A short story on replacing bad FETs and a warning about it

[zombiess]

I have a general rule which I've only ever had to apply twice luckily. That rule is to replace the entire bank (high or low side) at the same time with all new FETs if one blows, not just the single blown FET. I had a EB212 12 FET controller that I was unable to save due to it exploding the last phase bank furthest from the battery feed and also melting a leg from one of the electrolytic capacitors (huge current ripple), it also took out the driver section for that bank so the board was salvage. When I pulled the FETs off the board I thought I had 10 of the 12 FETs that were good, but it turns out only 8 were good. Two exploded and got kinda melty, pretty obvious they were no good. The other two on the opposite side of the blown phase tested out fine with a multi meter both in circuit and when removed and gave the same readings as a healthy IRFB4110. It wasn't until many months later when a friend needed some FETs for a project he was working on and I was giving him several to use that I removed from this blown controller for him that I actually found these two more were bad. This was part of his senior project for his BS in mechanical engineering (but he needed some electronics to drive some solenoids). Because I knew how important it was I decided to test all the used FETs I was giving him just to be sure I wasn't going to tank his project since he was in a rush finish. I rigged up a simple pulse generator with a PIC and had an LED blink triggered by the FET under test. All but two FETs tested perfect. The two that didn't test right would always trigger the LED on even when the gate signal was driven high or to ground, but would turn the led off if the gate was left to float. These were the two from the other side of the same phase that went pop.

Don't know how common this is, but I figured it was an experience worth posting with the current trend of people blowing up and fixing controllers lately.

 

[Alan B]

What procedure do you use when testing FETs with a multimeter?

 

[zombiess]

Compare them to a known good FET. Many blown ones show a short between drain and source, but it depends on how it failed. FETs have many different failure modes. Most people on here seem to experience the plasma failure.

 

[dnmun]

do you have a theory why they all blow at the same time?

 

[liveforphysics]

I bet the loss of that local cap (leg melting off) allowed large enough spikes that those other FETs died from spikes rather than thermally (how we normally kill FETs).

 

[Alan B]

A voltage transient on the FET would generally cause a short which is easy to measure with a multimeter. A high current surge would tend to open the device, either with major evidence of high current and heating/melting/etc damage, but in some cases they manage to blow open without much or any external visible sign.

In the case of my 24 FET that has failed twice, it operated fine for 15 miles, then failed on initial acceleration under load with no heat or visible damage. One FET was shorted, this was replaced, and the unit was thoroughly tested. It worked fine on the bench under test, even with a 33 pound wheel with Cromotor in a 17" moped rim with 3" tire, for many test accelerations no load (just the load of the wheel and tire). As soon as the wheel was on the ground the controller popped immediately on the first gentle start.

I don't know yet what is wrong, but it would seem that some FETs are not sharing current but none are shorted, so that testing proceeds fine but when a load is added and phase currents are higher and last longer the FET carrying the load fails.

In this case a 12 FET set for pretty much the same battery current, and about half the phase current is working just fine, and no load currents are quite low. So it would tend to indicate that the motor and wiring are okay.

Changing all the FETs in that side of the bank sounds prudent, but it is hard to know which other FETs were also stressed during the failure. It would be good if there was a more thorough way to test all the FETs while in circuit.

 

[ZOMGVTEK]

I've replaced FET's in various controllers about 10-20 times, and have always replaced only the ones that went short or melted. I'm not saying this is the BEST method, but it worked out fine for me.

I was running 11-12kW into a 12 FET 4110, until it blew when I was repeatedly holding the motor stalled for 5-10 seconds while I was tweaking the current. It handled it about 5 times in a row, until it just went poof. Only one FET was bad, sitting at a few ohms. I swapped it out with a fresh one and that controller has hundreds of miles on it at 8-9kW.

It probably helps that most of the time I cook FET's, the bike isn't moving very fast. Having all the energy of a decelerating bike going into a shorted FET would be bad.

 

[dnmun]

was that the hiside FET?

i assume that from your statement about it happening while stalled.

but he blew the loside FET's.

should be a totally different cause than losing the hiside gate voltage.

 

[ZOMGVTEK]

Yeah, that one was high side.

I suppose most of my failures were high side, but I did have a few low side failures and even a whole phase failure that trashed the board, but that one was on 48S and at high wheel speed.

If its a controller I'm fixing for someone else, I usually just shotgun the channel with new FET's. If it's one of my controllers, I will put back any FET's that measure good and just replace the bad ones.