Go Down

Topic: What killed this MOSFET driver? (Read 3474 times) previous topic - next topic

saximus

Mar 15, 2017, 12:24 am Last Edit: Mar 15, 2017, 12:26 am by saximus
Hello my knowledgeable friends. A coworker and I have had a bit of a conundrum and I was hoping for some assistance.
We are attempting to use a TC4431 as a high side MOSFET driver for some lamp dimming. The circuit was wired up per the attached picture. My colleague hooked everything up to his bench power supply with the current limit turned down to 0A. As he wound up the current he said the lamps flickered and everything shut off. This failure seemed to take out not only the driver but also the Arduino.

I did a little bit of post mortem and found that, with the Arduino disconnected, there was about 25V on the input pin (2) of the driver. So I can only assume something failed internally, allowing VDD to to pass into the PWM pin. I just can't work out how or why this would have happened though? As far as I can tell, he had it hooked up per the datasheet. The only thing that seems to be missing are the decoupling capacitors which I assumed should really only affect switching speeds but since the PWM frequency is only set to 50Hz and the load is well within the capabilities of the MOSFET, surely that can't be the issue?

MOSFET Datasheet

Driver Datasheet


DVDdoug

That's weird...


Sometime bypass caps can prevent oscillation, so it's probably a good idea to add them (next time).

Quote
I did a little bit of post mortem and found that, with the Arduino disconnected, there was about 25V on the input pin (2) of the driver.
That explains how the Arduino got killed.  But, I don't know why the driver died.

Is the MOSFET OK?  ...I wonder if the current limiting somehow "backfired"?   When current limiting kicks-in the voltage will be reduced and if VGS is too low the MOSFET won't turn fully-on and it could overheat and fry, shorting everything out, causing a chain reaction failure.     ...Just a wild-ass guess. 

Maybe next time...  Test the MOSFET alone (manually switching the gate between 28V and Ground).   You can also test the driver alone or with the MOSFET before connecting the Arduino.

You might also try adding a resistor (maybe 1K) between the Arduino and driver, then add a "backwards" protection diode between the Arduino's output and 5V.   The diode will conduct if the driver shorts and it's input goes over 5V, limiting the voltage to about 5.7V, and the resistor will limit the current (to protect the diode).



TomGeorge

Hi,
Have you got the arduino gnd connected to the TC4431 gnd?

I only see one connection between the arduino and the driver IC.

Tom.... :)
Everything runs on smoke, let the smoke out, it stops running....

MarkT

You are driving a mosfet that needs 12V of gate drive with 28V, and its absmax Vgs is +/- 30V, so
clearly you blew the gate oxide and thus the FET.
[ I will NOT respond to personal messages, I WILL delete them, use the forum please ]

756E6C

Did you use good ESD practice while handling? You may have "wounded" it.

MarkT

#5
Mar 15, 2017, 01:12 pm Last Edit: Mar 15, 2017, 01:13 pm by MarkT
I look at the datasheet for the TC4431 and yes it does say its a high-side driver, but clearly it is not,
(likely a copy/paste error when they did the datasheets for several devices)
its a low side driver, there is no charge pump or level shifting circuitry.  The high voltage rating is for robustness
in high power applications, it would normally be powered from 12V to 15V for MOSFETs and IGBTs
respectively.
[ I will NOT respond to personal messages, I WILL delete them, use the forum please ]

stuart0

#6
Mar 15, 2017, 02:17 pm Last Edit: Mar 15, 2017, 02:31 pm by stuart0
I look at the datasheet for the TC4431 and yes it does say its a high-side driver, but clearly it is not,
(likely a copy/paste error when they did the datasheets for several devices)
I noticed that too Mark. In the title it says "High Speed" mosfet driver, but then later in the text is says "High Side" mosfet driver. Perhaps they introduced a transcription error there.

I also notice that it has a "low voltage lock out" feature which should have protected it against the dubious test procedure in which the current limit was gradually increased. HOWEVER, if you follow the logic of the lockout it is clearly designed for n-channel mosfets, so it actually functions as a low voltage lock ON with the p-channel device!

This means that when the op gradually wound up the supply current limit, it pretty much guaranteed that the mosfet would go into linear mode at some point. This is exacerbated a bit by the lamp load in which a fair proportion of full load current can flow at reduced voltage. So this is also a possible cause of it's demise, certainly if left long enough in this mode.

MarkT

#7
Mar 15, 2017, 03:04 pm Last Edit: Mar 15, 2017, 03:06 pm by MarkT
Can I suggest a 2N2222 or similar, driving a 1k resistor to the gate of the MOSFET, and a 12V zener
in parallel with another 1k resistor between gate and source.  Unless your PWM is very fast that might be a simple fix.

For a true high side driver you'd need n-channel MOSFET and charge pump components as well,
and a 12V auxiliary supply for the driver.
[ I will NOT respond to personal messages, I WILL delete them, use the forum please ]

Jiggy-Ninja

I noticed that too Mark. In the title it says "High Speed" mosfet driver, but then later in the text is says "High Side" mosfet driver. Perhaps they introduced a transcription error there.
I don't think so. While it is clearly not a bootstrap driver that would be used for a high-side N-channel, the datasheet only says it is "suitable" for high-side driver applications. Presumably it would be used for a P-channel FET with something like 24V.
Hackaday: https://hackaday.io/MarkRD
Advanced C++ Techniques: https://forum.arduino.cc/index.php?topic=493075.0

stuart0

#9
Mar 15, 2017, 03:51 pm Last Edit: Mar 15, 2017, 03:52 pm by stuart0
I don't think so. While it is clearly not a bootstrap driver that would be used for a high-side N-channel, the datasheet only says it is "suitable" for high-side driver applications. Presumably it would be used for a P-channel FET with something like 24V.
Yes, within the devices limitations you can use it however you want.

As I pointed out before however, it's inbuilt under-voltage lockout is completely incompatible with p-channel high side switching.

Jiggy-Ninja

As I pointed out before however, it's inbuilt under-voltage lockout is completely incompatible with p-channel high side switching.
You are correct, I missed that little subtle detail. Doesn't explain how the driver and Arduino died, but I think I have an idea.

If the Arduino was powered from an independent power supply (like USB), then it would be applying power to the Input pin. This would forward bias the ESB protection diode that's bridging VDD and Input while the VDD voltage was being slowly ramped up. Semiconductors tend to fail short, so if the forward current through the diode was enough to damage it but not enough to fuse it open, it would increase the reverse leakage current so that when VDD finally went higher than the input voltage, it would backfeed current into the Arduino pin.

So the question for saximus is, what was powering the Arduino during this test? The bench supply, or something else?
Hackaday: https://hackaday.io/MarkRD
Advanced C++ Techniques: https://forum.arduino.cc/index.php?topic=493075.0

saximus

#11
Mar 15, 2017, 10:49 pm Last Edit: Mar 16, 2017, 12:25 am by saximus
Hello all, sorry for the slow response. Apparently the iffy email alerts failed me on this one. Thank you all so much for the replies.

Tom and Jiggy, the Arduino was powered through USB and yes the grounds were connected. I had a feeling it might would be a failure in that protection diode but didn't even think about overcurrent while VDD was ramping up. 

Mark, I'd completely overlooked the threshold Vgs as it was hard enough just finding one that could handle the 28V. So yep that's a fail but, if this counts as a defence, the MOSFET is still working fine.

I also obviously don't understand enough about the ins and outs of drivers because the comments about what to look for in a high side driver have gone over my head.

Funnily enough I originally designed an NPN driver for him but was never totally happy with it because most people on here advocate for the purpose built drivers and I didn't like limiting the gate current to what the BJT could handle. Maybe we'll go back to that option for the time being until I can learn a bit more driver selection.

As a slightly sideways question, is there any advantage/disadvantage to using an n-MOSFET driver instead of an NPN? I would imagine a sufficiently low Rds would give me more current driving ability with lower temperature?

TomGeorge

Hi,
N-CH MOSFET would be fine instead of NPN transistor, the low Rds being advantageous as well as lower drive current needed from the controller.

Tom.... :)
Everything runs on smoke, let the smoke out, it stops running....

MarkT

BTW do you have to use high-side switching?
[ I will NOT respond to personal messages, I WILL delete them, use the forum please ]

allanhurst

#14
Mar 16, 2017, 03:32 pm Last Edit: Mar 16, 2017, 03:43 pm by allanhurst
The TC4431 maximum peak siink output is 1.5A, and output resistance about 10 ohms.

If you switch the large capacitance of the gate with a 28v supply directly , an instantaneous 2.8A will flow.  The TC4431 may not like this.

Suggest you add say 2 x  27 or 33 ohms in series with the gate seperately to the  up and down drivers to limit this peak current to about 1A . Don't connect the up and down drivers together.

The switching will be slower, but  plenty quick enough.

And this method avoids any high breakthough currents ( ie if the conductions of the up and down drivers overlap)  within the chip as the output devices switch . This may be why Microchip provide these outputs seperately.

With these high fast drive currents I hope you're adding plenty of decoupling capacitance on the supply close to the mosfet - both low ( 100n) and high ( 100uF) values would be a good idea.

The inductance of wire is in the region of 1/2 nH / mm , so 100mm of wire is about 50nH. With a switching transient at about 1A in 20nS, this implies a voltage spike of L di/dt or 2.5 volts.. ( do the sums). So also keep all wiring very short.

The gate/source  drive limits of the mosfet  are  +/-30v, so you're pushing it, but the mosfet survived, so ( just) OK.

You could always add a 18v 1A zener in series with the 'low' driver and dispense with the resistors - this would also reduce the strain on the gate/source.

Allan

see attached for the 2 options

Go Up