3
\$\begingroup\$

I'm having trouble with TC4421A MOSFET drivers in a new PWM DC Heater design blowing. The 12V power supply used during testing is capable of 25A. While doing my initial testing, I started off with higher value load resistors with a 5 kHz PWM duty cycle of 25%. I successfully tested the design at up to 4A. That's when the TC4421A driver failed and literally caught fire. The MOSFET was not damaged. The PWM is generated by a PIC with a Vdd of 5V. My test setup had approx 2 feet of wire connecting the dc supply. Not the most ideal setup, but I didn't expect the driver to fail. The P-Chan MOSFET is rated at 80A and the driver is rated at 9A peak. Also, the 100 ohm gate resistor was not damaged nor was the PIC. I'm hesitating with any further testing until I at least make some attempt to keep the driver from blowing up. My ultimate goal is to be able to deliver up to 100W into a resistive load by varying the PWM duty cycle.

Thanks in advance for any advice.

enter image description here

\$\endgroup\$
  • \$\begingroup\$ "I'm <s>having trouble</s> blowing TC4421A MOSFET drivers in a new PWM DC Heater design." \$\endgroup\$ – Fizz Dec 6 '15 at 17:20
  • 1
    \$\begingroup\$ I want to see scope traces of the 12V rail, PWM, and output. Ideally at the same time on a 4 channel scope. \$\endgroup\$ – Matt Young Dec 6 '15 at 17:33
  • 1
    \$\begingroup\$ My first test was with a short across the 100 ohm gate resistor with the though that the relatively high gate capacitance might keep the driver from switching the MOSFET fast enough. That test ended up blowing both the MOSFET and the driver. The whole purpose of using a fast, high current driver is to switch the MOSFET as fast as possible to keep from overheating at the expected higher current levels. Adding a gate resistor is only going to slow the switching time. Might as well just switch the MOSFET with an NPN... \$\endgroup\$ – MotoDan Dec 6 '15 at 18:03
  • 1
    \$\begingroup\$ Not that it should matter in this circuit, but why not use an N-channel and take advantage of the better device performance? \$\endgroup\$ – Daniel Dec 6 '15 at 18:26
  • 1
    \$\begingroup\$ Regarding the temperatures, the MOSFET runs cool (so far) and the driver never had a chance to heat up as it blew immediately when the 4A load was applied. I've increased the wire size in my test setup and the ringing looks much better. Will try increasing loads and see what happens... \$\endgroup\$ – MotoDan Dec 6 '15 at 19:58
3
\$\begingroup\$

Quite possibly it's the two feet of wire that is forming an inductor of somewhere between 100nH and a couple of uH depending how your wire is laid across the bench.

With a pulse of 4A and 1uH the stored energy is 8 uJ and if this was dumped into the 4.7 uF capacitor when the load was turned off, the terminal voltage would rise a couple of volts BUT if the capacitor were an electrolytic with maybe an ESR of one ohm you could be starting to see voltages on the line to the chip that destroy it. The TC4421 has 20V as its maximum supply voltage.

Try twisting the power wires together and maybe also try increasing the capacitance.

Looking at your circuit, you have a 2 ohm load so the peak currents would be 6A and this of course means more voltage on the 4u7 capacitor so that would be more like a 3V glitch on the supply taking it to 15 volts. It's getting close! Each time the voltage gets bigger the current taken by the load gets bigger and maybe there is some point when this spirals out of control?


EDIT

My first test was with a short across the 100 ohm gate resistor with the though that the relatively high gate capacitance might keep the driver from switching the MOSFET fast enough. That test ended up blowing both the MOSFET and the driver.

The MOSFET in your schematic is rated at 16V gate-source voltage max: -

enter image description here

With the ringing and glitching as I've explained above, this is the likely failure mechanism for the MOSFET - puncturing the gate through to the source will be also pretty fatal for the driver too.

\$\endgroup\$
  • \$\begingroup\$ The 4.7uF cap is an SMT ceramic. I think there's a good possibility that the TC4421 may have seen something in excess of 20V with all of the potential ringing on the 12V supply line. My next attempt will be to shorten and increase size of the DC supply wiring. Will also scope the 12V at the driver to see what's actually going on. \$\endgroup\$ – MotoDan Dec 6 '15 at 17:57
  • \$\begingroup\$ I'm considering adding 14V zeners across the gate as well as the supply into the TC4421 driver - just in case there are any transients. I really can't risk these types of failures once we're in production. \$\endgroup\$ – MotoDan Dec 6 '15 at 19:26
  • \$\begingroup\$ That is going to help a lot I reckon but make sure the layout is tight and the 100nF cap nicely up at the legs of the chip. \$\endgroup\$ – Andy aka Dec 6 '15 at 20:30
  • \$\begingroup\$ He said the FET wasn't damaged \$\endgroup\$ – Russell McMahon Dec 7 '15 at 0:26
  • 1
    \$\begingroup\$ Ah - Schrodinger's FET! \$\endgroup\$ – Russell McMahon Dec 7 '15 at 12:57
2
\$\begingroup\$

OK, so to keep from roasting any more TC4421 drivers, I dialed down the 12V supply to 10V. Keep in mind this is a 350W switcher. In the scope trace below, the top (blue) trace is the load resistor which is now 1.5 Ohms, the lower (yellow) trace is the 10V supply measured at the PCB which is 8" (18AWG) from the supply. The FET turns on immediately, but dips about 35V below ground with 10-15 uSec of ringing. The PWM is running at 5 kHz at about 15% duty or about 1.5V average to the load resistor. The average current is only about 1A so the FET is loafing and should be fairly cool. The 80A FET is starting to get pretty hot which I'm sure is due to the turn off behavior. The 2 uS 5V bump in the supply line is probably due to the switching supply reacting to the load dump. I don't really need to run the PWM so fast since I'm just heating up a resistive load and may reduce the frequency to 100 Hz or so. This won't help the MOSFET turn off behavior which is my current concern due to the overheating. In the end, I will be drawing over 10A at max power to the heater. The heater will not be run continuously, but will be on for at least 15-20 sec at a time with about two minutes between so the duty will be around 15%. I'm pretty sure the MOSFET won't survive this given that it's getting hot with only a short 1A load.

enter image description here

\$\endgroup\$
  • 1
    \$\begingroup\$ Show us also the input logic signal and the gate voltage. \$\endgroup\$ – SunnyBoyNY Dec 8 '15 at 0:27
  • 1
    \$\begingroup\$ This should probably be included in the question - you ARE allowed to answer your own questions but this (excellent material) mainly expands the question related information. \$\endgroup\$ – Russell McMahon Dec 8 '15 at 5:41
  • \$\begingroup\$ Where is your scope gnd lead connected to. What does the trace look loike with scope ground AT the FET source? \$\endgroup\$ – Russell McMahon Dec 8 '15 at 5:41
  • \$\begingroup\$ The probe grounds for these traces was a the board GND near the MOSFET and driver. \$\endgroup\$ – MotoDan Dec 11 '15 at 17:32
0
\$\begingroup\$

From the Microchip data sheet:

These devices are essentially immune to any form of upset, except direct overvoltage or over-dissipation. They cannot be latched under any conditions within their power and voltage ratings. These parts are not subject to damage or improper operation when up to 5V of ground bounce is present on their ground terminals. They can accept, without damage or logic upset, more than 1A inductive current of either polarity being forced back into their outputs. In addition, all terminals are fully protected against up to 4 kV of electrostatic discharge.

Are you sure you wired it up correctly ? Do you have a genuine TC4421A ?

\$\endgroup\$
  • 1
    \$\begingroup\$ The wiring is correct and the TC4421A is believed to be genuine since it was purchased from Digikey. \$\endgroup\$ – MotoDan Dec 6 '15 at 17:37
  • 1
    \$\begingroup\$ It would have blown way before 4A if wired wrong. \$\endgroup\$ – Fizz Dec 6 '15 at 17:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.