Erider torque bb cycle analyst on/off behavior

Joined
Sep 8, 2019
Messages
555
Location
USA, CA, Bay Area
Erider torque bb cycle analyst on/off behavior (1:12 video demo)


The CA is reporting RPMs while constantly spinning the pedal to be on/off/on/off in a pattern. The motor is getting power in the same on/off pattern. This also happens when in toque sensing mode as well.

Some details:

  • Frankenrunner controller
  • DD_45 motor
  • 3.14 CA software (not latest stable, I discovered, but pretty close)*
  • 72v nominal system.
  • PAS is powered via a DCDC step down (12v) since the CA says you can't run these from the CA when on a >52v battery.

* I see that 3.15 added "ERider_T9" which has defaults for my sensor, but I've just input those manually as a custom torque sensor, so not really a big change.

Maybe I'll give the latest beta a try; the main thread on the CA seems to have some rather positive feedback on the PAS situation
 
Last edited:
Strange--I've never had this behavior with a cadence sensor on the CA (regardless of the torque sensor); what are all of your settings in the CA? I suspect one or more of them is contributing.

Alternately you can look at my settings files for the SB Cruiser trike, I've posted them as zip files in a few threads recently for others to compare theirs with to help figure out problems.
 
The CA settings (in hex output). I can't fathom what would be configured in there to cause it to happen. I pretty strongly suspet the BB might have a hardware glitch. I'll probably swap over to a regular magnet pas sensor this weekend to confirm.
 

Attachments

  • xclass-v3-2b3.zip
    607 bytes · Views: 3
The only significant differences I see besides system specific stuff are below; my values are in parentheses after yours.

Power Limit Settings
Amps Feedback Gain 150 (5)
Power Feedback Gain 15.0 (5.0)

I don't know if it will make any difference, but it is something to check.
 
I have system based on different microcontroller - MPe but with Erider torque sensing BB. Does this situation happen when you ride as well? Or is this happening only when you test it like on the movie ? In my system I can't get motor assist on when moving pedals by hand - it is because I am not able to apply enough force to activate it. However on my screen I have readings on "mass on pedal in grams" that is constantly changing. So maybe you should calibrate torque sensor in Cycleanalyst ? In idle mass on pedal shpuld be O.
 
Kiwago said:
I have system based on different microcontroller - MPe but with Erider torque sensing BB. Does this situation happen when you ride as well? Or is this happening only when you test it like on the movie ? In my system I can't get motor assist on when moving pedals by hand - it is because I am not able to apply enough force to activate it. However on my screen I have readings on "mass on pedal in grams" that is constantly changing. So maybe you should calibrate torque sensor in Cycleanalyst ? In idle mass on pedal shpuld be O.

It happens while in actual use as well. It's extremely noticeable as well, the motor cutting in and out like that jerks you around quite a lot.
 
Resolved!


tl;dr: these sometimes need you to ground the battery negative to the frame.
 
Last edited:
So, I am back with a new build and running into this torque sensor showing a new inconsistency issue.

I can scope the RPM and Torque lines and see exactly what I expect (as long as the frame is grounded). RPM gives out the 36pulses per rotation and torque goes from 1.5-3.5v.

The CA, however, seems to lose the RPM signal above, say, 30rpm and starts reporting it as 0rpm. Oddly, despite having the CA in torque mode, the loss of the RPM signal causes the CA to stop sending power to the motor. (The CA debug screen also shows some odd error letters with the A and W going capitals despite this thing being in the air and getting nowhere near any cut off values...)

As noted, I've got the frame connected to the battery negative, and verified it's a solid connection, but that isn't resolving the issue.

Adding to the oddities: I changed the CA config to be "Basic Assist" (instead of torque) and it does not exhibit the dropping of the RPM signal in that setup.

I did take a moment, though, since I had the scope out, to show what it looks like when the frame is un/grounded with the erider:


With the cadence assist working, that's honestly good enough for now. Torque would be better, but since I'm in the preorder line for the all axle 45mm with its built in torque sensor, I'll just leave good-enough alone for now. Just thought I'd come document some of what I've hit for posterity.
 
I can scope the RPM and Torque lines and see exactly what I expect (as long as the frame is grounded). RPM gives out the 36pulses per rotation and torque goes from 1.5-3.5v.

The CA, however, seems to lose the RPM signal above, say, 30rpm and starts reporting it as 0rpm. Oddly, despite having the CA in torque mode, the loss of the RPM signal causes the CA to stop sending power to the motor.


Which CA FW version are you using? There is an option in at least one of the later (3.2x?) FW to use pure torque input, but previous versions require the RPM signal even in torque mode.

(I can't remember the specific menu choice differences; been quite a while since I was looking at that, since my use case would work better with pure torque for startup, then pure cadence once I'm moving, and I don't think that was an option).


Ok, I poked around the big CA beta thread, and found this
and skimming that page and the next there is info that seems to imply (without reading it all in detail) that if a cadence signal is being received it will try to use it. If true, you can disconnect the cadence wire it would then just use the torque signal only.

Downside being that it would never respond to cadence, so if you're in a gear ratio where torque is insufficient to cause motor output where you need it, and you can't change gears to something that would provide torque output sufficient for that, you'd be stuck with lower motor output than you need (whereas if cadence was still working you could pedal faster to get more output, etc).

I might not have interpreted it correctly, in which case it won't help. :(

(The CA debug screen also shows some odd error letters with the A and W going capitals despite this thing being in the air and getting nowhere near any cut off values...)
That sounds like a bug, and the other part might be too.

Is it the same FW version on this CA as on the previous working setup?


As noted, I've got the frame connected to the battery negative, and verified it's a solid connection, but that isn't resolving the issue.

Adding to the oddities: I changed the CA config to be "Basic Assist" (instead of torque) and it does not exhibit the dropping of the RPM signal in that setup.

I did take a moment, though, since I had the scope out, to show what it looks like when the frame is un/grounded with the erider:

That's a nice clean signal; where's the scope ground and scope signal monitored at? (shouldn't matter, since the CA works in the BA mode it can obviously read it correctly; I'm just curious).
 
Which CA FW version are you using? There is an option in at least one of the later (3.2x?) FW to use pure torque input, but previous versions require the RPM signal even in torque mode.
3.2b3
Ok, I poked around the big CA beta thread, and found this
Digging into the release notes, it seems the old "rowbike" (wheel torque now) mode does something like where it will use torque only. There's also a bunch of info about how the RPM/Torque interact in the release notes. There's probably something to uncover there if I wanted to really dig into it.
where's the scope ground and scope signal monitored at?
I actually took the ground from the frame and the signal near the cranks. I suppose it could be getting dirtied up somewhere along the line with interference. Maybe when I'm redoing the handlebar bundle I'll scope from up there as I rearrange things and see if I'm doing myself in with pulling the signal along a 12v line and other such things. I...don't really expect that to be the case, though.
 
Back
Top