IRF460 Failures

[ScottI2R]

Hello Everyone,
I am experiencing continuous IRF460 failures in an EDM generator. This results in catastrophic pcb damage. This device replaces the previously used IRF340 because of its larger current handling capability. It is used in a closed loop high voltage regulator (PWM driven) configuration. Basically, 300VDC in, max 255 VDC out to be supplied to a pulse board which switches the requested o/p voltage via 3 parallel irf 830's (at freq's as low as 20 nanoseconds). It appears that the sequence of destuction occurs in this order: First the IRF460 shorts (this is the common mode of failure, I have NEVER found one that opens). At this point, the entire DC supply is allowed to pass through a 1 ohm 2 watt non-inductive shunt resistor, which quickly gives it up and opens. Then the supply side must seek a different route back to its return, and it finds that the sensing circuitry is its best option. This is when the pcb gets burnt, scorched, and otherwise rendered "unrepairable".
I have resolved to using 2-amp pico fuses to limit the supply current in the event of the IRF460 failure. This will prevent the loss of the shunt and prevent the destruction of the pcb. However, there is still no solution for blowing the IRF460. The device is socket mounted and has a vertical rib heatsink mounting. Thermal compound is used and the following is the operating parameters: 1) Max Iout is 1.25A (due to Ilimit of the pulse board) 2) 4.0uSec on time / 7.0uSec off time.
I do not know the maximum current the non-regulated supply is capable of. We do have significant issues with higher W shunts, as this is pcb mounted and inductance IS a factor. Please note that these parameters are for the final output of the EDM generator, and NOT what the IRF460 is running at. In essence, I believe that the heat sinking and cooling is NOT sufficient for the device. The Tj max is 150 degrees C and at the current we are running, this would fail around 145+ Deg C. (Unfortunately, this is a customer unit and these are all the test I have so far. This unit is being run at its maximum capability as well.) Hmmm...Who WAS that salesman? We have larger units that will DO this application without trouble. (Sorry, had to vent).
Any suggestions anybody?

 

[OperaHouse]

I had this problem with an EDM from China using 4049 driver boards, they turned to toast. The problem there was resistance from bad connections in the current path that forced current to the board. Personally, I would never use sockets with any high current path. Is the high speed kickback diode still good? How many do you have of these in parallel and what is the max output of the machine. Are you sure the power supply doesn't fail first causing overvoltage and then output failure.

 

[jimkirk]

When you substituted the higher current devices did you modify the circuitry driving the gates? Input capacitance for the IRF460 is about three times that of the IRF340. This could slow switching times, maybe even cause loop stability issues.

 

[VEBill]

This is not likely, but worth mentioning on the off-chance:

http://www.google.com/search?q=irfp460+counterfeit

Note - reports are concerning IRFP460 (with a P).

 

[OperaHouse]

You beat me to it, I was just checking the specs when forced to go out to the hot tub for 45 minutes. Bigger isn't better and you don't get something for nothing. Think he fell into the one size replacement fits all trap. While he may not have trouble with the high current section of of the EDM that only operates at 60-80V, it is much harder to switch 250V tickler fast without a lot of heating. That gate capacitance is a killer.

 

[sreid]

The statement that the FET is running hot supports the theory that the switching is slow. This, of course, is always (?) caused by inadequate gate drive.

 

[IRstuff]

There may also be an issue with drain-gate Miller capacitance. A larger current device is often implemented as one with a wider channel and higher absolute overlap capacitance. With any sort of useful gain in an inverting application, that overlap capacitance will also slow down the gate drive.

TTFN

 

[ScottI2R]

Thanks Everyone for the advice. Now in reply to some of your questions:
OperaHouse: the max output is about 1.3A and all the connections were OK. It does use 4049 (6)from the same pkg and these drive a 2n4123 and a 2n5401 (in a push pull configuration). These two xistors directly drive the IRF460 thru a 470 ohm gate resistor. There is only one IRF460 in use in this generator. Oh, and the snubber diodes are fine.
JimKirk: No, I did not modify the drive ckty. (OOPS!) I guess this is what happens when you are forced to perform in a "reactive" mode as opposed to a "proactive" mode. After all the Engineers were "Let go", these kind of issues all come my way. I am a Field Service Engineer and I have not done any major design work since college. I DO enjoy it though! Now, if only I can get a workorder to assign this to!!! Thanks for the help! I appreciate it.
IrStuff: Noted. Now all I have to do is figure out the impact of the Miller capacitance and look up the difference between the 460 and the 340. Back to the books I guess...Thank you for the help.

Ok, another question if I may. Assuming the push-pull transistors will handle the additional current, can I reduce the gate resistor to an amount where the original time constant is the same? ( essentially 3x smaller than the original since the capacitance is 3x greater. I do not have any specs for the transistors in front of me at this time.) Or, should I just practice asking," Would you like Fries with that?"

A quick line up of this ckt. is as follows: 8 bit DAC drives diff amp MC1458 which feeds voltage sense comparator TL494. This in turn drives a HCPL2601 which drives all 6 4504's (single pkg) and this drives the 6 4049's which drives the 2 transistors which drive the FET. Which then burns up, which in turn, drives me to drink!

That is in a nutshell everyone. BTW, the DAC-08 is supplying the requested (by the operator) voltage to the sense circuit. This in turn drives the circuit as required to maintain the needed output voltage.

Thank you again to EVERYONE for your help. I am sorry it took a week to reply but I have been incredibly busy as of late.
Thanks,
Scott

 

[sreid]

Yes, the 470 ohm gate resistor is limiting the the current to charge the gate capacitance and magnifies the Miller Effect. This may have been the reason for failures with the original FET. Check the gate drive transistors' current rating but 100 ohms max would likely be more appropiate.

 

[ScottI2R]

Thank you Sreid. As soon as work allows me to tackle this issue, I will take care of the problem once and for all (I hope!)As for now, I will go google and see if I can find the current limits for those driver transitors. Thank you again!
Scott

In a hundred years, it isn't going to matter anyway.

 

[OperaHouse]

Look at ZTX718 & ZTX618 complementary drivers with a B-E resistor of about 15 ohms. These can really drive but watch the voltage. Only transistor I've seen that is graded in couple volt increments. We were driving just off the 4049 pack, no drivers, with just 20 ohms.