Kelly Controller with Motenergy motor problem.

[MrCherbike]

We have built an electric Suzuki GS 550. Motor is Motenergy ME1905 with Sin/Cos analog encoder and Kelly 8080IPS controller. Batteries are 2P/30S Headway 40152S for 30Ah and 96V. It has a Domino potentiometer throttle. Has been operating about one year now. Generally works very well, but of course a problem has been there since the start that has me baffled here which I am getting tired of chasing and now am asking for ideas.
Roughly about one in thirty starts from a dead stop get a shutdown "Hall Error". I had to push it home without shutting it off to get that reading since this Kelly drive has no easy form of telemetry and I connect it to a PC. If you shut off the isolated 12V to the controller it resets and you are on your way and it may never happen again on that drive. Only in the first ten to twenty feet, never at speed or accelerating while cruising. I am aware that locked rotor(like from a dead stop) is the most difficult for the drive to perform. The actual voltages of the Sin/Cos were measured from a breakout box I installed for that purpose, and they are dead-on and the thing performs the Identification Angle with no problem. I am discounting loose crimps and such things because it only happens during initial acceleration and also the machine gets a lot more jolts and jarring form road bumps. Contact with Kelly has not been forthcoming with ideas. All the parts are new(withing one year). I have been wondering if there is any other type of input to that "Hall Error" besides something to do with the Sin/Cos system. It also might trip as soon as you take it out of the shed or if it is warm. I have physically moved the high current wires to be as far apart as possible, and rotated motor 180 degrees in it's mount to achieve that. Also am using all the shielding connections provided and installed steel enclosures for the control plugs and cabling. The 12V supply is of the isolated switching type and never drops out(the main contactor stays energized).
Lastly I wish Kelly would provide more detailed information in their manual like other manufacturers of VFD devices do and more properly translated it would give more confidence to us! Attaching a few pics have many more including full schematics. The motor has been rotated 180 degrees in it's mount since that last pic.
Thanks in advance!

 

[amberwolf]

If it only happens during high current, the most likely problem is noise from the phases induced into the position sensor signal or power / ground lines.

Do you have an oscilloscope? If so, you can try to see if there is noise there.

Otherwise, you can try further shielding the signal / power / ground lines from the phase wires. Physically isolating the two sets of cabling from each other (phase vs sensor) as the simplest first step, but it sounds like you've already done that, so if that isn't working, you can add actual RF shielding around the sensor cable set, grounded only at the controller end. If that's already been done and doesn't work, then you can try small bypass capacitors on the power/ground lines at the sensor end (using them on the signal lines will alter the actual signal, so that's a last resort).

 

[MrCherbike]

I do have a scope and have considered looking at it. I can install the center stand which I use for all kinds of testing but I would have to hold the rear brake to get a load otherwise it would take a dyno since the scope is stationary and I also would need to watch it. But possible. The motor has the shield grounded inside it and the Kelly has no shield on the six-inch wires with plugs that come out of it for control wiring. I am not excused from doing what I would do in industry and try the other end for shielding and disconnect the shield inside that motor. It is easy to get at under the side cover. Also with some research I have learned what you said that the device producing the radiated field can also be contained with a shield. I already tried a resistor/capacitor filter on the 5V supply right at the breakout box, and also a separate linear 5V regulator derived from 12V elsewhere on the machine and those didn't work. The control and power leads are definitely as separated as I can get them on this layout, and I think it should be far enough. I kind of wish it was more robust like an incremental encoder would be. Neither Kelly nor the manufacturer of the encoder talks about impedances of those lines but I am suspecting it is not low based on the low current consumption of the Sin/Cos board and the higher the impedance is the more likely it is to be affected by any fields near them right? Of course having the shield grounded at the motor and controller would create a ground loop so that is out. I guess when it comes to noise it can always end up being some Voodoo fixes.
Thanks for taking an interest! If anything comes to you in a vision let me know!

 

[MrCherbike]

Excuse the aliasing here between my phone and the scan rate of the scope. But anyway it is just what I read with a meter with center of 3V and plus/minus 1V so from 4V to 2V separated in time by 90 degrees. No regularly occurring noise to be seen. But it will take this kind of thing to rule out some possibilities and keep looking.

 

[amberwolf]

You'd see any noise bad enough to cause a problem pretty easily. Unless their controller software is really crappy, it would be able to to filter minor noise not visible in a scope trace like that.

So...unless theres' stuff you can see at higher or lower frequencies than what the scope is set to in that image, it's probably not electrical noise causing the problem. (unless it only happens very intermittently, just at the instants it causes the error itself).



I don't know the specific controller's features / options / etc well enough to suggest further tweaks to it's settings that might prevent the error.

 

[MrCherbike]

After some altering of the shield connections and installing a steel box to contain the Kelly controller plugs and breakout box for the Sin/Cos encoder did not fix the problem I then verified that the output of the encoder is in fact fifty ohms by testing.
The next thing that got done was to make the frame of the machine 12V ground(negative). It was not built that way and really should not need to be done(the lights and other 12V devices all have negative leads run to them). So anyway at the same time as doing that I also connected the steel box containing the controller plugs and connected the negative conductor for the controller at that point inside that box to the frame locally. I have now traveled about 110 miles under conditions that would instigate the fault and usually would expect some faults in less than ten miles or so. No faults have happened since. I am still testing but it seems that this has fixed it.

 

[amberwolf]

After some altering of the shield connections and installing a steel box to contain the Kelly controller plugs and breakout box for the Sin/Cos encoder did not fix the problem I then verified that the output of the encoder is in fact fifty ohms by testing.
The next thing that got done was to make the frame of the machine 12V ground(negative). It was not built that way and really should not need to be done(the lights and other 12V devices all have negative leads run to them). So anyway at the same time as doing that I also connected the steel box containing the controller plugs and connected the negative conductor for the controller at that point inside that box to the frame locally. I have now traveled about 110 miles under conditions that would instigate the fault and usually would expect some faults in less than ten miles or so. No faults have happened since. I am still testing but it seems that this has fixed it.

Thanks for posting your solution; it will probably help others with similar problems.

The solution points to the likelihood that something in the electrical system (probably the motor, most likely windings on the stator) is inducting to the frame, and the frame is then radiating that to the signals that are having a problem. Grounding the whole frame eliminates that in a similar way to shielding the signals.

FWIW, I think there have been some people with Erider torque sensors that have had to ground the BB / frame to B- to fix issues with them,

 

[MrCherbike]

It was definitely getting to the point of no more ideas except to start replacing parts. So anyway even though I can't explain why it is cured I don't have any problem taking the win there.
Thanks for following this issue and your suggestions amberwolf.