Hall sensor replacement = chattery behavior

[detraquez]

Hello all,

I have searched extensively and haven't found the answer to my questions so here goes.

Setup :
MCS Cruiser
Grin Front Hub V1
72V 2.5kWh Battery
Phaserunner with MT plugs
Cycle Analyst

Introduction :
Original torque arm mount gave out and lead to severed motor cable.

Initial repair :
Opened the motor, replaced motor cable and upgrade to L1019 connector and created an L1019 adapter cable to MT plugs on phaserunner and used the new frame clamp for the torque arm. While I was at it, I replaced the worn out bearings. When everything was sealed back together, ran the autotune to find out there was a bad hall.

Second repair :
After diagnosing an always on hall sensor probably damaged by a short when the cable severed, replaced with a new one. Once that was done, autotune works fine, hall map ok. The only oddity is that the new hall sensor seems to be more reactive that the two original ones. Tested with a magnet, the signals coming from the older ones are a little "sticky" when I approach and remove the magnet. The new one has immediate reactions when approaching and removing magnet. All three are "normally on".

So I lace it all back up, temp read ok, speed ok, hall start ok but a little jerky. Then things get weird. As soon as I put on the power, the motor is vibrating quite severely and isn't running smoothly at all. I tried autotuning several times to no avail. I tried running the motor in all three modes :

Hall only - somewhat jerky start but acceptable, lots of vibrations up to 30kph.
Hall start, sensorless run - identical behaviour.
Sensorless - No start from standstill as we can expect but a buttery smooth operation above 0.5-1kph.

Hypothesis
It seems like the phaserunner isn't switching to sensorless when runner even in hall start sensorless run mode. It seems to be a hall issue because sensorless operation is smooth and quiet as can be. I can't figure out what to do and I would be thrilled to figure this out software wise because I don't really feel like opening up the motor again and replacing the other two hall sensors and relacing the wheel. Maybe the different quality hall sensors lead to bad timing ? If anybody has ever experienced this or has any idea what to do, please advise !

Happy riding !

Arlen

 

[goatman]

did you try resetting the phaserunner before you did the autotune?

 

[amberwolf]

If the hall sensor is a different model / part number, it may be very different magnetic detection response. Replacing the other two with teh same kind would fix this.

If the hall sensor is installed facing the other way from what it had been, it may also respond quite differently than before. Most sensors all face the same way in most motors (all labels in, or all labels out), but some of them the middle one is flipped on purpose.


As a side note, it is very common for cable damage to short phases to halls, and if this happens it often damages the controller itself as well as the hall in question (and if that hall is *also* used for the Cycle Analyst speedo sensor, it can damage the CA as well, sometimes actually destroying it's MCU so it wont' even boot).

I don't know if such damage would allow the autotune to still run normally, but if the autotune automatically just goes sensorless when it doesn't read valid halls, then it certainly could.

Additionally, if the PR has a fallback mode like some controllers, it can run on just two halls signals, but it may not run well like that. I don't know if it will automatically fallback to complete sensorless from this point, or be stuck trying to use just the two halls. (IIRC the present generation of Grinfineons, a totally different controller technology than the PR, is like this--you have to actually unplug the sensors to get it to run sensorless).


It is also possible you have a problem with the new cable or connectors, which would also cause the same symptoms as a bad hall sensor. You can do a continuity test but you'd need to open the motor to test the motor end of the cable. The best way to do such a test is to run directly from the inside of the motor cable connetions to the controller end of the cable at the controller itself, assuming you have access to the controller end wiring. This will also test the actual L1019 connector.

 

[detraquez]

did you try resetting the phaserunner before you did the autotune?

No, I kept the old setting and simply ran autotune to avoid manually reconfiguring voltage, amps, ramps etc.

 

[detraquez]

If the hall sensor is a different model / part number, it may be very different magnetic detection response. Replacing the other two with teh same kind would fix this.

This is what I wanted to avoid but my gut feeling is that it would solve the issue.

If the hall sensor is installed facing the other way from what it had been, it may also respond quite differently than before. Most sensors all face the same way in most motors (all labels in, or all labels out), but some of them the middle one is flipped on purpose.

All facing the same way in the grin motor. I replaced the one on the right.

As a side note, it is very common for cable damage to short phases to halls, and if this happens it often damages the controller itself as well as the hall in question (and if that hall is *also* used for the Cycle Analyst speedo sensor, it can damage the CA as well, sometimes actually destroying it's MCU so it wont' even boot).

Indeed, but in this case the controller and cycle analyst seem to work fine except in sensored mode. Sensorless is buttery smooth ! I actually tried with another phaserunner (L1019 version) and the problem was exactly the same. That said, I had loaded the old parameter file to keep all original settings and ran the autotune to remap halls. Resetting the phaserunner might indeed be a good idea and manually configuring the rest.

I don't know if such damage would allow the autotune to still run normally, but if the autotune automatically just goes sensorless when it doesn't read valid halls, then it certainly could.

That happened after the first repair when I discovered there was a bad hall. After fixing, mapping the hall worked fine.

Additionally, if the PR has a fallback mode like some controllers, it can run on just two halls signals, but it may not run well like that. I don't know if it will automatically fallback to complete sensorless from this point, or be stuck trying to use just the two halls. (IIRC the present generation of Grinfineons, a totally different controller technology than the PR, is like this--you have to actually unplug the sensors to get it to run sensorless).

With PR, my preferred mode is hall start and sensorless run which would solve the issue because the startup would work even though slightly jerky and then would run smoothly in sensorless.

It is also possible you have a problem with the new cable or connectors, which would also cause the same symptoms as a bad hall sensor. You can do a continuity test but you'd need to open the motor to test the motor end of the cable. The best way to do such a test is to run directly from the inside of the motor cable connetions to the controller end of the cable at the controller itself, assuming you have access to the controller end wiring. This will also test the actual L1019 connector.

Continuity checked, all cable are ok.

Thanks for the feedback !

 

[goatman]

I had a little blob of solder work its way to the magnets, if your opening the motor back up and see metal shavings , that could cause the juttering

 

[amberwolf]

If the hall sensor is a different model / part number, it may be very different magnetic detection response. Replacing the other two with teh same kind would fix this.

This is what I wanted to avoid but my gut feeling is that it would solve the issue.

Then before opening:
What part number did you replace the problem one with?
What part number was the original?
If they are the same then they respond the same.
If they are different, you can check their spec sheets to see what their magnetic response is supposed to be, and what their output type is supposed to be (latching vs nonlatching, opencollector vs not, linear vs switching, etc).
If that is the same, then they are still the same for your purposes.
If any of those are different, then they may be causeing the problem, and you are best off by replacing with the original p/n.

Indeed, but in this case the controller and cycle analyst seem to work fine except in sensored mode. Sensorless is buttery smooth !

Which it would be, if only the hall sensor line on the MCU was damaged. If the CA speedometer runs off the halls but is not connected to the line that was damaged, then it would not be affected, and you could see exactly what you describe.

I actually tried with another phaserunner (L1019 version) and the problem was exactly the same.

This means that it is not likely to be a hardware problem in the PR itself.

With PR, my preferred mode is hall start and sensorless run which would solve the issue because the startup would work even though slightly jerky and then would run smoothly in sensorless.

Is that the result you see?

If it is not, then you have a different problem somewhere. Or you have multiple problems.

Continuity checked, all cable are ok.

Checked from inside the motor at the connection points inside? If not, you don't know yet.