V12 Cutout tracking

[supercurio]

22 minutes ago, Paul A said:

People could be injuring themselves by conducting a heavy, awkward test.

A test that is now in doubt as being reliable, as there are reports of wheels failing despite having earlier passed the stress test.
 

Even if the Zen Lee comment that Ewheels will be conducting a recall does not eventuate, it seems pointless to do the test.

Agree, at this point it is wiser to skip the test and avoid riding a V12 altogether if you own one.
The rare exceptions would be if your V12 already proved itself extensively with jumps, stairs and bumps.

For most, the stress test will give a false sense of security.

What's worse: the units which passed the stress test would fail in more dangerous situations than what we've seen so far: higher speeds, bigger jumps, steeper stairs, bigger bumps.
All that increasing the risk of severe injury.

IMHO, wait for the replacement boards. They should have a number of differences including stronger MOSFET and likely the same as V12 HT boards.

Edited January 12, 2022 by supercurio

[FidoUK]

24 minutes ago, pico said:

I would rather say Surge rather than Transient.

My first though when I rode the wheel was, Geez the V12 is very aggressive in shutting down the motor when the wheel falls off.

(Hey, this is what I do, my wheel when doing tricks falls more than the casual rider) The V10F is a real fiasco in that respect. Takes forever to stop the wheel spin. This was noted in the V10F thread.

I do understand feedback loops. I also know that you can also control the gain of the loop.

My first thought was that WOW tons of gain (VERY GOOD for tricks) lets tone it down.

My theory from what I have observed is that it is very easy for the software to be OVER aggressive when SLOWING the wheel and that the ferro magnetic is saturated by pulses width way too large to be absorbed by this 16 inch wheel material. I suspect also variations in the over current detection and variation in quality of the windings would explain different wheels behaviours...

I could be wrong... 

In the mean time I am doing single legged tic-tocs but in COMMUTER mode. I am staying away from the OFF ROAD....

We have similar theory with our V12, stay away from OFF ROAD mode.

The wheel arrived last Thursday, passed the "test" successfully (off-road) so all good, we rode it like maniacs for two days in "commuting" mode, speed bumps, up and down curbs, the V12 is an amazing machine! 2 days passed, we switched to Off-Road mode and board failed after a 4cm 2mph standing jump testing Grizzla pads, 2 x MOS failed. 

We have replaced the MOS yesterday and again passed the "test" successfully, OFF-Road mode. Road test feels good tested with different riders style and weigh but commuting mode only for now.

Could a software update limiting braking power and overall performance specially OFF ROAD setting be the easiest way for Inmotion to get some time until proper investigation is complete?, what do you think@Cecily Inmotion

 

 

[supercurio]

6 minutes ago, FidoUK said:

We have similar theory with our V12, stay away from OFF ROAD mode.

The wheel arrived last Thursday, passed the "test" successfully (off-road) so all good, we rode it like maniacs for two days in "commuting" mode, speed bumps, up and down curbs, the V12 is an amazing machine! 2 days passed, we switched to Off-Road mode and board failed after a 4cm 2mph standing jump testing Grizzla pads, 2 x MOS failed. 

We have replaced the MOS yesterday and again passed the "test" successfully, OFF-Road mode. Road test feels good tested with different riders style and weigh but commuting mode only for now.

Could a software update limiting braking power and overall performance specially OFF ROAD setting be the easiest way for Inmotion to get some time until proper investigation is complete?, what do you think@Cecily Inmotion

It's likely that the firmware handles differently some of the transient power demands between commuter & off-road, as well as depending on the extra pedal sensitivity settings.

But keep in mind that the original cutouts, before the test was published occurred both in off-road and commuter.

Therefore commuter mode will provide no guarantee that any of these V12 are safe to ride.

[RockyTop]

I believe the word is tolerance. It has very little. When adjusting electrical, mechanical devices I use the term “widow of acceptability” … the window is too small. 

[FidoUK]

Totally agree, tolerance needs to be higher, to restore confidence now. This data was available since IM released the wheel and testers/reviews units didn't have any issues even pushing the wheels to the edge, somehow IM managed to handle the same tolerance before? 

 

[Rawnei]

1 hour ago, supercurio said:

IMHO, wait for the replacement boards. They should have a number of differences including stronger MOSFET and likely the same as V12 HT boards.

It's very probable that only those with broken wheels/boards will get replacement boards, in that case just waiting is not an option if your wheel is in working order.  

[supercurio]

1 minute ago, FidoUK said:

Totally agree, tolerance needs to be higher, to restore confidence now. This data was available since IM released the wheel and testers/reviews units didn't have any issues even pushing the wheels to the edge, somehow IM managed to handle the same tolerance before? 

We typically don't know if early testers or reviewers had cutouts or burned MOSFETs, especially for pre-versions.
That's usually something they would report to the manufacturer they have a relationship with, but not publicly.
That'll depend how much this person is willing to risk the relationship in exchange of transparency. And for pre-prod models, that might not even be relevant. 

One reviewer reported MOSFET failure where the wheel should have protected itself, like when @Jack ex-KS tested the V12 on a steep climb and it failed. In this case, Inmotion answered they would look into it. However we know today that instead even more fragile wheels were mass produced later on

[supercurio]

6 minutes ago, Rawnei said:

It's very probable that only those with broken wheels/boards will get replacement boards, in that case just waiting is not an option if your wheel is in working order.  

And that would be a disaster, since after a successful stress test, the only way to receive a robust replacement board would be by crashing first.

In my opinion the only two acceptable option would be

1. Recall of all wheels (best, costly)
2. Voluntary recall, with board replacement of all wheels for whomever requested it (people less informed will still crash, cheaper)
3. Replace only the burned boards either during stress testing or by user cutout (not acceptable)

 

[skedra]

7 minutes ago, supercurio said:

And that would be a disaster, since after a successful stress test, the only way to receive a robust replacement board would be by crashing first.

In my opinion the only two acceptable option would be

1. Recall of all wheels (best, costly)
2. Voluntary recall, with board replacement of all wheels for whomever requested it (people less informed will still crash, cheaper)
3. Replace only the burned boards either during stress testing or by user cutout (not acceptable)

 

I think voluntary recall may be the way to go with distributors notifying the customers, and IMO InMotion should cover the cost of that recall in this case, especially considering how many wheels may be out there with the issue. Doubt it will happen, but one can wish.

[Paul A]

https://www.paypal.com/au/webapps/mpp/paypal-buyer-protection

 

Paypal.

Shop with confidence.

Ensure you get what you pay for with Buyer Protection.

 

If your order is significantly different to the seller’s description or doesn’t arrive, you have 180 days to file a dispute. It could be that the event you bought a ticket for was cancelled, you didn’t receive everything you bought, or the item was defective, damaged or counterfeit. We can protect you for the full purchase price plus shipping costs, up to $20,000 per item.

To help protect your purchase, make sure you:

• Pay with PayPal
• Keep your account in good standing (so don’t dispute everything!)

[Denny Paul]

22 minutes ago, Paul A said:

https://www.paypal.com/au/webapps/mpp/paypal-buyer-protection

Paypal.

Shop with confidence.

Ensure you get what you pay for with Buyer Protection.

If your order is significantly different to the seller’s description or doesn’t arrive, you have 180 days to file a dispute. It could be that the event you bought a ticket for was cancelled, you didn’t receive everything you bought, or the item was defective, damaged or counterfeit. We can protect you for the full purchase price plus shipping costs, up to $20,000 per item.

To help protect your purchase, make sure you:

• Pay with PayPal
• Keep your account in good standing (so don’t dispute everything!)

Maybe if you bought directly from inmotion to put the squeeze where it belongs. Otherwise you're just hurting your dealer. In a niche hobby like this you'd be burning one of only a few bridges available. 

Best to just communicate with your dealer and work it out. That's what Paypal would do anyway with buyer protection; they just use the funds as collateral to ensure everyone plays nice.

[Paul A]

Was thinking more along the lines of using Paypal if purchasing from Aliexpress.

https://sale.aliexpress.com/__pc/XJew3VffFA.htm

[pico]

On 1/12/2022 at 11:14 AM, RagingGrandpa said:

Fortunately, EUC's do have current-limiting logic in their controls, including direct measurement of motor current.

The more I think about it the more I doubt of the reaction time of their current limiting algorithm. Judging by the test they propose. This is a reaction time test( can it survive the disturbance without failing) and not a load test (with full carpet burns!).

The way I see it: sometimes it does not see well enough a surge in current (when braking) and poof! This seems to happen irrespective of battery voltage.

BTW they possibly (pure speculation of my part!)  corrected their algorithm also in the V10F and suddenly on a PROVEN design they are popping Mosfets like crazy!  On a 84V wheel!

Interesting! 

Edited January 13, 2022 by pico

[supercurio]

Status update:

I had a call with the staff of @skedra's dealer this afternoon.
They own one of the wheels which passed the stress test as instructed by Inmotion and later failed.
Their stress test execution was good enough to make several other V12 units fail. They have a video of it so I encouraged them to publish it for everyone to review. In the meantime, I checked the footage and it was well done.

Now, how did their "test passed" V12 died? With a 5cm jump, at 2km/h.

Alternative stress test suggestion #1

(I made that up, not official whatsoever)

So I may suggest to try a new stress test with your V12. Here's how:

1. Set the V12 to off-road mode
2. Install jump pads on your V12
3. Make sure the tire pressure is 30PSI or more
4. Hold something on your side to stand on your wheel immobile, or do so at very low speed.
5. Lift the wheel like you would while riding to jump onto a curb or over an obstacle. Target: 5cm high.
   Non perfect technique is desirable, it's great if the wheel starts spinning in the air, and abruptly stops when back on the ground
6. Repeat

I can't promise it'll do anything else than give you a good workout and practice EUC jumps.

Alternative stress test suggestion #2

Same, I made that up. It's easier to run than #1

1. Set the V12 to off-road mode
2. Find a place outside where tire marks don't matter
3. Disable V12 anti-spin feature
4. Lift the wheel unevenly a few cm
5. Let it start to spin a bit
6. Lower it back to the ground: the goal is to have the wheel stop rotating suddently

Make sure you're not destroying your tire with crazy burn-outs tho.

I hope #1 or #2 can reproduce the conditions accurately enough which made the wheel die after passing the test. #1 is tougher on the wheel, but needs advanced riding technique and experience.
I can't try yet because my wheel is still at the distributor.

Let us know how it goes, take care 🙏

Edited January 13, 2022 by supercurio

[Paul A]

@supercurio

Just like to say thank you for your stamina and continuous efforts in improving safety for all of us.   

 

[pico]

@supercurio I second @Paul A 

Thank you!

[yoos]

1 hour ago, pico said:

reaction time of their current limiting algorithm

Are you implying that there is some basic feedback loop that dictates current judging by tilt angle and then there is a sophisticated but slow high-level algorithm that watches over the basic one and tries to intervene if the current gets too high? 

I thought all EUCs function in the same way: there is one single algorithm which takes into account everything: tilt, speed, acceleration, temperatures, currents, voltages etc at any given moment and perhaps also the recent history of those values so it can better adjust dynamically. So, there would be a single "mind" inside the EUC that decides on optimal current every moment. With the speed of modern CPUs there seems to be enough time to run a non-trivial algorithm even between each pulse of the PWM! Or am I getting it completely wrong?

edit:typo

Edited January 13, 2022 by yoos

[pico]

33 minutes ago, yoos said:

basic one and tries to intervene if the current gets too high

In essence you are not wrong.

This is low level (or should be). Can be implemented in hardware or software or both. 

The idea is whatever the main loop is doing (to keep you balanced) one should limit the current drawn by the 3 phase bridge driving the motor. (any multiple of 6 Mosfets) 

Failure to do so will result in unwanted high current that can kill the Mosfets.

If you are not diligent enough also kill in order the fuse and/or the batteries (ask Gotway). Depending on design choices. 

Quite a few safeguards in a wheel.

This is the order I would like things to fail.

Warning and or tilt back/Shut down by overcurrent protection/Fuse/Mosfets/Traces - battery cables - phase cables

 

 

Edited January 13, 2022 by pico

[RockyTop]

Sorry In advance. -  This does not look like a few bad wheels or a few bad parts. It looks like a weak design and any V12 given the right circumstances will fail. It looks like these test are pushing close to the edge. Sometimes they go over and sometimes they don’t. These test would be stressful to any wheel. I fear that these test just aggravate the components and push them closer to a fail. 

[RagingGrandpa]

1 hour ago, yoos said:

there is some basic feedback loop that dictates current judging by tilt angle and then there is a sophisticated but slow high-level algorithm that watches over the basic one and tries to intervene if the current gets too high

That summary is correct.

A simplification of the self-balancing controller is:

In real systems (like EUCs) we add more logic before the "Motors" element, to limit the "Controller Output" (FET drive command) according to protection criteria such as 'too much current'...

[yoos]

7 minutes ago, RagingGrandpa said:

self-balancing controller

If I understand it correctly, this is a generic PID controller, [not really specific to EUC and balancing], which uses analog integration and derivation to optimize phase currents. (the "angle" in the diagram is apparently the angle of the current vector in 2D space and has nothing to do with EUC tilt). Then there should be a EUC-specific balance controller somewhere that checks for overcurrent etc and, more importantly dictates the overall current/power/torque, i.e. implements the "algorithm part" which makes different EUCs behave differently and features soft/hard modes etc. I thought that this EUC-specific balance controller is digital, not analog (i.e. instead of using hardware integration, differentiation etc it would use a microprocessor and memory and run any algorithm installed via the app). Would that be correct?

[RagingGrandpa]

But actually, the diagram above represents the digital self-balancing control logic quite well. 

• "Feedback" is the movement of the EUC, caused by the motor torque, and becomes measured by the IMU "Sensor"
• "Current Angle" is not electrical current (an awkward choice of words: think 'currently present' and not 'amps'). It just means the actual angle of the EUC pedals, and is the thing that the "Sensor" (IMU) estimates (with some additional signal processing not shown in the picture).
• "Error" is the physical angle in degrees, that the pedals need to move to become level again.
• Different gains (P,I,D) and other logic at this level are used to create the hard, medium, and soft modes you mentioned.
• "Controller Output" represents a commanded force, and for an electric motor this is inherently an electrical current command. Note that this is not a speed command. Motor force can speed up and slow down the motor, and the controller really doesn't care about it- because it is controlling pedal angle, not speed.

BLDC motor commutation (the "angle of the [force] vector in 2D space" you mentioned) is a 'solved problem' and not related to self-balancing control, and not shown in the diagram. That low-level control of the motor phases can be treated as part of the "Motors" element... just in reality its logic happens to be located in the EUC control board, and not in the wheel motor.

In this discussion, we can take the commutation behavior for granted, because self-balancing and current-limiting can be achieved just by modifying the magnitude of the "Controller Output."

.02

 

Edited January 13, 2022 by RagingGrandpa

[RockyTop]

@RagingGrandpa good explanation. In addition the controller logic/ software makes a big difference on demand. How far off angle will it allow in a given time, how aggressively will it correct and how smoothly will it compensate? Dose it try to correct everything in the next pulse or divide the force in the next ten pulses. This is something that KingsSong seems to be really good at. 

Edited January 13, 2022 by RockyTop

[yoos]

@RagingGrandpa Thanks for the explanation! If I understood correctly, the diagram is still a simplification: a genuine PID control might be too simple, even with added current caps. What I am curious about is whether the control is exercised by a CPU (using discrete data collected at a high rate from all sensors and meters), using algorithms written in high-level language or whether they use an analog controller (i.e. complicated circuit which functions continuously) while the CPU is just used to change its settings (sorry, I am a theoretical physicist and my knowledge of these questions is evidently lacking, but I do have the capacity to learn)

[RagingGrandpa]

16 hours ago, yoos said:

What I am curious about is whether the control is exercised by a CPU 
or whether they use an analog controller

Sure 

The control is performed on a CPU using discrete-timestep data collected at a high rate from all sensors using algorithms written in high-level language.

Gotway uses the STM32F CPU, for example.

The CPU has timers and sub-functions that make it easier to separate the high-speed topics (like the 20kHz PWM output that triggers FETs to activate and deactivate) from the medium-speed topics (like 100Hz pedal-angle feedback control) and low-speed topics (like 1Hz board temperature monitoring and reporting data to the smartphone app).

(I think there are no 'analog' controllers involved in self-balancing. There are analog topics involved in low-level circuits to achieve things like voltage regulation... but they're not participating in the motor control algorithms.)

 

Edited January 14, 2022 by RagingGrandpa