Is the Prop killing my MOSFET Drivers?

Philldapill

I'm building a high current PWM controller, driven by the propeller. It's going to have all sorts of cool features that only the propeller could do, like constant current feedback, power accumulation etc., but that's beside the point.

I am using the UCC27323 Mosfet Driver IC from TI. For some reason, when I connect even a fairly small load to my mosfets, the DRIVER, not the FETs themselves, explodes and gives a puff of smoke. I've taken the frequency down to 100Hz and almost instantly, the driver blows when a load is attached. I don't think frequency is really the issue here. I've read all the application notes about the UCC27323 and all I can find is how it keeps saying to keep the output of the driver as close to the mosfet as possible, and maybe even add a small resistor in between the output and gate. The Driver's datasheet uses a 10nF load as the testing load, while my mosfets are only 4.4nF all-in-all.

Just a little side note... When I have my mosfets on the board, and the driver, but no·load on the mosfets,·everything is fine. I see on the scope that the gates are switching just as they should. However, the moment I put a load on it, I get hit in the face with hot pieces of plastic chip casing...

Attached is the EAGLE·board schematic I made. Sorry, I don't really make electrical diagrams for simple·circuits like this. I hope you can follow it.


Any thoughts???

BradC

Philldapill said...
all I can find is how it keeps saying to keep the output of the driver as close to the mosfet as possible, and maybe even add a small resistor in between the output and gate.

Excuse my complete ignorance here, but why would you keep "as close to the mosfets as possible" yet "maybe even add a small resistor in between the output and gate".
That seems incredibly counter-intuitive to me.

I would have _thought_ (and remember I'm uneducated in anything electronic - "would you buy a transistor from this man??") that keeping things close together would be to minimise resistance and capacitance, but adding a resistor would do exactly the opposite.

Someone hit me with the clue stick please!

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cardinal Fang! Fetch the comfy chair.

Philldapill

You need the driver as close as possible to reduce any parasitic inductances/capacitances. The mosfet is already essentially a capacitor, or at least appears that way to the driver since the gate has a capacitance. If you have long stray wires or traces between the gate and driver, the result will be an RLC tank circuit, and the mosfet gate will begin to oscillate very rapidly. The driver hits the gate with up to 4A when it first charges/discharges the gate, and given the right frequency of puslation from the driver, the gate will start oscillating, turning the mosfet on and off very rapidly, thereby creating alot of heat in the mosfet.

The reason for the small resistor in between, is to reduce the "Q" of this natural RLC circuit. I thought the same as you did at first, "Isn't the goal of the driver to charge/discharge the gate as QUICKLY as possible?". The resistor DOES in fact reduce effiency directly, but it also indirectly aids in keeping the circuit stable by reducing any oscillation between the gate and stray inductance.

As for the explosions, I just tried something a little different(added another bypass cap), and this time the top of the chip literally hit the ceiling. Something is wrong... The funny thing is, I have another circuit with essentially the same setup, and it works just fine. I am running one of these tiny TO-252 package mosfets at 30kHz, 5A and it's not even warm, nor is the driver.

Carl Hayes

Your diagram is not easy to interpret, but I'd bet a lot that you only think you're driving the gate of each FET.· I'd give big odds that you're actually connected to the source or drain, perhaps both.· So, when you connect a load, the driver output is now connected (through the load) to one rail of the load's supply.· Boom.



▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
· -- Carl, nn5i@arrl.net

Philldapill

That's exactly what I figured, Carl! The way it goes boom, suggested to me that the Driver was actually driving the load somehow, but everything that I've looked at says it's ok... Here is what I KNOW - The GROUND plane on the diagram is the ground of my battery(-). My Load connects between the Positive Battery Terminal, and the LOAD(-) Plane on the diagram. I know the MOSFET pins are correctly arranged on the board, i.e. Source is connected to the negative(GROUND) terminal of the battery, and the drain is connected to the Negative Load plane on the board. The Gate is connected through a resistor, to the ouput of the driver. Other than that, I can't figure it out, but you are right, it seems like I'm shorting the output of the driver somehow.

scotta

I don't see bypass caps next to the mosfet driver. Also, the slow nature of the
opto-isolator could be causing a slow rise/fall time being applied to the input
of the mosfet driver. The internal translator circuits could be driving OUTA
high before OUTB resulting in cross conduction.

Just a thought.

Philldapill

Well, the bypass caps are actually the green vias next to the socket. I'm cheap. I do my PCB's at home the quick and dirty way.

As for the Isolators, I should have specified. These are no opto-isolators. These are the ISO721 digital 150MBPS isolators from TI. They are extremely fast. I think their rise and fall times are faster than the driver itself!

As for the cross conduction, I doubt it... The datasheet explicitly says it is just fine to parallel outputs/inputs in this way. My original board was using a UCC27322 which is a 9A, single output version. I was getting the same results(failures), and that is why I actually switched to the 4A, dual driver version. Apparently, same problem.

Thanks guys. Even if I shoot your ideas down, they help me look at the problem from another angle, and I appreciate it.


Problem still has yet to be solved.

Carl Hayes

What FETs are you using? Are you sure of the pinouts?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
· -- Carl, nn5i@arrl.net

Philldapill

STB60N55F3. I would love a second opinion about if my pinouts are correct on the diagram... I've looked at this so many times, it's hard to see a problem, even if it's right there, staring me in the face.

Phil Pilgrim (PhiPi)

Some observations:

1. If you're using an opto-isolator, yes, it may well be too slow (rf: scotta). What is the part number?

2. If your 12V power is connected on the lefthand side of the board, as you indicate, you've got motor current returning through that skinny trace under the driver. This is not good, because it will alter the voltage that the driver sees on its supply terminals. You need to cut that trace where it joins the groundplane and run a wire directly from the power input to the groundplane.

3. Are you using protection diodes across your load? If not, you may get dangerously high transient voltages feeding back and killing your driver.

4. Always use bypas caps (rf: scotta)! Also, depending on the size of your load, you will require a large (1000uF or more) electrolytic cap where the power comes into the board.

-Phil

Philldapill

@Phil Pilgrim

1. The isolator is not an opto-isolator. It's rated for 150MBPS, and is faster than the driver itself. It is an ISO721M digital Isolator from TI.

2. Under these test conditions, I am using an old headlight that draws about 7.2A under a 12.65V Battery. There should be minimal inductance in the lamp. As for the skinny trace on the board, I don't know where the current would be going - The Negative terminal of the battery is connected to this GROUND Plane with a good sized clip. I see your point about the possible elevated voltage where the trace meets the plane, but I have these traces coated in thick solder and I really don't feel that it is causing that much of a problem... I have a bypass caps all over - 0.1uF ceramics at each device, and 100uF at the driver as well, plus a 1000uF cap at the supply which isn't shown in the diagram because it was an after thought once I made the board(durrrrr).

3. No, I am not using protection diodes on THIS board. My previous one in which I used the 9A version of the chip, yes I did use a 2N1007 diode, but it may not be fast enough. Regardless, it still blew with the old board.

4. Bypass caps are everywhere.

scotta

Could the gates of the mosfets be shorted to ground or supply ?

Phil Pilgrim (PhiPi)

Reiterating Carl's question, what's the part number for your FETs?

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!

Philldapill

Sorry, I thought I had posted it. The Part number is STB60N55F3 from STMicro.

I'm sure the gates are NOT shorted to ground or supply. The circuit works just fine when no load is attached, but the momeny I attach CERTAIN loads, it goes poof. I've just found that when I put a 1Ohm resistor across the Positive Supply and Drain of the mosfet, it works fine, gets warm, and hums at 1kHz, as it should. The resistor is pulling around 12A, but when I put the headlight in the same configuration, it goes poof. The headlight pulls about 6A directly connected to the battery.

soshimo

Maybe your load impedance is too low with the light? If you are drawing 6A with a 12V supply your impedance is 2ohms. Maybe try connecting the 10ohm resistor in serial with the light?

Harley

Incandescent lamps, when cold, draw many times the current of the hot state. Like maybe 10x!

I learned this back in the late 50's when my lamp drivers failed way sooner than expected. Plotting with low current, found tungsten had quite a low cold resistance vs. that when at rated voltage and current. Nature and it's many surprises.


▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Harley Shanko

Philldapill

Soshimo, that's the funny thing about the low impedance. I tried the same thing with a 1ohm power resistor, and it works great! For some reason the lamp causes it to fail. On my previous prototype board, using the same chip, and as far as I can tell, the same circuit, the lamp works fine. I'm ready to throw this whole thing in the trash and say forget it, but the engineer in me is obsessive about figuring out the problem...............

Leon

Do you have a schematic? The PCB layout isn't very useful.

Leon

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Amateur radio callsign: G1HSM
Suzuki SV1000S motorcycle

Phil Pilgrim (PhiPi)

You do realize that incandescent lamps present a much lower resistance than their running resistance before the filament heats up, right? The initial, transient load may exceed the ratings of your components. Since you're PWMing the input, could you start with a lower duty cycle until the filiment incandesces and work your way up from there? Another option is to add some series inductance as a sort of ballast. A properly-sized inductor will limit the inrush current until the filment has a chance to get hot.

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!

Philldapill

One more note. I just tried my big-ol-motor @ 100Hz, 20% duty. Worked fine. I have a couple of diodes from the Battery(+) to the Drain of the mosfets, and a diode from Battery(+) to source. I think that should cover any voltage spikes. However, when I increased the frequency from 100Hz to 1Khz, kaboom. More smoke... Now, I replaced the chip(for the um-teenth billion time), and the motor works fine again @ 100Hz. This is my last chip so I'd rather not blow it again.

@Leon,
Sorry about the bad schematic. It's a pretty simple circuit. I'll attached a circuit diagram in a minute.

@Phil,
Yes, I've tried that. I modified my psmAsm file so that I have 0-1000 duty cycle control, rather than 0-100. I started at 5(0.5%) duty cycle, and it did the same thing. Lots of smoke, and a racing heart rate.

soshimo

Forgive my ignorance, it's been a few years, but isn't PWM power controlled by the duty cycle? Since the power is determined by the average value if the duty cycle stays constant the power should stay constant, regardless of the frequency. Or am I way off? If that is the case though, why would increasing your frequency (not your duty cycle) have any affect on what the load sees?

Philldapill

If I increase the frequency, two things happen. First, the driver has to charge/discharge the mosfet more often. Each time the mosfet gate is charged/discharge, the driver has to dissipate some power. If I took the frequency to the extreme, then the driver consumes a huge amount of power and will over heat.

The second thing that can happen, is when driving an inductive load, each time the mosfet turns off, there is a voltage spike. At lower frequencies, this is probably not too bad, but at high frequencies, the repetitive voltage spike can damage components such as diodes by overheating them.

Philldapill

Just curious... What effect will inadequate capacitor bypassing at the power pins of the driver have?

soshimo

Ah, I had to draw out some waveforms to see, but yes, you are right, increasing your frequency by 10 (as you did) will increase the number of pulses by 10 (if the duty cycle remains constant). I just heard PWM in the beginning and was too lazy to reverse engineer your board layout to a schematic. Though, you might want to think of doing that quickly as it might be easier to troubleshoot.

Also, I forgot to add, there are plenty of solutions for flyback from inductive loads and it looks like you have used some of them already (filter caps, diodes, opto-isolation, etc..), I don't that that is your problem here.

Post Edited (soshimo) : 12/18/2008 1:18:14 AM GMT

Phil Pilgrim (PhiPi)

It's looking more and more like you need to limit the inrush current caused by the initial, nearly-dead-short load. In lieu of an inductor, something from this list might be a better solution.

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!

Philldapill

Are you talking about the inrush current that the mosfet Drain-Source has through it, or the inrush driver current from driver to gate to source?

soshimo

Sorry for the hijack - hey Phil, those ICL devices are pretty cool. They look just like PTC fuses - only it seems like they work exactly opposite, when cold presenting very large impedance and when heated up presenting very low impedance. Cool beans!

Phil Pilgrim (PhiPi)

I meant the drain-source current. That's where the surge occurs. It's still unclear why the drivers are getting zapped, since the gates are supposed to be insulated from the action; but since it seems to be load-dependent, that's where I'd start.

BTW, what gage wire — and how long — comes from the power supply to the board?

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!

Philldapill

I'm using 12AWG(I think), or there about. All my wire is this size, incuding to the load. EDIT: The wire length from battery to board is about 18", and from board to load is about 24" total(2 alligator clip wires).

I always thought that the gate capacitance and current into the gate was independant of the load Drain-Source Current.

By the way, I changed out my resistor from the driver to the gate. I put a 124Ohm in the 4.7Ohm's place. Same thing, no change. I'm starting to wonder if there is something causing the internal circuitry of the driver to turn both sides of it's bridge on at the same time. With the 124Ohm resistor in place, the MAXIMUM current into the gate from the driver, even if it were shorted, would be 12.6V / 124Ohm = ~101mA. I seriously doubt this could possibly be causing the driver to blow instantly. I've found that when one of these blow, they aren't hot to the touch, which makes me think there is an internal short that is extremely fast, but builds enough pressure to blow the device apart, stopping the short before any heat can accumulate.

Post Edited (Philldapill) : 12/18/2008 2:15:02 AM GMT

Philldapill

Ok, so here's a theory. SOMETHING is causing the Half-Bridge inside the driver to turn the high and low side on at the same time, somehow. I'm not sure if it's just inadequate dead time where neither are on, but I think both of them being on at the same time causes a direct short from the positive rail, to ground, causing the violent explosion, yet lack of heat.

Now, the cause of this is still a mystery, but I now have about a 315Ohm resistor between driver and gate, and it seems to work like a champ, yet my mosfet now gets pretty warm at higher frequencies(above 30kHz)... [noparse]:([/noparse]

Phil Pilgrim (PhiPi)

By using the larger gate resistor, you're effectively limiting inrush current by slowing the turn-on. Without knowing the precise mechanics of it, I'm convinced that this is an inrush current problem and that whatever you can do to limit those currents at the load (via an inrush current limiter) will solve your problem.

-Phil

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
'Just a few PropSTICK Kit bare PCBs left!