Replacing Controller MOSFETs (Sabvoton72150)

Joined
Apr 6, 2021
Messages
16
Hey guys! I'm new here but I joined after reading several threads relating to my issue, and I could use some more help. I replaced an entire bank of FETs in my controller after a torque arm failure caused phase wire damaged, shorting out all my halls and one whole phase bank (blue).I repaired the harness and replaced all the halls is my QS205 V3 before I realized the FETs where blown also. So after replacing the 8 FETs in the blue bank with some phat overrated bois (like $5 a piece) I actually got the motor running awesome! All the way to 115kph! But only under no load....as soon as I flip it over and sit on it to test it either over amps and shuts off or if I give it some help it runs but cogs bad and is weak. I read something about someone having the same issue that had some water damage to just one FET In their controller, and that made me think that maybe one or 2 of the FETs in another bank ( I think the green one because the blue and green wires had broken and had burn marks on them) may also be blown out and thats why it can't handle the amp load with weight... what do you guys think? I'm ready to pull the trigger on more of these FETs and replace the rest...I'd like to hear from some of you experts! Thanks in advance.
 
The problem with blown FETs is that it's not uncommon for the gate drivers to get damaged or destroyed during the FET failure (battery voltage can get leaked into the gate pin, then back thru the drivers).

You can also get FETs that are just damaged and not blown, so they appear to be ok until they get pushed harder, and then they act different (or fail then).

Also, if the new FETs are not identical to the old ones (not just hte same part number and manufacturer, but also the same characteristics, which typically means from the same batch), then you could have differences between sets that creates differences in current flow that causes unwanted behavior. That's typically just with different FETs in the same half of the same phase, but some poeple have had trouble even if it's a whole phase's FETs that were replaced with slightly different ones and the other two phases were "ok". Most gave up at that point so I don't know if they had other unknown issues contributing to it....

Another problem is that with motor wire damage, phases and halls often get shorted together, and sometimes the 5V supply does too (often there's a diode in series with this inside the controller, so the 5V line may remain undamaged...but sometimes not). If the 5v line was damaged, everything that's on it (including throttle, PAS sensor, MCU, 5V regulator, etc) may have damage that may be obvious or may only happen under certain circumstances.

If the halls get shorted with the phases, not only can the halls in the motor be destroyed, but the MCU hall inputs can also be damaged or destroyed, which leaves the motor unable to correctly read motor position. Some controllers "fallback" to either a two-hall mode, interpolating the third, if at least two still get valid signals, and some fallback to sensorless. Some don't have a fallback, so may either not operate at all, or may run rough or not be able to start under even a small load from some wheel positions. (and may be noisy at start even without load from the same positions).
 
amberwolf said:
The problem with blown FETs is that it's not uncommon for the gate drivers to get damaged or destroyed during the FET failure (battery voltage can get leaked into the gate pin, then back thru the drivers).

You can also get FETs that are just damaged and not blown, so they appear to be ok until they get pushed harder, and then they act different (or fail then).

Also, if the new FETs are not identical to the old ones (not just hte same part number and manufacturer, but also the same characteristics, which typically means from the same batch), then you could have differences between sets that creates differences in current flow that causes unwanted behavior. That's typically just with different FETs in the same half of the same phase, but some poeple have had trouble even if it's a whole phase's FETs that were replaced with slightly different ones and the other two phases were "ok". Most gave up at that point so I don't know if they had other unknown issues contributing to it....

Another problem is that with motor wire damage, phases and halls often get shorted together, and sometimes the 5V supply does too (often there's a diode in series with this inside the controller, so the 5V line may remain undamaged...but sometimes not). If the 5v line was damaged, everything that's on it (including throttle, PAS sensor, MCU, 5V regulator, etc) may have damage that may be obvious or may only happen under certain circumstances.

If the halls get shorted with the phases, not only can the halls in the motor be destroyed, but the MCU hall inputs can also be damaged or destroyed, which leaves the motor unable to correctly read motor position. Some controllers "fallback" to either a two-hall mode, interpolating the third, if at least two still get valid signals, and some fallback to sensorless. Some don't have a fallback, so may either not operate at all, or may run rough or not be able to start under even a small load from some wheel positions. (and may be noisy at start even without load from the same positions).

Awesome! I really appreciate the response AmberWolf. The nice thing about my adventure in making these repairs is that my buddy at work is an electrical engineer. I'm just a talented welder/fabricator but can learn to fix anything. I'm NOT going to give up on this controller, as even if I replaced the 16 remaining FETs, and each gate driver for each phase, would still be much cheaper than purchasing new motor and comtroller.

I know the sabvoton won't run sensorless so that still leads me to believe there is another FET or 2 that are damaged, would that be safe to assume?

My buddy says they cannot be tested in circuit so ill have to pull them all. I'm also wondering if there's any specific way to tell that the driver is bad or if i should just order those as well since they are pretty cheap?

My buddy also looked at the data sheet for the FETs we got and said I shouldn't have any compatibility issues replacing the whole phase, he actually picked the replacements. He said it may just handle the load a little better even. (This is just his opinion, im not saying anyone is right I just like to hear different opinions.)

I also showed him a reply that you (I think, maybe one of the other Forum Daddy's here.) posted to someone who had sent you a blown Sabvoton to take apart and you were telling about how they are just clones now with way too little capacitance and very weak FETs and other components, he said it was bound to fail even if my cable didn't short out, so I'm not opposed to doing more surgery on this big heat sink of a board!

Lastly, I have 2 sets of halls that I replaced, one stopped working completely after some of these tests and over current faults, and the other set has 3 halls that test perfectly on the oscilloscope and still works and makes it operate the way it does now.

Edit... these are the FETs I'm using (that I put in the whole blue phase) IPP041N12N3GXKSA1

Any words of encouragement to keep going, friend?
 
TurboWelderMonkey said:
I know the sabvoton won't run sensorless so that still leads me to believe there is another FET or 2 that are damaged, would that be safe to assume?
Probably that or the gate drive on the repaired phase.

I had a controller I was experimenting with that something happened and my hand on throttle moved and the motor jerked and caused the board to land on a metal tool, and it damaged the controller in a way that even after replacing blown FETs left it "working", but not like it used to.
https://endless-sphere.com/forums/viewtopic.php?f=4&t=21830&p=320682#p514878
file.php

https://endless-sphere.com/forums/viewtopic.php?p=515777#p515777
I never got around to completely fixing it as it wasn't that great anyway, though I still have it and sometimes think about using an oscilloscope to check out the gate drive of each phase relative to each other to see if that's the cause...but I never have enough time/energy for even the stuff that's critical.... It coud also be that the FETs used were different in characteristics. There's video and scope pics and stuff if you're interested, in the last link.


My buddy says they cannot be tested in circuit so ill have to pull them all.
They can be tested in circuit as a group (whole top or bottom of each phase), but not individually. So for instance you wouldn't know if just one was causing a problem, or not working at all, just if the whole paralleled set was wonky.


I'm also wondering if there's any specific way to tell that the driver is bad or if i should just order those as well since they are pretty cheap?
To test them, you'd probably need an oscilloscope to see the waveforms, since you already know they are driving the FETs, just that something is not happening as it is supposed to. You'd probalby see something differentn between the undamaged phases and the damaged ones.

It probably won't hurt to just replace the gate drivers of the damaged phase before doing the FETs in all the phases, just to see if it makes any difference. But if you want to see wha'ts going on and learn even more, and have access to even the most basic oscilloscope (there are cheap "front ends" available online that connect to your phone using OTG/USB, for instance, though I've never used any of them; also cheap old standard benchtop used ones on ebay and the like), you could see the waveforms and even post pics of the results. :)


My buddy also looked at the data sheet for the FETs we got and said I shouldn't have any compatibility issues replacing the whole phase, he actually picked the replacements. He said it may just handle the load a little better even. (This is just his opinion, im not saying anyone is right I just like to hear different opinions.)
Most likely the different FETs aren't teh cause of the problem--it's usually only if the Qg (gate charge) or RDSon is significantly different from the others taht you get different behavior enough to cause a problem. Usually it doesn't matter...but there are a lot of different kinds of characteristics of a FET, and sometimes certain ones are needed for certain applications...I am not expert enough to help with those, but you can read about them in various custom motor controller and powerstage design threads by those that are. :)

A big enough difference in gate charge means the FET may turn on or off with enough difference in timing to the others that it creates problems, or that it takes more than the gate driver can provide (especially if it's already damaged).


I also showed him a reply that you (I think, maybe one of the other Forum Daddy's here.) posted to someone who had sent you a blown Sabvoton to take apart and you were telling about how they are just clones now with way too little capacitance and very weak FETs and other components, he said it was bound to fail even if my cable didn't short out, so I'm not opposed to doing more surgery on this big heat sink of a board!
Was probably someone else; I don't think I've had a sabvoton here yet, but there have certainly been enough blown ones posted about. ;)
 
Hell yea man! I just got 15 more of those really good FETs today, and I have gate drivers arriving Monday, but they are the wrong ones, lol I ordered 2112s and they are 2110s. My buddy looked at the sheets and said they may work in place but I better get the right ones. I'm going to test the original ones today before I order the correct ones, and I have tons of great equipment and several ociliscopes here at my work shop so yea ill post pics of the results!!!

amberwolf said:
TurboWelderMonkey said:
I know the sabvoton won't run sensorless so that still leads me to believe there is another FET or 2 that are damaged, would that be safe to assume?
Probably that or the gate drive on the repaired phase.

I had a controller I was experimenting with that something happened and my hand on throttle moved and the motor jerked and caused the board to land on a metal tool, and it damaged the controller in a way that even after replacing blown FETs left it "working", but not like it used to.
https://endless-sphere.com/forums/viewtopic.php?f=4&t=21830&p=320682#p514878
file.php

https://endless-sphere.com/forums/viewtopic.php?p=515777#p515777
I never got around to completely fixing it as it wasn't that great anyway, though I still have it and sometimes think about using an oscilloscope to check out the gate drive of each phase relative to each other to see if that's the cause...but I never have enough time/energy for even the stuff that's critical.... It coud also be that the FETs used were different in characteristics. There's video and scope pics and stuff if you're interested, in the last link.


My buddy says they cannot be tested in circuit so ill have to pull them all.
They can be tested in circuit as a group (whole top or bottom of each phase), but not individually. So for instance you wouldn't know if just one was causing a problem, or not working at all, just if the whole paralleled set was wonky.


I'm also wondering if there's any specific way to tell that the driver is bad or if i should just order those as well since they are pretty cheap?
To test them, you'd probably need an oscilloscope to see the waveforms, since you already know they are driving the FETs, just that something is not happening as it is supposed to. You'd probalby see something differentn between the undamaged phases and the damaged ones.

It probably won't hurt to just replace the gate drivers of the damaged phase before doing the FETs in all the phases, just to see if it makes any difference. But if you want to see wha'ts going on and learn even more, and have access to even the most basic oscilloscope (there are cheap "front ends" available online that connect to your phone using OTG/USB, for instance, though I've never used any of them; also cheap old standard benchtop used ones on ebay and the like), you could see the waveforms and even post pics of the results. :)


My buddy also looked at the data sheet for the FETs we got and said I shouldn't have any compatibility issues replacing the whole phase, he actually picked the replacements. He said it may just handle the load a little better even. (This is just his opinion, im not saying anyone is right I just like to hear different opinions.)
Most likely the different FETs aren't teh cause of the problem--it's usually only if the Qg (gate charge) or RDSon is significantly different from the others taht you get different behavior enough to cause a problem. Usually it doesn't matter...but there are a lot of different kinds of characteristics of a FET, and sometimes certain ones are needed for certain applications...I am not expert enough to help with those, but you can read about them in various custom motor controller and powerstage design threads by those that are. :)

A big enough difference in gate charge means the FET may turn on or off with enough difference in timing to the others that it creates problems, or that it takes more than the gate driver can provide (especially if it's already damaged).


I also showed him a reply that you (I think, maybe one of the other Forum Daddy's here.) posted to someone who had sent you a blown Sabvoton to take apart and you were telling about how they are just clones now with way too little capacitance and very weak FETs and other components, he said it was bound to fail even if my cable didn't short out, so I'm not opposed to doing more surgery on this big heat sink of a board!
Was probably someone else; I don't think I've had a sabvoton here yet, but there have certainly been enough blown ones posted about. ;)
 
I actually got the motor running awesome! All the way to 115kph

Does motor go faster after the repair? This could still indicate a sensor issue.

Lastly, I have 2 sets of halls that I replaced, one stopped working completely after some of these tests and over current faults, and the other set has 3 halls that test perfectly on the oscilloscope and still works and makes it operate the way it does now.

Are you sure the halls are placed ok and wired ok? A scope readout will show that halls are triggered but did you also check timing and pairing with the correct phase BEMF?

Easiest could be if you have a spare controller/to borrow - if it runs poorly also with a new controller then your motor sensors or wiring is the likely error, not the fets.

Which halls did you use, did you place them in the exact same location as before?
 
larsb said:
I actually got the motor running awesome! All the way to 115kph

Does motor go faster after the repair? This could still indicate a sensor issue.

Lastly, I have 2 sets of halls that I replaced, one stopped working completely after some of these tests and over current faults, and the other set has 3 halls that test perfectly on the oscilloscope and still works and makes it operate the way it does now.

Are you sure the halls are placed ok and wired ok? A scope readout will show that halls are triggered but did you also check timing and pairing with the correct phase BEMF?

Easiest could be if you have a spare controller/to borrow - if it runs poorly also with a new controller then your motor sensors or wiring is the likely error, not the fets.

Which halls did you use, did you place them in the exact same location as before?


It went the same top speed unloaded as it did before. And I got the exact same halls but Honeywell instead of generic. I wired them exactly the way they came out. How can I check timing? Because yes the scope shows them being triggered just fine. Its noisy but my buddy stuck a stepped resistor on it and it turned into a clean up and down square type wave when wheel was turned.
 
hall should turn on when the BEMF of the phases crosses zero according to below picture.
Hall sensor ”A” goes high when C+/A- phase BEMF crosses zero etc.
5461B7BE-D519-476D-A183-35C72B40420D.jpeg

”Exact same halls but honeywell” won’t ever happen :wink:
They all have different switching characteristics. SS411 is often recommended here due to their constant output.
A56937AB-6F16-4A47-851A-7DE6731B8B27.jpeg
 
larsb said:
”Exact same halls but honeywell” won’t ever happen :wink:
They all have different switching characteristics. SS411 is often recommended here due to their constant output.
A56937AB-6F16-4A47-851A-7DE6731B8B27.jpeg

Those are the ones I got! SS411 Unipolar. Because someone else on here that replaced them in my motor said those worked for him because the bipolar did not. And thank you for the graphics, I will check that out on the scope tomorrow at the shop!
 
Also I forgot to mention that it does fail the automatic hall test with any and every phase combo, but let's assume my halls are fine, could a bad gate driver cause this symptom also?
 
Oh my goodness. No. I checked my digikey invoice and the halls I ordered were the freaking s441... so is that my issue then??? :oops:

Edit-- checked the post I read about the correct halls that work and I guess the s441a is the one he said to use so I have no idea lol.
larsb said:
”Exact same halls but honeywell” won’t ever happen :wink:
They all have different switching characteristics. SS411 is often recommended here due to their constant output.
 
The 441 is unipolar, and the 411 is bipolar. The difference is:
https://www.electronics-tutorials.ws/electromagnetism/hall-effect.html

Bipolar sensors require a positive magnetic field (south pole) to operate them and a negative field (north pole) to release them.

Unipolar sensors require only a single magnetic south pole to both operate and release them as they move in and out of the magnetic field.

Either should work in your system, but the bipolar type is more commonly seen in hubmotors.

Some controller manufacturers (kelly) have sometimes recommended using unipolar halls, but I haven't seen a specific reason given.

If you want to see the operational difference, you could install an extra hall in the *other end* of the same slot as the middle hall sensor (which is usually in between stator teeth, while the other two are usually in grooves carved out of the teeth), using the other kind of sensor from whichever one you have in there to compare to. If you can't fit an extra signal wire thru the axle for the test hall (or it is too much of a PITA just for a test like this), you could temporarily repurpose the speed or temperature sensor wires if your motor harness has those.

Then you can directly compare those two signals on screen to each other, as they'll be about the same start and end times for each pulse.


The main potential problem I have noted with hall signal sensing on some controllers is that certain ones use a *lower* voltage pullup on the signal lines than the typical 5v, which is already pretty low. That means there is less contrast with the "on" state (grounded), in an electrically noisy environment like a motor controller. They should be using a *higher* voltage pullup, to give greater contrast...but they probably don't because it makes their input circuits to the MCU more complicated (have to buffer the higher voltage down to MCU-input-compatibl levels), so adds a few tenths of a cent of cost to each controller.
 
TurboWelderMonkey said:
Also I forgot to mention that it does fail the automatic hall test with any and every phase combo, but let's assume my halls are fine, could a bad gate driver cause this symptom also?

No, probably not. If your motor runs ok in noload it should be able to drive it to pass the hall test if halls are ok.

I never used sabvotons so i don’t know how stable their test is. On other motor/controller combinations i’ve used that didn’t pass the autotest there were issues with the halls even though i could run motor halfway decently under load.
 
larsb said:
TurboWelderMonkey said:
Also I forgot to mention that it does fail the automatic hall test with any and every phase combo, but let's assume my halls are fine, could a bad gate driver cause this symptom also?

No, probably not. If your motor runs ok in noload it should be able to drive it to pass the hall test if halls are ok.

I never used sabvotons so i don’t know how stable their test is. On other motor/controller combinations i’ve used that didn’t pass the autotest there were issues with the halls even though i could run motor halfway decently under load.

OK interesting. AmberWolf seems to be pushing towards my gate drivers being bad. Those arrive tomorrow so ill let you guys know what happens. I also replaced the whole green phase with those new high power FETs.
 
It'll be interesting to see the conclusion on this. In hindsight the halls should be ruled out before turning to the gate drivers check. I wonder what would happen if you spun the motor at noload and slowly increased load and current by braking the wheel until controller cuts out while looking at the hall signals.
 
larsb said:
It'll be interesting to see the conclusion on this. In hindsight the halls should be ruled out before turning to the gate drivers check. I wonder what would happen if you spun the motor at noload and slowly increased load and current by braking the wheel until controller cuts out while looking at the hall signals.

Good idea. I will try and check this today before I but the new drivers in. But I also think that it could have maybe just been that a couple of the green fets were blown closed, but still enough to run on like 2.5 phases. There was a nice smoke mark under 2 of the green phase mosfets when I was pulling them out but there wasn't an open short on the whole bank like there was with the blue phase that had 5 that were completely dead (shorted open i think?) And 3 that tested ok on diode test...
 
after finishing the drivers and fets on the damaged phases, I'm still having the same issue... so larsb may be right that something is up with the halls. We checked them on the scope at the controller input and what do you know? Just like AmberWolf said, one of them pulls up close to 5v and the others were about half. We had the idea to add larger pull up to both the low ones. After that it got them to pull up around 4.4v ish and the motor seemed to run a little bit better... but I could still grab it and stop it with my hands while it was spinning not too fast and it would start making the angry magnet noises.

Another note is that larsb asked if it runs faster now... and it does! 136kph instead of 115kph. Although I'm not sure on the orientation of the 3 speed switch on those tests so that could have something to do with that. I'll check again today. But here are some photos of the noisy hall signal we were getting while running and the pull up on the halls we made...
 

Attachments

  • 20210414_190629.jpg
    20210414_190629.jpg
    324.7 KB · Views: 573
  • 20210414_181151.jpg
    20210414_181151.jpg
    296.5 KB · Views: 573
  • 20210414_184916.jpg
    20210414_184916.jpg
    354.4 KB · Views: 573
That signal looks totally off. It’s supposed to be a square wave with ON for 180 degrees electrical and OFF for 180 degrees.

Is the signal constant with peaks both high and low or do i read it wrong?

For measuring and alignment of sensor this guide can be used:
https://www.kollmorgen.com/en-us/developer-network/hall-sensor-alignment-process/
 
larsb said:
That signal looks totally off. It’s supposed to be a square wave with ON for 180 degrees electrical and OFF for 180 degrees.

Is the signal constant with peaks both high and low or do i read it wrong?

Yea my buddy said that signal has absolutely no place in electronics hahaha.

And kinda its just the high peak for both showing the difference in voltage pull up before we added the pull ups
 
It doesn’t make sense, the hall voltages must’ve been equal in the originally designed part. Question is then what would cause the voltage difference on two of the sensors? That cause could be part of your issue. Maybe you should follow the traces and see if there are more damages along the way, or poor connections.
 
larsb said:
It doesn’t make sense, the hall voltages must’ve been equal in the originally designed part. Question is then what would cause the voltage difference on two of the sensors? That cause could be part of your issue. Maybe you should follow the traces and see if there are more damages along the way, or poor connections.

My buddy thought maybe we had fried the pull ups in the controller for the 2 halls that read low voltage which is why we added the new ones. But yes I agree. The only damage I know of is the damage to the insulation of the phase wires and a couple of the hall wires all of which I re-insulated enough so that there aren't any shorts to axel or other wires. I thought that could be causing some of the noise but my buddy doesn't really think so. I will take a closer look at the traces and I'm about to test the Controller on my 48v 750w hub motor in a few.
 
TurboWelderMonkey said:
We are currently finding that some of the diodes at the hall inputs of the controller may be fried... may have found the cause

You should be testing the motor halls without the controller even connected.
 
Back
Top