Blown up SIR! NFET death

[itsmoked]

Hello,
In the continuing saga of thread248-184216 I have another question you esteemed EEs may be able to help me with.

Things were going pretty smoothly until today.
Remember this is a charger that is voltage controllable. Today for the first time I ran a routine that stepped the voltage up thru 960 steps to an expected 30V. Interestingly the voltage went to 35.8V before topping out.

I have a circuit that allows me to insert some resistance if needed and when not needed to shunt it. It consists of a highside driver and a power Nfet <see below>

The FET is:

http://www.fairchildsemi.com/ds/FD/FDB8441.pdf

The highside driver is:

http://www.micrel.com/_PDF/mic5014.pdf

Here's what I've pieced together: I ramped the output. It hit 35.8V. I altered my code to not ramp as far. Hit 33V. I adjusted further so now it hits 29.9V which is just fine.

While sitting there I added load. The current climbed to 3.5A. I held there and scanned the board for hot spots. None of the switcher components nor our 'friend' the blocking diode were even noticing the load. But this FET sure was. It was ratcheting up as I watched. It hit 90C <case> and I dropped the load. I measured the gate drive and noticed that it was the same as the drain voltage. Huh..?... Should be ~15V higher by my understanding of the highside driver data sheet.

Well the data sheet sez 30V max standard operating voltage for the MIC5014. It also sez it will take 60V hits (load dumps). I ran at 35.8V for maybe 5s. Perhaps that cooked it. So I rip and replace it.

Re-connect everything and re-ramp to 29.9V. Add load. Same thing. FET starts to space-heater on me. I dump the load at 90C again. Takes about 30 seconds to reach 90C. Head scratching, I decide the FET may be bad. I check with a voltmeter. All three terminals are pretty much the same voltage. The FET is now a 1/3 OHM resistor. Period!

Studying the data sheets I recognize that the FET will toast if Vgs exceeds +/- 20V. This means that if I command the FET off while the supply is above 20V... Toasty Critters! Well I didn't command the FET off! More head scratching. That could certainly be a weakness though. One CPU hick-up and piffft. After a third read of the MIC5014 data sheet I see, "to protect your system the MIC5014 will shut off the gate drive if 35V is exceeded." So the MIC5014 shut off the FET while it was being fed 35.8V no doubt obliterating its gate structure.

So now that you see the problem anybody want to suggest a solution to prevent a recurrence? Another words, if the gate drive is pulled to ground by the MIC5014 the gate should never gets less than 15V below the source. Keep in mind this is a slow switch just replacing a relay so there are no speed concerns as long as it switches in less than 100mS.

Any suggestions would be appreciated.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[OperaHouse]

A 18V zener from gate to source and resistor to gate drive. Since low speed, this could be a couple K

 

[LionelHutz]

I'd think series connected back to back source to gate zeners at about 18V with a 50 ohm or so gate resistor would fix the problem.

I did look at the datasheet and it does seem like it will pull the gate output to the Gnd pin potential on turn-off but I didn't see anything discussing this as being a possible issue or anything on protecting the FET or IC from different transient event possibilities.

Have you looked to see if there are some app notes for that series of IC's going into more detail. I know IR does that with their driver IC's and reading the app notes tells you a bunch of stuff their datasheets do not.

Most FET info says you should always have a gate resistor connected as close to the FET as possible to avoid high frequency oscillations.

 

[BrianG]

Unless your design has to have a ground-referred battery negative return (i.e. not a fully "floating" charger output) and if your FET is simply an "isolation relay" why not change the high-side switch for a low-side switch? That should make the voltage differences easier to control.

 

[Noway2]

A back to back zener like LionelHutz suggested might do the trick. You would have to connect the zener anodes together with the cathode connctions at the gate and the source.

You would need both zeners to keep the source voltage from climbing to the gate voltage via forward conduction of the off mode clamping zener. The second zener would have to have a minimum Vz that is greater than the max Vg-Vs (17 V in this case). I am wondering if a standard diode would work for this as it would provide a higher reverse breakdown voltage while still only adding ~.7V when zener would be conducting.

The resistor would be needed to current limit the (sinking) gate drive pin. I didn't see anything in the data sheet that shows how much current this pin can take. The data sheet says that the logic pins have a +/- 2uA current limit, but I don't think this applies to this pin. From the block diagram in the data sheet, it looks like when the driver goes into shut off mode, the charge pump is isolated from the gate, which is switched to ground, so sinking into the gate pin may be ok. The data sheet shows a maximum 50mA source current, so I would choose a resistor to current limit it to less than that.

 

[jimkirk]

Along the lines of BrianG's suggestion, if you have to have a high line switch, I prefer using PFETs. They are slightly more expensing, but controlling them is far easier, pretty much offsetting the increased cost.

 

[richs]

Hiya-

In keeping with Noway2 excellent post:
"The resistor would be needed to current limit the (sinking) gate drive pin. I didn't see anything in the data sheet that shows how much current this pin can take."

My "intuition" suggests a pass transistor in leiu of just a simple resistor. I think a simple BJT PNP would most likely do the trick. I haven't thought it through.

Just my 2 cents.

Cheers,

Rich S.

 

[Noway2]

Jimkirk, The idea of using a PFET is a good one. The problem with them though is that due to the lesser carrier mobility the RDs-on of the P devices is about 5x larger. At higher output currents, say 10Amps for example, both voltage drop and power dissipation become a problem, especially at higher operating temperatures. You are correct, though, a P-Chan device would be a cinch to drive.

Richs, I think you and I are sharing some intuitive thoughts here. I too was thinking that somehow a pass transistor would be the right choice (based on feeling).

I have a thought about using a P-channel device as a crowbar. It could be turned on by the high side driver pulling the gate control low. According to the circuit above, there is a current limiting resistor in the circuit. A second resistor could be added at the source terminal of the P-Chan 'crowbar'. This way, when the gate control goes to ground, the P-chan conducts and the Source voltage can be pulled down to a safe limit via the two resistors. Of course, a voltage divider would be needed between the P-Chan source and the gate control to keep its Vgs within limits, but those values can be chosen to keep the current sunk into the high side driver within safe limits. Or even a second transistor could be used to turn on the 'crowbar' transistor and this would eliminate any current into the gate control of the high side driver.

Sorry if the above sounds complex, but I don't know of any way that I have available to post or paste a circuit schematic.

 

[itsmoked]

OperaHouse; Certainly sounds easy. A solution I can try easily and would alter the least stuff.

Mr. Hutz; You're thinking back to back to protect on the Vgs going too high too? I think that is a freebie by the MIC5014 not driving over 17V above the source. As for "close to the gate" resistors, that's not a problem. On this SMD board anywhere is 'close'.

I just combed the site for apps which had nothing specific other than the admonition of "protect the Vgs!" They do show a protection Zener in a circuit and interestingly they never mention any resistor.. Seems you are just forcing their charge pump into a short and they don't care.

Wait!! A single Zener provides.. Protection from both overdriving and under-volting the Gate. In the overdrive direction if the driver raises the gate voltage too high then the Zener avalanches to the source kneecapping the voltage to the Zener voltage. If on the other hand the Gate voltage ever drops below the Zener's forward voltage (~0.5V) the Zener conducts with just the forward voltage drop.
This means if I'm not troubled by the Gate ever being over-driven I could just use any old diode as That is the only direction I'm after for protection.

Still think I should add a resistor.

Hi BrianG. The design is way down the road from a jump to low side... Many boards, multiple batteries - lowside becomes problematic.

Howdy Noway2! Ahhh I see the MIC5014 data sheet shows the charge pump isolated during OFF states and the Gate internally FETted to ground.. That would explain their lack of a resistor. But I would think any capacitance hung out on the source could blast that internal FET. So I agree totally that there needs to be some resistance, 50mA (or less)sounds good. I'll look at the BOM and try to find a value already present. I'll also try an already present regular diode.

Thanks jimkirk. I will look at that next time.. Since I'm working on a layed out board a change like that is not very desirable at this stage. Good thought though.

Greeting richs. You lost me.. LOL

Noway2 back again? Atta boy! Man you guys are transistor happy! Haha! There is no doubt that a system as described would work. But I need to keep the BOM down as my customer has a sharp pencil which he stabs at me when I get too pricey. Thanks for the clarification on the PFET I knew there was a reason I didn't just use a second one as I have one on there now. But it only needs to pass a few amps not 10.

I will try the (1)regular diode and (1)regular resistor today and get back to you all with the results. That is unless someone starts screaming "NO" in the defence of the FET.

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com

 

[Noway2]

"But I need to keep the BOM down as my customer has a sharp pencil which he stabs at me when I get too pricey."

Hahaha!

Sounds like you need to explain to the customer the benefits of what you are doing and how ultimately they will be money ahead.

 

[sreid]

How about a 15 V voltage regulator (78L15) between the input voltage and the gate driver V+ input. At voltages below 15 volts, the regulator tracks the input voltage but limits the supply voltage to 15 VDC.

 

[OperaHouse]

Am I the only one here that can't for the life of me understand why you need two back to back zeners? I've read that post till I'm blue in the face.

 

[LionelHutz]

OperaHouse you're right that a single zener would work in this application. I was just thinking that a FET will turn off quicker when you take the Vgs negative and that you just want to keep the gate voltage within 20V of the source.

Keith, I would still protect the FET gate under all conditions gy using a zener and not just a diode. Protect the FET with the zener and the IC output with the resistor.

I'd have to ask why you didn't use something like a PNP transistor and then just drive the base from a small NPN connected emmitter to gnd switched on by the PIC? 1 x power PNP, 1 x small signal NPN and 2 x resistors.

 

[itsmoked]

Hello Mr. Hutz.

I installed a SOT23 Si diode and a 10K resistor and it worked beautifully! I understand your (single) Zener method but Zeners cost more than the Si's I already have splattered around the board. I have two other Zeners but neither(5V & 30V) are close to the value I'd need - meaning yet another part for the BOM. Furthermore the MIC5014 has a Zener in it that serves the same basic function.

I ran the output up to 30V and then switched the MIC5014's control pin ON and OFF a few times. The Gate voltage dropped instantly to the Vsource - 0.5V each time then went back to Source + 16V when ON.

I didn't use a transistor instead of a FET as I need to keep as much heat as possible off the board. 0.2V x 10A = 2W I'd have to contend with instead of 2.5mOhm x 10² = 0.25W. The difference is, in one case, I need serious board plane area or a heat sink verses the other that comes down to 'just' mounting the part and its case alone will deal with the 1/4W.

Thanks for the help everyone!

Keith Cress
Flamin Systems, Inc.-

http://www.flaminsystems.com