Sherman S cutout on jump landing: an early pattern emerging (Patton as well)

[supercurio]

Yesterday, Alexis Caison posted a slow motion of a fine jump landing which resulted in his Sherman S cutting out on a Facebook post, blaming it on himself for a harsh landing.

The video shows a fairly clean landing however. The only imperfection in technique is the motor starting to spin backwards mid air due to a slight amount of backwards lean.
Nothing out of the ordinary tho. I landed so much worse comparable jumps on my OG Sherman many times, even digging a hole on the landing spot (my bad)

 

Thing is, it is not the first cutout of its kind. The same occurred during prototype testing in August/September 2022 in China.
A friend dug out the video and here's an edited version, which highlights the similarity.

 

The main potential root causes are:

• Firmware issue
• Hall sensor poorly secured in the motor, as @MrMonoWheel identified in his unit

 

What's next?

More investigation is needed. For now, the Sherman S is not a good candidate for jumping practice.

Alexis Caison injured his knees (graphic) due to lack of protection so I would urge every Sherman S rider to never omit gear in case they would get some planned, or unplanned air time.

Help needed

Please report in this thread any similar occurrence, of a cutout on landing. Except maybe if it was a terrible one with the wheel having no chance to recover.

Edited March 27, 2023 by supercurio

[Mango]

It looks like an overtorque situation with an extended suspension that the wheel couldn't keep upright in time as the suspension compressed during landing. You see the wheel was trying desperately to keep itself balanced after the landing.

Not a cut out.

[supercurio]

29 minutes ago, Mango said:

It looks like an overtorque situation with an extended suspension that the wheel couldn't keep upright in time as the suspension compressed during landing. You see the wheel was trying desperately to keep itself balanced after the landing.

You didn't specify which video you were commenting about.

In all cases, let's reason about the the simple physics involved :

1. The motor must rotate according to the horizontal axis, forward speed of the vehicle to continue balancing on landing
2. The suspension has only an effect on the vertical axis

Conclusion: The suspension or lack thereof has no impact on the torque required during landing.
The only exception would be if the EUC is tilted forward or backwards significantly, but it is not the case in both videos attached.

In both videos, we can observe a cutout, with the motor/controller making no effort to balance, seemingly disengaging entirely before the rider's weight is transferred to the vehicle and suspension compresses. This is why we can confidently describe them as cutout and not overlean.

Edited March 24, 2023 by supercurio

[Rawnei]

If anything surely suspension must take stress of the motor as there is less force to deal with on impact which is softened by the suspension.

It would be interesting to see this video without slow motion.

[supercurio]

48 minutes ago, Rawnei said:

It would be interesting to see this video without slow motion.

Here you go!

 

[Rawnei]

9 minutes ago, supercurio said:

Here you go!

 

No beeping while in the air so doesn't seem motor was spinning that fast.

[supercurio]

4 minutes ago, Rawnei said:

No beeping while in the air so doesn't seem motor was spinning that fast.

Yep there was only an EUC World disconnection beep (unrelated)

When the wheel spins out in the air it can quickly beep before landing indeed. Then really no good reason for the wheel to cutout in this case.

Edited March 24, 2023 by supercurio

[chanman]

Was there really backwards spin in the air in the first video? I thought maybe that was a camera trick, I'm not quite sure we could take it at face value and there doesn't seem to be backwards tilt to the wheel. I guess in slow-mo it's pretty clear and we shouldn't have to worry about the aliasing tricks since it's super high frame rate relative to the wheel rotation.

The landing is interesting, in slow mo you see it actually start balancing shortly after the landing at the impossible forward angle instead of falling all the way over. 

Maybe a torture test ala s22 motor is in order to see if hall sensor issue can be reproduced, otherwise I'd think some firmware current limitations on the landing might be the culprit.

Edited March 24, 2023 by chanman

[supercurio]

I just heard that the LeaperKim Patton also has the same issue of cutout on landing, in case the motor is spinning backwards.
That's despite its high torque characteristics, and would confirm the theory developed above.

Apparently LeaperKim is investigating now - to be continued.

[Tasku]

We would need to see logs. Maybe it is some kind of high amp spike? Could there be some protective measures against that and during hard landing amps could spike high? Just a thought.

[Rawnei]

Since no other EUC has this I'm guessing it's not intended and it's a bug rather than a feature, good find and good that they are looking into it.

[supercurio]

29 minutes ago, Tasku said:

We would need to see logs.

I look forward to controllers storing and sending high frequency logs. Unfortunately at the moment, with a sample rate between 2-3 Hz what we get is useless for that kind of short live events unless you can reproduce it hundred to thousands of time. Until there's a crash, or you get extremely lucky.

An unpleasant methodology.

29 minutes ago, Tasku said:

Maybe it is some kind of high amp spike? Could there be some protective measures against that and during hard landingamps could spike high? Just a thought.

Yeah it could be a protective system, sensor data filtering error, even a small logic error in the code.
Knowing the main conditions able to trigger the cutout: landing with the motor spinning backwards, looking at the source is likely to give a good idea.

I would suggest to pay extra attention to Sherman Max with the new silent controller firmware too which removes the magnets pulling sound at low RPM, in case the root cause is a bug in the new motor control algorithm.

 

Then it could also be a problem with the hall sensor or something else related to hardware, and that would end up much more problematic, an Abrams-like scenario.

[supercurio]

Although it's too early to tell if it's significant here, it's still interesting to think about the effect of the hall sensor moving around on bumps and landings.
Would the hall sensor still provide a qualitative enough signal when its distance to what it measures oscillates or varies?

When the motor turns backward on landing, it also means that it reaches zero and low RPM on impact. Low RPM is when hall sensor control is used instead of sensorless, making the system prone to being vulnerable to poor hall sensor signal - even if it's not an issue at any higher speeds.

Edited March 27, 2023 by supercurio

[Rawnei]

Maybe it would also be relevant to know if these wheels have that high speed more enabled or not?

[M1thr4ndr]

When the accelerometer is tilted backwards in the air (wheel spins backwards), its trying to go the opposite way free spinning right? When ground contact happens all the power and torque slams the body forward (opposite direction tiltback), because its thinking its going too fast backwards. Isn’t this what was happening?

[supercurio]

19 minutes ago, Rawnei said:

Maybe it would also be relevant to know if these wheels have that high speed more enabled or not?

High speed disabled on the recent cutout, and most likely high speed non-existent on the early test prototype.

[supercurio]

12 minutes ago, M1thr4ndr said:

When the accelerometer is tilted backwards in the air (wheel spins backwards), its trying to go the opposite way free spinning right? When ground contact happens all the power and torque slams the body forward (opposite direction tiltback), because its thinking its going too fast backwards. Isn’t this what was happening?

Yes correct.

When in the air the wheel doesn't know it's in the air, only its tilt angle so it makes the motor rotate accordingly. Simplifying: the tilt acts acceleration/braking as bidirectional throttle.

When ground contact happens, we can think of a few forces at play (assuming flat ground)

1. inertia from motor rotation, from the added weight of the magnets, rim and tire
2. forward inertia from the unsuspended and unsuspended weight of the wheel: 
3. forward inertia from the weight of the rider
4. feet of the rider applying more force to the front of the pedals initially

In this list:

• 2 and 3 help the wheel carry forward without any extra torque requirement on landing compared to riding on the ground the whole time
• 1 and 4 are what will make the wheel dip forward. The faster the motor rotates backwards the more that inertia will make the wheel brutally slow down at the contact patch point. The more the rider presses on the front of the pedals the more it'll require torque to remain level as well

There's no indication than OG Sherman or Sherman Max suffered from this previously (mine included), despite OG Sherman having a lot less torque than Sherman S or Patton. So it's a new issue, only reminiscent of the Abrams which sometimes failed on react on time to sudden torque demands.

[inMotional]

Let me add another similar incident to the case. 
Trying to push my new Sherman S up the stairs resulted in a cut off. I was holding it by the lift handles and tried to give it some momentum by going back and forth while being mid-stairs. At some point the wheel switched off. IMO while it is not a cut-off caused by a jump it has the same physics involved. Moving back and forth in low speed and then hitting against an obstacle. I'll try to reproduce it if possible

[Tasku]

15 minutes ago, supercurio said:

So it's a new issue, only reminiscent of the Abrams which sometimes failed on react on time to sudden torque demands.

Yeah that. Where the wheel is in steady conditions, it would make sense to regulate or protect behaviour so it would not cause damage to some other parts. This could be some new protection, maybe just trying to protect battery from extreme amp drains or just some other protection type that works well in general use, but not maybe with jumps. Maybe the air time and spinning wheel to wrong direction is a part of it. Maybe it is scenario that needs multiple triggers. More of a guessing game here in the audience.

While jumping, try to control the device with your feet, so that the wheel is working in correct way in terms of spinning. (yet not to freespin cutoff while mid air)
 

[M1thr4ndr]

2 hours ago, supercurio said:

Yes correct.

When in the air the wheel doesn't know it's in the air, only its tilt angle so it makes the motor rotate accordingly. Simplifying: the tilt acts acceleration/braking as bidirectional throttle.

When ground contact happens, we can think of a few forces at play (assuming flat ground)

1. inertia from motor rotation, from the added weight of the magnets, rim and tire
2. forward inertia from the unsuspended and unsuspended weight of the wheel: 
3. forward inertia from the weight of the rider
4. feet of the rider applying more force to the front of the pedals initially

In this list:

• 2 and 3 help the wheel carry forward without any extra torque requirement on landing compared to riding on the ground the whole time
• 1 and 4 are what will make the wheel dip forward. The faster the motor rotates backwards the more that inertia will make the wheel brutally slow down at the contact patch point. The more the rider presses on the front of the pedals the more it'll require torque to remain level as well

There's no indication than OG Sherman or Sherman Max suffered from this previously (mine included), despite OG Sherman having a lot less torque than Sherman S or Patton. So it's a new issue, only reminiscent of the Abrams which sometimes failed on react on time to sudden torque demands.

If the wheel is less powerful, the in air backward spin would not slam the body forward upon ground contact right? 

[supercurio]

1 minute ago, M1thr4ndr said:

If the wheel is less powerful, the in air backward spin would not slam the body forward upon ground contact right? 

If the wheel is less powerful, in a worst case scenario it would have less time to accelerate the motor in the backwards direction. But it will also have more difficulty matching rotation speed and vehicle speed on landing, so that's probably a net zero in the end.

I shared the topic in the Facebook community by the way and found that comment from Justin Ptnd spot on:

Quote

this is exactly the way to use an euc an downhill jump mountain bike trail, really disappointed to have a cut here, it's perfectly normal use

 

[chanman]

Has anyone seen the control boards? Some people attributed the Abrams problems to not having enough capacitance with its 2x560uF cap config and added more themselves.

I know the S22 has 5x680uF, which is more than 3 times that,  and my V11 has 2x470uF (also only 100V rated which is not a good margin, but I digress), but not sure if there's a standard or necessary amount for extreme demands. Maybe similar issue here in dealing with the current spike which would be pretty large. I've blown a couple V11 boards under similar circumstances.

From some digging it looks like the 2x560 config is pretty common on a lot of models, which don't have similar problems, so probably more firmware related in trying to be conservative in not blowing the board, but the overlean as a result? Assuming the issue is indeed current output and not hall sensor related.

Edited March 27, 2023 by chanman

[supercurio]

15 minutes ago, chanman said:

Has anyone seen the control boards? Some people attributed the Abrams problems to not having enough capacitance with its 2x560uF cap config and added more themselves.

There has been a lot of confusion mostly due to Kevin at Erides who associated capacitors and Abrams cutout mistakenly.

For the Abrams, the cutouts were due to poor / inconsistent signal from the hall sensor, leading to the controller not being aware of some instantaneous torque change requirements like those occurring when overpowering the motor intentionally, or road bumps. Essentially, the controller would extrapolate from previous hall sensor data, which works if it can be and not at all otherwise.

LeaperKim tried to work around the hall sensor signal quality issue by running sensorless most of the time, but still relied on hall sensor below a certain RPM where riders would still cutout.

The problem was eventually fixed quietly by dealers replacing many Abrams motors, despite there has been no public recall. With the motor control algorithm LeaperKim had at the time, running fully sensorless was not an option.

The technology exists now however (see VESC sensorless full torque at 0 rpm), and LeaperKim opted for a new motor control algorithm since, reported to be open source code from STMicro. It is characterized by the disappearance of magnet pulling sound at low RPM.

15 minutes ago, chanman said:

I know the S22 has 5x680uF, and my V11 has 2x470uF (also only 100V rated which is not a good margin, but I digress), but not sure if there's a standard or necessary amount for extreme demands. Maybe similar issue here in dealing with the current spike which would be pretty large.

Like everyone else I hope that it's a software bug and not a strict limitation of the mainboard.

The fact that Patton is also reported to be affected could indicate that it's more a software issue, because that one has a much more torque oriented motor / tire size combo than the Sherman S - requiring lower current spikes to handle imperfect landings.

[chanman]

22 minutes ago, supercurio said:

not a strict limitation of the mainboard

Appreciate the clarification on the Abrams, I would be inclined to agree, at least on capacitance alone. I'm now in a rabbit hole, so for posterity, it looks like the capacitances in the 1-1.5mF range have been standard on the big wheels. The RS which people have been doing extreme jumps with is 2x680's for example. I do wonder the motivation behind the departure, the S22 is 2.5x this and the V13 is almost 3 times with 18x220uF. I assume engineers aren't driving up the BoM for no reason or just to make the boards look better.

Edited March 27, 2023 by chanman

[inMotional]

20 hours ago, inMotional said:

Let me add another similar incident to the case. 
Trying to push my new Sherman S up the stairs resulted in a cut off. I was holding it by the lift handles and tried to give it some momentum by going back and forth while being mid-stairs. At some point the wheel switched off. IMO while it is not a cut-off caused by a jump it has the same physics involved. Moving back and forth in low speed and then hitting against an obstacle. I'll try to reproduce it if possible

It happened again today with no intention whatsoever. Rolling up the same stairs with no significant force applied. It switched off with no beeps at all.
Now turning the high speed setting ON and will follow the wheel's behavior