Smoke and sparks coming from controller. Trying to learn what happened so it doesn't happen again.

[Cr1sp]

TLDR: What happened (most likely) is that hall sensor ground shorted to phase wires with the controller turned off but still plugged into the battery when brake disc rotor bolts chewed through insulation as the wires exist the axle. The short caused sparks and plume of smoke so there was a decent amount of energy. Since the controller was off, how was this possible? Is there always voltage in the phase wires even when the controller is turned off?

Full story:
Today I went out on a ride on my QS 273 and Sabvoton 72200-M bike. About a mile in I got an error on my display "Controller protect" which happens once every few months when I go over a hard bump, turning off and on the controller makes it go away and I can continue, not today. Everytime I turned it back on and tried to set off with the throttle, the bike would nudge forwards before cutting off power and throwing "Controller protect" error. Only having allen keys with me, I took off controller cover and did a quick inspection of wiring and connections but everything looked fine so I decided to put the cover back on and just push it back home since I was a mile away.

As I started pushing it, I noticed that there was resistance in the motor which told me that there's something wrong with my phase wires but the resistance wasn't constant, it was on and off and the frequency would change depending how fast or slow I was pushing the bike, at a quick look, I noticed that the 6 bolts holding on the disc rotor to the motor, were rubbing against the phase/hall wires as they come out of the axle and as the wheel was turning, the bolts would touch the phase wires, short them out causing resistance then there would be a gap which would get rid of the short and repeat, this would also explain why the controller kept going into protection mode after nudging forwards because it would nudge forward, a disc rotor bolt would pass the phase wires shorting them causing the controller to go into protection mode.

Since I knew the issue, I ripped of the zip ties holding the phase wires to the frame so I could move them or reposition them so they would not be touching the bolts so I could at the very least get home, as I moved the phase wires, suddenly sparks started shooting out from underneath my controller cover followed by a constant plume of smoke for solid 2 seconds, I quickly started moving the phase wires forward and backwards and the sparks and smoke stopped. I quickly disconnected the battery after that. On the bright side after that smoke show, there was no longer resistance in the motor allowing me to push it home

When I got home, first thing I did was grab a fire extinguisher and then try reconnecting the battery which was uneventful but the controller was not turning on nor the display. I also noticed that the whole controller was covered in soot, at closer inspection I noticed that there was a large amount of soot at the wires as they come out the controller with a melted black wire which is the hall sensor ground, here's a picture, all the black stuff that you can see is not dirt and it's straight soot, the controller was very clean.

I haven't taken off the motor yet to inspect where the motor wires were rubbing but what most likely happened was that the bolts also rubbed off insulation from the hall sensor ground wire, when I was moving the wires to get them away from the disc rotor, the hall sensor ground must've shorted to the phase wires. I'm unsure what was causing the sparks but the smoke was most likely the ground wire just melting because it's thin and had a lot of resistance.

I tested my display and it still works so since it doesn't turn on with it plugged into the controller, this means that the controller doesn't even pass the battery voltage to the display so it's probably decently cooked. Since there was no longer resistance in the motor after the smoke show, most likely the FETs also blew which on the bright side it allowed me to push the bike home without much effort which I don't care since the controller was very likely cooked anyway even if the FETs weren't blown since it doesn't pass voltage to the display. My guess is that the hall sensor are probably dead too but fortunately QS 273 has two sets so I can use the other one.

When I was moving the phase wires right before sparks and smoke, the controller was turned off but still plugged into the battery so I'm unsure how shorting ground to the phase wires caused this? I'm unsure how controllers work but if the controller is turned off, shouldn't there be no power in the phase wires? I'm trying to figure this out as I always want to learn from my mistakes to not repeat them. I already ordered a new controller which should be with me in about a week which is more than enough time for me to take off the motor and repair the insulation.

 

[amberwolf]

When a controller is "off" but battery is still connected to it, the phase FETs still have full battery voltage on them, but the LVPS that makes 12v and 5v / etc doesn't get battery power, so the brain (MCU) and sensors, throttle, and other logic are unpowered.

Depending on exactly what happened in there, it's possible the controller is unrepairably toast.

The large amount of smoke in phase wire shorts most commonly is from the phase FETs, and sometimes capacitors (rare), and likely the wiring and/or PCB itself since the resistance went away (blown FETs usually short Drain to Source, which is why a blown controller tends to "lock up" the motor). Replacing the FETs is possible, but damage to other parts like their gate drivers is common, and if the wrong kind of failure happens, can even reach all the way to the MCU outputs that control the gates...and that can destroy the MCU, which is not replaceable (unless you can get the firmware or code to flash a new one with and have the tools to deal with all this).

Wiring damage is usually easy to fix, but PCB damage is often not. If it's a multilayer board things can happen inside the layers that destroy things you can't even see were there (and aren't now).

If the phases just touched each other, it's likely that they'd all have been at the same state (all FETs off) so there probably would not have been anything to short, with no voltage difference between them.

If the phases just touched the hall ground, that would potentially only short battery voltage thru the upper half of the phase bridge FETs to ground. What fails inside at that point depends on the ground path from there. You could get lucky and only the ground wire itself burned, or possibly the ground trace on the PCB if it wasn't part of a ground plane.

If the phases touched the hall signals, that passes battery voltage directly into parts of the controller only intended to handle a few volts. Typically the hall signals go right to the MCU inputs, *and* they are indirectly connected (via the pullup resistors) to the 5v supply (which goes to the MCU, the LVPS itself (well, comes from there), other logic circuits, your throttle, the halls, any PAS or other sensors, etc. Battery voltage fed in there can damage or destroy any or all of those parts.

 

[Cr1sp]

When a controller is "off" but battery is still connected to it, the phase FETs still have full battery voltage on them, but the LVPS that makes 12v and 5v / etc doesn't get battery power, so the brain (MCU) and sensors, throttle, and other logic are unpowered.

Depending on exactly what happened in there, it's possible the controller is unrepairably toast.

The large amount of smoke in phase wire shorts most commonly is from the phase FETs, and sometimes capacitors (rare), and likely the wiring and/or PCB itself since the resistance went away (blown FETs usually short Drain to Source, which is why a blown controller tends to "lock up" the motor). Replacing the FETs is possible, but damage to other parts like their gate drivers is common, and if the wrong kind of failure happens, can even reach all the way to the MCU outputs that control the gates...and that can destroy the MCU, which is not replaceable (unless you can get the firmware or code to flash a new one with and have the tools to deal with all this).

Wiring damage is usually easy to fix, but PCB damage is often not. If it's a multilayer board things can happen inside the layers that destroy things you can't even see were there (and aren't now).

If the phases just touched each other, it's likely that they'd all have been at the same state (all FETs off) so there probably would not have been anything to short, with no voltage difference between them.

If the phases just touched the hall ground, that would potentially only short battery voltage thru the upper half of the phase bridge FETs to ground. What fails inside at that point depends on the ground path from there. You could get lucky and only the ground wire itself burned, or possibly the ground trace on the PCB if it wasn't part of a ground plane.

If the phases touched the hall signals, that passes battery voltage directly into parts of the controller only intended to handle a few volts. Typically the hall signals go right to the MCU inputs, *and* they are indirectly connected (via the pullup resistors) to the 5v supply (which goes to the MCU, the LVPS itself (well, comes from there), other logic circuits, your throttle, the halls, any PAS or other sensors, etc. Battery voltage fed in there can damage or destroy any or all of those parts.

Thanks for the reply, I always thought that the ignition cable/blue wire in the display connector powers on a contactor/relay inside the controller that is normally open so even with the controller off but battery plugged in, there's only power up to the controller terminals and that's it. I guess this is one way to learn this.

I did take the motor off today and it does appear that the disc brake rotor bolts were rubbing on the phase wires, it was probably very barely or they were only rubbing when I was going over bumps with the wiring bouncing and touching the bolts because I have about 4000 miles with this setup and there's very small nick in the insulation so if it was permanently rubbing, I feel like it would've shorted months ago, here's some pictures:




Only the yellow phase wire has damage, the others appear to be fine, also the white fiber loom where all the hall sensors run inside is untouched so my theory of the hall ground also being damage and shorting to the motor doesn't seem like it was the cause, even if it's damaged inside, that fiber protective loom is not conductive. I also did hall sensor test with ebike tester and both hall sensors sets work fine even the one set that I was using at the time so they appear to be undamaged which is surprising as I thought that set was cooked.

I don't have any more theories on what caused the sparks and smoke and why the hall ground melted where it exists the controller (the ground on the motor side appears to be fine where it's exposed and all the soot is at the controller side), the sparks and smoke started when I was fiddling with the phase wires where they exist the axle and they stopped when I started moving them so I'm confident that was the cause. My only other theory is that I have few unplugged connectors inside my controller box which is metal, when I was moving the phase wires, maybe I bumped the bike and unconnected ground touched the box making the whole bike grounded and at the same time the damaged phase wire was touching the motor so there was a short and the reason why the hall ground melted is because that's the longest and thinnest ground on the bike so it had the most resistance and started melting first? This is probably a long shot.

I have a cheap generic 72v controller that I don't mind killing so I'll repair the damaged insulation and try that controller and see if the bike works, I don't want to wait for the Sabvoton to arrive just for me to plug it in and instantly kill it because the issue is still there. I will also electrical tape all unused connectors just incase they caused this.

EDIT: It appears my spare controller is dead, I ordered another generic one on amazon prime that will arrive tomorrow. I did however test the motor very quickly by plugging it into my other working bike and the motor spins fine in the air which is a good sign as the motor is the most expensive part. I got rid of my disc rotor bolts and swapped them out for stainless steel countersunk allen bolts and now the phase wires have plenty of clearance and should never rub again.

 

[amberwolf]

Thanks for the reply, I always thought that the ignition cable/blue wire in the display connector powers on a contactor/relay inside the controller that is normally open so even with the controller off but battery plugged in, there's only power up to the controller terminals and that's it.

That would be a better way to work it, but costs a lot more, and would be larger (contactors / relays for the currents in these controllers aren't really all that small) so I don't know of any typical controllers that are built this way.

Some controllers manufacturers like Kelly expect builders to use a contactor (and a precharge circuit) and even include this in their diagrams. Some more advanced controllers like Sevcon have contactor and precharge control lines built into the controller, with programmable control over how they operate and when. But they don't include the contactor inside.


When you reinstall the motor, see if there is a way to not put the wire exit directly against the frame. If you can fit "C" shaped washers (which you'll probably have to make out of regular fender washers) between frame and axle shoulder, to space the frame farther away from the motor face, it will give you enough space to not worry about rubbing, if you then tie down the cable to the frame right at the axle area to hold it away from the rotor/etc. (there probably isn't room to do that ATM).

 

[fechter]

With most ebike controllers, there is always power to the DC bus but there should be NO power on the phase wires unless one of the FETs is shorted. The FETs should all be "off". There may be a tiny amount of leakage which might show up with a voltmeter, but not enough to do anything.