Mix matched FETs controller repair - what can go wrong?

[matt912836]

I've had a Sabvoton 96100 and 96120 both sitting around for years because each had a blown set of fets

Rather than order a new set of matching ones, I took the impatient route of taking a set of working ones from one and using them to replace a set of blown ones in the other.

The 96100 has 8 irfb4115's per phase and the 96120 has 10 irf135b203s per phase.

I took 8 working irf135b's from the 96120 and replaced the 8 blown irfb4115's in the 96100. It now has 16 irfb4115's working on two phases and 8 135b's on the last phase.

Before replacing the fets the controller registered as shorted at the positive and negative on a voltmeter and could not even be plugged into any power source. After replacing the fets, the controller powers on and spins the wheel just fine. Haven't taken it out for an actual ride to give it a load but on the bench spinning the weight of a heavy dirt bike wheel seems to be working just fine, including running flux weakening and slide regen.

What possible problems may I run into, assuming all the remaining FETs are healthy and aren't compromised at all from the first initial failure? Both FETs work at the same voltage and the 135b's are actually capable of more amps. As long as I remain within amp limits of the lower spec original fets, will this repair work just fine in the long run?

 

[amberwolf]

Probably...it's been done successfully, but:

FETs have a couple of "critical" characteristics that matter for this.

The RDSon of the lowest ohms FEt(s) in the parallel set will pass the most current. Because it's lower resistance it shouldn't get much hotter than the others, but if it does, and still keeps passing more current, it will keep getting hotter than the others. If the current sharing is really poor, most of the load could be taken up by just these FETs, leaving them more prone to failure.

The fastest rise and fall times of the FEt (how fast it turns on and off) will also take the geater part of the load, but will heat up less during those times than the others do. But again, if these FETs take most of the load, they could be more prone to failure if they can't handle that load.

They *might* just all average out, but things switch very quickly and current flows and voltages thus change quickly, so exactly how it will behave, I couldn't say. Could probably be modelled in circuit software if ti had all the parameters of the controller parts, motor, and battery, and the load on the motor could be modelled.


So, the less ahrd you are pushing the controllers the less worry there is...the harder they're pushed, well...


The ohter part of the quesiton is, what caused the original failures?

If it was an external fault (short, open, environmental, etc), you can avoid those and prevent them from triggering another failure.

If it was an internal controller issue (manufacturing problem, software bug or just poor sw design, hardware design flaw, etc), then it can still cause the same failure.

 

[matt912836]

Probably...it's been done successfully, but:

FETs have a couple of "critical" characteristics that matter for this.

The RDSon of the lowest ohms FEt(s) in the parallel set will pass the most current. Because it's lower resistance it shouldn't get much hotter than the others, but if it does, and still keeps passing more current, it will keep getting hotter than the others. If the current sharing is really poor, most of the load could be taken up by just these FETs, leaving them more prone to failure.

The fastest rise and fall times of the FEt (how fast it turns on and off) will also take the geater part of the load, but will heat up less during those times than the others do. But again, if these FETs take most of the load, they could be more prone to failure if they can't handle that load.

They *might* just all average out, but things switch very quickly and current flows and voltages thus change quickly, so exactly how it will behave, I couldn't say. Could probably be modelled in circuit software if ti had all the parameters of the controller parts, motor, and battery, and the load on the motor could be modelled.


So, the less ahrd you are pushing the controllers the less worry there is...the harder they're pushed, well...


The ohter part of the quesiton is, what caused the original failures?

If it was an external fault (short, open, environmental, etc), you can avoid those and prevent them from triggering another failure.

If it was an internal controller issue (manufacturing problem, software bug or just poor sw design, hardware design flaw, etc), then it can still cause the same failure.

The original problem was caused when I was testing weather I bled my brakes properly, holding down the brake and applying throttle to see if it could lock it up. Normally any other time that would cause an overcurrent at most, but that time it decided to cause a popping sound and the controller failed to shorted on the positive and negative. Thankfully the battery was fused. Controller is unlocked so you can enter any parameters, I definitely had it turned up past stock when it did so. The same result can happen if you don't have the hall phase combination properly figured out and you give it too much power testing it out, which is what I believe happend to the sacrificed controller.

But it seems to have just been the FETs that failed, because replacing them even with a different type seems to have been able to bring it back to life! Just came back from the maiden voyage and the controller runs seemingly flawlessly! It's a shame I didn't try this earlier.

I may also try to replace all the ones I took from the sacrificed controller along with the ones which are damaged and bring that one back to life too, and even throw the originals into this one, once Im ready to place a nice order for brand new FETs. These high voltage sabvs still go for $400+ each so just getting one back up and running was more than worth the experiment, throwing no more than a hundred bucks into some brand new FETs seems like a good deal to me if I can get both back into original condition!

 

[amberwolf]

The original problem was caused when I was testing weather I bled my brakes properly, holding down the brake and applying throttle to see if it could lock it up. Normally any other time that would cause an overcurrent at most, but that time it decided to cause a popping sound and the controller failed to shorted on the positive and negative. Thankfully the battery was fused. Controller is unlocked so you can enter any parameters, I definitely had it turned up past stock when it did so. The same result can happen if you don't have the hall phase combination properly figured out and you give it too much power testing it out, which is what I believe happend to the sacrificed controller.

"Oops" Unfortunatley even with factory limits on many controllers you can still pop them due to poor firmware or hardware design (or bookkeepers deciding to use cheaper parts in critical areas to save a bit of $$ per unit).


I blew up a "methods" 18FET in a similar fashion over a decade ago on CrazyBike2 trying to basically "rooster tail" the motor wheel in the front yard dirt, but traction was just too good and KAZOWIE goes the controller. But it was not fixable by changing just the FETs.

But it seems to have just been the FETs that failed, because replacing them even with a different type seems to have been able to bring it back to life! Just came back from the maiden voyage and the controller runs seemingly flawlessly! It's a shame I didn't try this earlier.

That's good--I've seen the gate drivers destroyed, and at least one that took out stuff all the way back to the MCU output for that bridge.

I have tried to repair several controllers by swapping known-good FETs over, only to run into those types of problems. (the gate drivers are usually surface mount tiny parts, whcih I am not very good at working with).

I may also try to replace all the ones I took from the sacrificed controller along with the ones which are damaged and bring that one back to life too, and even throw the originals into this one, once Im ready to place a nice order for brand new FETs. These high voltage sabvs still go for $400+ each so just getting one back up and running was more than worth the experiment, throwing no more than a hundred bucks into some brand new FETs seems like a good deal to me if I can get both back into original condition!

It would be a good deal to do it if you ahve the time...and if one of them doesn't work you still have all those as spare parts for "next time".