Failing hall sensor symptoms?

[E-HP]

I had a weird issue where was getting erroneous speed readings on my Cycle Analyst. The CA harness I’d made used the motor’s yellow hall sensor wire as the speed source. Continuity testing passed but the issue persisted until I switched to using the blue hall wire. Speed reads correctly now, but I didn’t know why the yellow didn’t work, although it had previously for several thousands of miles. What I’m noticing now is that the motor runs as smooth as silk and silently, but the motor is less smooth above 30 or so. My question is, do hall sensors fail all at once, or are there cases where they work well until they get past a certain speed and can’t switch on and off quickly enough causing the motor to run more roughly at higher speeds? Has anyone experienced this type of issue?

Edit: the motor has only exceeded 100C a couple of times and over 90C 5 or so times, and for very short periods (the time to find a spot to pull over). It has around 8k+ miles on it, and copper windings were bright and looked like new when I replenished the Statorade.

 

[fechter]

I've never seen one fail partially. Either good or bad, but anything is possible. Much more common would be some kind of intermittent connection, either a connector or bad solder joint.
You might want to try measuring the hall signal voltages to make sure they all look the same just turning the motor by hand.
Another possible issue might be noise pickup from the phase wires.

 

[Zambam]

Edit: the motor has only exceeded 100C a couple of times and over 90C 5 or so times, and for very short periods (the time to find a spot to pull over). It has around 8k+ miles on it, and copper windings were bright and looked like new when I replenished the Statorade.

Where on the motor was the temperature measured?

 

[E-HP]

I've never seen one fail partially. Either good or bad, but anything is possible. Much more common would be some kind of intermittent connection, either a connector or bad solder joint.
You might want to try measuring the hall signal voltages to make sure they all look the same just turning the motor by hand.
Another possible issue might be noise pickup from the phase wires.

Thanks. I’ll take some measurements.

Where on the motor was the temperature measured?

temp sensor in the windings.

 

[E-HP]

I've never seen one fail partially. Either good or bad, but anything is possible. Much more common would be some kind of intermittent connection, either a connector or bad solder joint.
You might want to try measuring the hall signal voltages to make sure they all look the same just turning the motor by hand.
Another possible issue might be noise pickup from the phase wires.

Thanks. I’ll check the halls and probably replace both sets of hall connectors if the hall checks pass.

 

[afzal]

I had a weird issue where was getting erroneous speed readings on my Cycle Analyst. The CA harness I’d made used the motor’s yellow hall sensor wire as the speed source. Continuity testing passed but the issue persisted until I switched to using the blue hall wire. Speed reads correctly now, but I didn’t know why the yellow didn’t work, although it had previously for several thousands of miles. What I’m noticing now is that the motor runs as smooth as silk and silently, but the motor is less smooth above 30 or so. My question is, do hall sensors fail all at once, or are there cases where they work well until they get past a certain speed and can’t switch on and off quickly enough causing the motor to run more roughly at higher speeds?

In general, accuracy of hall sensor based position estimation by controller reduces as the motor speed increases, especially with high erpm motors - controllers like VESC switch to sensorless at higher speeds.

You mentioned Y hall sensor was working fine earlier, so here, perhaps with the modifed wiring, noise pickup as mentioned by fetcher.

 

[E-HP]

I believe noise was the issue. The halls were fine, so after checking the connectors, I moved the conductors around to route as far away from the other conductors as much as possible in the limited space, and that seemed to smooth out the motor at higher speeds. It also appears to have resolved a separate issue mentioned on my build thread regarding odd current readings that were bouncing around and going negative (I had tried switching the direction of the shunt, but that made current and watts negative all the time, but now it’s working). I have the hall wiring going into the battery box, but they don’t need to, and the connections could be done all outside of the box. I may route them outside and maybe add some shielding of some sort, as the permanent fix.
Thanks for the help and suggestions!

 

[fechter]

There are some pretty big magnetic fields around the phase wires at higher power levels.
If signal wires are run parallel to these, it basically acts like a transformer.
Physical separation helps. Avoiding parallel runs helps. Better filtering on the signal inputs helps.

 

[docw009]

I have thought about how the Hall sensors in a motor can work in such a noisy environment, They do use latching Hall sensors for the phases which must reduce signal bounce, but many sources say the internal speed sensors are non latching, The speed sensor does flash at a slower rate. like 1/16 slower if you have 16 magnets, I have never seen bad speed readings on my KT controllers using the internal speed sensor in the hubmotor, but when I used an external wheel sensor, most of the time I would see speedometer jitter, which I attribute to noise pickup. Sometime i would reroute the cable and that fixed the jitter. It would only show up when coasting when there is no motor noise.

An interesting bit of info is in the external speed sensor used on Tong Sheng middrives. Magnet spacing on this sensor is part of the TSDZ2 lore. It has to be adjusted right. If set too close, the speed might read 2x or 4x. In addition, the stock firmware and OSF firmware respond different. I can run the OSF at closer magnet spacings that cause the factory firmware to read 3X, I wonder if they filter it with software, Anyway, magnet spacing may make noise a factor too,