.

Any thoughts on what might be happening here?

B

 

[blcpro]

The schematic didn't attach the first time... here it is again:

 

[imcjoek]

I'm not very good at this electronics voodoo, but I will offer a thought...

Don't you possibly get a very large in-rush current when initially turned on because the capacitors are discharged and tied directly to ground with no current limited resistors? Could that be causing the S->G junction to fail closed? IE: The only impedance in your circuit at the moment of the great on-turning is the capacitor ESR (which you have halved by putting two in parallel) and any stray inductance from your leads/pcb traces.

 

[blcpro]

This may cause the Source-Gate junction to fail? Yes, I would expect a large in-rush current, but my thermal printhead might demand fairly large currents (as much as 7 amps, in the worst case) for short bursts. Granted, that's really what the large capacitors are for, but I wouldn't want the voltage and power to dip any while printing due to some current limiting resistors. However, I'd like some more info about why the S-G junction might fail in this case. I would think that if the S-G junction failed, the mosfet would be always off, instead of always on?

B

 

[blcpro]

Oh, but if the inrush current exceeded the S-D junction limit? Would that be the cause of the failure? Or is that exactly what imcjoek meant to say?

 

[imcjoek]

S-D is what I meant, yes.

At initial startup a worst case looks like:
Mosfet goes rapidly to Vgs(-10V) and RDS-on(0.022ohms)
Capacitor ESR could be ~0.1ohms for 2200uF aluminum electrolytic. So 0.05ohms when you put the two together.
You now have 24V across 0.072ohms = 333A, if only for an instant.

In reality the mosfet's gate capacitance will limit how quickly it reaches RDS-on, and the entire circuit should have some inductance to limit the current ramp rate. So the real amperage is probably less, but still a potentially big number.

Is this your problem? I don't know. Just pointing it out.

 

[blcpro]

That very well may be the problem... I'll have to do some more testing and see if an ICL (NTC thermistor type) might fix the problem. Thanks very much for bringing this to my attention.

B

 

[geekEE]

Do the large capacitors have to be on the drain of the MOSFET? Couldn't you put them on the source of MOSFET instead? You can put some smaller, faster caps on the drain.

Glenn

 

[LiteYear]

Yep, there's a momentary short through that MOSFET when it turns on, probably limited by the power supply. Rather than adding series resistance, what usually works in this case is to slow the turn on of the mosfet. As long as you're not cycling too often the mosfet will handle it fine, and it saves adding a constant power loss. Adding capacitance from G to ground ought to do it. You'll just have to study the datasheet and probably take some measurements to figure out how much.

 

[blcpro]

Thanks for the suggestions, I'll keep them in mind while I re-evaluate the circuit.

B

 

[iop995]

Reduce R84 to 470ohm. Depending of switching frequency requested by circuit, may reduce initial turn-on MOS current by adding a capacitor between gate and ground; value depend of switching freq, MOS capacitance and ramp current desired.

 

[OperaHouse]

I would add a capacitor, but NOT TO GROUND. That can cause two problems. When power is first applied to the device it turns on the FET momentarily. That may not be a problem. The real problem is the cap can over voltage the FET gate when power is first applied. Place the cap parallel to the gate drain resistor. You don't want to trade one problem for another.

 

[OperaHouse]

I would also think about changing the 2K/2K voltage divider to a much higher value. No need to have to add a larger capacitance than necessary to get the right RC time constant.

 

[LiteYear]

Good point Operahouse. Totally missed that it is a P-channel MOSFET. My suggestion is therefore upside-down.

 

[blcpro]

Thanks for your comments.. I have been experimenting with larger resistor values and different ratios, and I can see some slower turn-on times. Also, I have experimented with some inrush current limitors, and so far the problem has not happened again. I am re-designing the circuit to take these factors into account. Thanks again for the help!

B

 

[blcpro]

UPDATE:
I have been testing some of my ideas and some of the suggestions posted here before I redo the board layout. I have attached an updated schematic with my changes so far. What do you guys think? After some testing, I discovered that I probably did not need as much capacitance as I had previously. Right away this will reduce the ESR and inrush current. Additionally, I was testing some resistor/capacitor combinations to slow down the turn-on time of the transistor T1. My solution also includes a diode so that the turn-off time will be quick when it needs to be. Also, I moved some of the bulk capacitance to the source side of the mosfet. This further reduces the inrush for when T1 turns on with uncharged caps. But I did add some fast ceramic and low-ESR aluminum caps (C125 and C97). Overall, I have reduced the total capacitance used for the print load by about 50%. My tests so far look good, and I don't see any voltage dip when the printhead is printing. I saw quite a bit of voltage dip when I tried an inrush current limiter in series with the mosfet drain, so I decided to slow the turn-on time and reduce the ESR. Thanks again for your contributions. I'll welcome any additional comments.

 

[OperaHouse]

There is absolutely no need for R102 100 ohm. This is insignifigant campared to the transistor driving resistances. Off hand I would think that C126 1uF is much larger than it has to be. making it smaller and the 100K larger would give better turn on/ off.

 

[blcpro]

Thanks, OperaHouse.

In my test setup I'm actually using 1Mohm and 2.2uF cap to keep the turn on time very slow. I'm able to measure the inrush current, and it doesn't spike more than about 9 amps. The MOSFET can handle up to 40A (so it claims), and previously, before the adjustments, I saw such a huge spike I couldn't even measure it to its peak (over 80A for a brief few uSecs). I'll consider removing the R102. Thanks for pointing that out.

Any possibility of the gate resistor R103 at a very large value (say > 1Mohm) being so large that it won't be able to exceed some gate leakage currents and leakage from the capacitor C126? I just wonder if you keep increasing that gate resistor value that it won't be able to fully turn on the MOSFET.

B

 

[blcpro]

Also, for this MOSFET, the gate threshold voltage seems pretty low (typically -1.9V), so I think thats why I needed such a large value for C126, so that the gate charge builds as slowly as possible in that region.

 

[OperaHouse]

My gut feeling is you could use 100K and .1uF and tell the diode to stay home. ON can mean different things. Is that when it starts conduction? Then it will self current limit and partially charge the cap. The event can't take that long. Keep staring at it and you will begin to think phantom turn on from Miller Effect.