Breaking crash analysis

[Paco Gorina]

Last day I submitted the Ninebot One E+ to some breaking tests for a project. The idea was to test it against a bicycle to and compare the negative accelerations so we have data for a negotiation with the authorities proving that EUC’s may break as fast as a bicycle. By the way, they are.

I connected 9BMetrics and started to make hard breaks.

 Everything was going fine when in a breaking the Ninebot suddenly cuted current and I was ass in the pavement. Fortunately I already have that outcome into account and was protected enough so not a big problem.

Afterwards the Ninebot didn’t start till I disconnected and re connected the battery. Perhaps someone may explain this.

So now it begins the forensic analysis of the 9Bmetrics Recordings.

 

I have loaded all the recording in 9BMetrics .9bz format to Dropbox. It is just a .zip that has .csv for each magnitude. It is available at :

https://www.dropbox.com/s/bslxvdivd07se6s/20161130_202301.9bz?dl=0

Well, first considetation is that all was fast, fast, fast.

 

Here we have a graphic of the last moments. Angle scale is in the right axis. Other magnitudes correspond to left axis, Speed in km/h Current in A and Pitch and Roll in º.

From starting to break to end of the event was less than a second.

When sufficiently enlarged the development of the crash shows things were not as simple as a cut.

My interpretation of the sequence of events is as follows:

Time	Event
221.5	Start of Breaking. Pitch begins to increase because I push hard on the rear of the paddles. Speed continues to increase and current begins to go down to negative values as the controller tries to maintain the paddles horizontal.
221.8	Pitch increases less (but continues increasing). Controller begins to reduce and stop current decrease.
221.9	Controller has reversed current flow and compounds with my torque and pitch jumps up.
221.8	I reduce push out of equilibrium and fall. Without my weight pitch recovers a little but the controller continues to jump current up so eventually continues to drive pitch up.
222.13	Finally, the controller begins to reduce current. There are some changes that are reflected in speed
222.21	Wheel begins to oscillate in roll direction
222.43	Current is cut for some reason but current was already going to 0. Anyway, stability is already compromised and fall of the wheel is inevitable.

That’s just an event but my first interpretation is that the sudden increase in pitch (or the sudden change from pushing in front to rear) has been too much for the controller that has been driven out of its stability zone and has begun to oscillate as feedback system do when they have a too much gain.

Power may have been cut by either the mainboard or the BMS (sorry, didn’t have a voltmeter then to measure voltages) due to the sudden increase to more than 50A. That is about 3000W !!!

What do you think? Do you have other registers that may help to understand those types of failure?

Once again I was testing heavy breaking so it is not easy to find ourselves in this situation if we don’t look for it. I did previous breakings with negative accelerations of 0.7g while usual ones when driving are from 0.1g to 0.3g and had no problems.

 

 

[Paco Gorina]

I had a little time and have made a comparison with a normal braking (one that ended well) from the same tests. I have got one very similar.

As you can see in the graphic (dark color = Crash, lighter colors = OK) they are very similar. In fact speed and current are a litthle over in the crash but not by much.

Big difference is the excursion of current into negative values is able to stop pitch increase in the correct one but not in the crash. Also between 0,7 and 0,8 s the controller decides to change current and as a consequence pitch goes up and I go down. From here all is done.

Other posible differences are :

Sampling (s/100)	OK	Crash
Speed	1,83	4,52
Current	1,81	4,41
Pitch	1,93	5,17
Voltage(v)	59,14	58,68

So we have a lower voltage (may have affected) but curiosly a big increase in sampling period.

I don't know if it is an artifact of the sampling and transmitting to 9BMetrics but linked to the crazy current reversal of the controller, one may think that the controller really choose a wrong way, was working with excesive delay and got messed.

Another interesting think is that maximum current when braking seems to go about 15A. I have not seen a lower value.

 

 

[Frode]

1 hour ago, Paco Gorina said:

 

 

Yes, this looks quite disturbing. That current reversal during breaking (@ 0.75 s) is indeed strange. To me it looks like the controller suddenly reverses from braking to acceleration when you are still trying to brake. There is no clue here to why, except maybe that it seems to happen when the pitch reaches ca. 6 degrees. The only thing that comes to my mind is a possible bug in the controller software...

 

What is even stranger is that the speed actually is decreasing while the current is sky-rocketing. Where is this current drained if not through the motor? And if the 30-50 amps goes through the motor, why is speed still decreasing? Did you hit a major hump in the road?

[Smoother]

You've done some good work here. Unfortunately I'm not able to help you analyse your charts.  I did, however try to imagine, at what point on the graph you went flying through the air, purely as an academic exercise, you understand. ?

[Paco Gorina]

No hump absolutely flat

[Paco Gorina]

I think I went flying a bit after 0.8. As you see pitch recovers a little as my torque goes to 0

[Chriull]

On 1.12.2016 at 7:02 PM, Paco Gorina said:

Afterwards the Ninebot didn’t start till I disconnected and re connected the battery. Perhaps someone may explain this.

This means normally that the overdischarge protection of the BMS kicked in.

BMS cut-off could also appear imho from overtemp, undervoltage - but in this cases the BMS "reconnects" once the situation is within limits again.

You managed speeds noticable above 20km/h without tiltback with the NB One E+?

The current reverse could be that the controller switched from regenerative breaking to "plugging" braking (connectiong the battery in "reverse" to the coils) which needs "enormous" power but creates more braking torque?

If so, maybe an explanation could sound like this:?

The current change starts once the big pitch increase happens (from ~7 to ~12°) - with this some stabilizing of the pitch occurs together with an speed decrease.

But as you are already out of balance you push the wheel forward with increasing pitch, the current rises but the breaking force is not enough to counter.

Once the current reached ~55A the BMS cutoff occured and the capacitors (are they big enough? - imho not) supported the wheel for another ~0.3s? Or just the board had supply for another 0.3s and the motor was already without voltage...

 

[Paco Gorina]

7 hours ago, Chriull said:

This means normally that the overdischarge protection of the BMS kicked in

Thanks for the info, I was not sure about that and didn't try to measure the voltage at the mainboard. 

7 hours ago, Chriull said:

You managed speeds noticable above 20km/h without tiltback with the NB One E+?

Yes that is interesting and may be part of the explanation of what happened. Perhaps the controller wanted some tiltback so it didn't go so fast after the increase in pitch and when it tried was too late.

7 hours ago, Chriull said:

The current reverse could be that the controller switched from regenerative breaking to "plugging" braking (connectiong the battery in "reverse" to the coils) which needs "enormous" power but creates more braking torque?

I have drawn some lines to mark interesting events. I don't believe the surge was in the direction to break. Reason is that at event 2, when current begins to reverse, pitch increases much much faster. If it created more torque by inverting the current then it should reduce pitch increase and not increase it.

I think event 3 is beginning of my instability, it should reduce my push down it may push it forward. As I push less pitch reduces the increase and when I fall (between 3 and 4) it decreases but the system is not happy so it continues to increase current.

Behavior from this point may be strange as inertia moment of the wheel without the driver is very different.

Probably the BMS does not cut with a current spike but wants some time of high current. That could explain the delayed current cut.

The big question about the current inversion may be explained by an instability of the closed loop system that begins to oscillate or because we have gone to a position in variable space that has not been considered or simply a sign in some condition, for example, suppose that the variable to define current in the controller is stored in a ones complement value. At the middle (if it is 8 bit between 127 and 128) there is the change in sign. The system is decreasing the current by subtracting one. If in some place it goes from 0x80 to 0x7F the value goes from -128 to +127, glups.

Another possibility is an interaction between the current loop and the controller loop. Somewhere in the forum someone explained that in some of this devices, they don't program the motor controller. It is standard and work in current control. I suppose the subsystem has a current or torque command and a loop controls the current drawn by the motor. That means there are some loops controlled by  other loops and seem difficult to model and analyze out of normal or fairly lineal conditions.

I think there is a lot of information in the reduction of the sampling frequency. That could mean the controller is locked in some undesired loop.

I have added a detail of 0,4s around the interesting events with markers for the points and also for the current points in the normal event.

[Mono]

 

On 12/5/2016 at 1:11 PM, Frode said:

Yes, this looks quite disturbing. That current reversal during breaking (@ 0.75 s) is indeed strange.

It's not surprising to me, but expected. As pointed out by Chriull, when regenerative braking becomes insufficient to meet the pitch input, the controller switches to active/power braking. I believe I can feel this switch happening also with my Gotway. 

Quote

To me it looks like the controller suddenly reverses from braking to acceleration when you are still trying to brake. There is no clue here to why, except maybe that it seems to happen when the pitch reaches ca. 6 degrees.

All consistent with a switch to power braking. 

Quote

The only thing that comes to my mind is a possible bug in the controller software...

What is even stranger is that the speed actually is decreasing while the current is sky-rocketing. Where is this current drained if not through the motor? And if the 30-50 amps goes through the motor, why is speed still decreasing?

Because the motor is "used actively" for non-regenerative power braking.

I have seen graphs in previous posts which also record voltage during braking. IIRC, there we see an voltage increasing during regenerative braking which becomes a voltage drop during the power braking period.

[Smoother]

I don't have the technical knowledge to post "factual " info about what's going on inside an EUC, but my understanding and reasoning lead me to the conclusion that  @Mono is correct. The response from the EUC to inputs, is proportional.  It responds with a wide range of power outputs that, hopefully, are linear and appropriate to the input.  EUC can regeratively brake, they can also power brake.  In almost every braking situation, the control board must switch between the two seamlessly so that the response continues to match the input.  So, as @Mono states, at some point, when regenerative can no longer return the equation ( input+ output) to zero, it must step up its game and resort to power braking.  I can't say where on the graph that is, because I've also heard that some wheels report positive voltage, no matter the direction ( from the batteries or from the wheel when regeneratively braking).  

I have to keep reminding myself that the EUC is blind.  It can't see hills, pot holes, rocks, nothing.  The algorithms (if, that's what they are) accept pedal angle, battery voltage ( and maybe amps) speed of wheel, wheel angle (left to right) battery temp(BMS) control board temp, and probably other stuff I've missed .  From that it has to figure out what to do next.  I'm sure that when extreme speed meets extreme pedal angle, meets extreme current, all bets are off, and it's anyone's guess what will happen next, or do we already know ?  FP.

[Paco Gorina]

1 hour ago, Smoother said:

I can't say where on the graph that is, because I've also heard that some wheels report positive voltage, no matter the direction ( from the batteries or from the wheel when regeneratively braking).  

Ninebot reports current with sign. Seems positive is acceleration, negative breaking and regenerating but who knows. May be the supossed current, not the measured one.

 

[Mono]

9 hours ago, Smoother said:

when regenerative can no longer return the equation ( input+ output) to zero, it must step up its game and resort to power braking.  I can't say where on the graph that is, because I've also heard that some wheels report positive voltage, no matter the direction ( from the batteries or from the wheel when regeneratively braking).

I guess you mean current rather than voltage. The switch point is quite likely where the current starts to shoot up from large negative values to large positive values in a very short time frame. Here is my version on the below events, somewhat different from the above comments: at point 2 the controller starts to switch the braking policy and then between point 3 and 4 power braking takes place. It becomes physically effective at the tick shortly after point 3, where we see the steepest deceleration in this trial. At point 4 some exception handling sets in, which is likely the point of no return to a normal riding state. At point 5 the rider starts to fly/butt-plant backwards thereby pushing the wheel further forward away from him.

Well, what do I know, but all in all it doesn't look like the perfect control algorithm  

22 hours ago, Paco Gorina said:

I have added a detail of 0,4s around the interesting events with markers for the points and also for the current points in the normal event.

 

[Smoother]

I don't have the technical knowledge to post "factual " info about what's going on inside an EUC, but my understanding and reasoning lead me to the conclusion that  @Mono is correct. The response from the EUC to inputs, is proportional.  It responds with a wide range of power outputs that, hopefully, are linear and appropriate to the input.  EUC can regeratively brake, they can also power brake.  In almost every braking situation, the control board must switch between the two seamlessly so that the response continues to match the input.  So, as @Mono states, at some point, when regenerative can no longer return the equation ( input+ output) to zero, it must step up its game and resort to power braking.  I can't say where on the graph that is, because I've also heard that some wheels report positive voltage, no matter the direction ( from the batteries or from the wheel when regeneratively braking).  

I have to keep reminding myself that the EUC is blind.  It can't see hills, pot holes, rocks, nothing.  The algorithms (if, that's what they are) accept pedal angle, battery voltage ( and maybe amps) speed of wheel, wheel angle (left to right) battery temp(BMS) control board temp, and probably other stuff I've missed .  From that it has to figure out what to do next.  I'm sure that when extreme speed meets extreme pedal angle, meets extreme current, all bets are off, and it's anyone's guess what will happen next, or do we already know ?  FP.

[Smoother]

@mono . Your analysis seems reasonable.  Something I noticed: the speed reduction in the crash graph doesn't seem very much. He was going 28 kmh indicated and ends up around 18kmh before the "fit hit the Shan" and then the wheel takes off by itself.  Contrast that with his non butt plant run, which had a high of about 26kmh and a low of 9kph in the same time frame.  That's an crashing reduction in speed of 10 kph, and a stable reduction In speed of 17kph in the same time frame.  It doesn't make sense to me.  It's not like he was braking harder, in fact the non crash braking was harder than that of the crash, by definition and in fact.  It has something to do with that sharp current up turn, which the other, more aggressive braking didn't have.  Something is  rotten in State of Denmark.(Hamlet)

[Frode]

20 hours ago, Mono said:

It's not surprising to me, but expected.

But it seems that 50 amps reversed braking current give less breaking than 15 amps into the battery. Look at the pitch angle. It skyrockets at the same instance as the current reverses. Things are not consistent here...unless something is wrong.

[Mono]

4 hours ago, Frode said:

But it seems that 50 amps reversed braking current give less breaking than 15 amps into the battery.

We are talking about a time frame of a little over 0.1 second. This might well be the time where the rider started to get off balance without yet having lost the contact to the wheel. Like that he would give a very strong lean input, which could easily counteract any brake effort in particular at that speed, where, I believe, you can (easily) outlean any Ninebot. The changing pitch angle fits to this course of events.

5 hours ago, Frode said:

Look at the pitch angle. It skyrockets at the same instance as the current reverses

Right. My handwaving explantation is that there is a (small) delay between end of regeneration and begin of power brake at which the pitch angle gets a little too loose. I believe I feel the same effect on a Gotway, which under hard braking becomes a fraction of a second loose before a more severe braking force steps in.

4 hours ago, Smoother said:

@mono . Your analysis seems reasonable.  Something I noticed: the speed reduction in the crash graph doesn't seem very much. He was going 28 kmh indicated and ends up around 18kmh before the "fit hit the Shan" and then the wheel takes off by itself.  Contrast that with his non butt plant run, which had a high of about 26kmh and a low of 9kph in the same time frame.  That's an crashing reduction in speed of 10 kph, and a stable reduction In speed of 17kph in the same time frame.  It doesn't make sense to me.  It's not like he was braking harder, in fact the non crash braking was harder than that of the crash, by definition and in fact.  It has something to do with that sharp current up turn, which the other, more aggressive braking didn't have.  Something is  rotten in State of Denmark.(Hamlet)

I have no definite answer, but small differences in whatever can lead to the difference between crash or no crash here, for example battery temperature which is not even recorded. The crash run is a little faster and with a stronger acceleration before braking. The straight current decrease between points 1 and 2 looks already a little suspicious (together with the different pitch behavior) compared to the more noisy decrease in the non-crash graph, but I have really no idea what to make of it. 

I have not much of a doubt that this controller is far from being optimal, but I also reckon that programming such controller is much less trivial than one could ever dream of without having it done at least once.

[Frode]

On 7.12.2016 at 3:20 PM, Mono said:

at point 2 the controller starts to switch the braking policy and then between point 3 and 4 power braking takes place

 

Why should it need to switch "braking policy"? If one inspects the blue "speed OK" curve, one can see for time between ca. 0.65 and 0.7 that 15 amps generate much more torque (braking force) than 50 amps "plugging braking" did in the "speed not OK" case (if that plugging braking hypotesis is correct). I just don't get this. Is not the torqe (braking force) generated by the motor proportional to the current sent through it? And if it is, where are the 50 amps sent/cooked?

 

[Mono]

6 hours ago, Frode said:

Why should it need to switch "braking policy"?

Because, I suspect, that regenerative braking is less effective than power braking. I have seen several sources of confirmation that power braking actually takes place, at least on other wheels than the Ninebot.

Quote

If one inspects the blue "speed OK" curve, one can see for time between ca. 0.65 and 0.7 that 15 amps generate much more torque (braking force) than 50 amps "plugging braking" did in the "speed not OK" case (if that plugging braking hypotesis is correct).

No, that is not what one can see. One can only see that -15 amps lead to more deceleration than +50 amps (at somewhat different speeds though), which does not necessarily imply that it generated more torque. You assume that all other variables were equal, and I highly suspect that they were not. I suspect that in the +50 amps case the weight with which the rider was leaning into the wheel (i.e. pushing it forward) was much larger than in the -15 amps case. I also suspect that without such kind of counter force (like with a free spinning wheel) that you could never have currents that high. 

Quote

I just don't get this. Is not the torqe (braking force) generated by the motor proportional to the current sent through it?

I would believe so, give or take. 

Quote

And if it is, where are the 50 amps sent/cooked?

Interesting question in any case when we see high currents. I guess the answer is similar to where are they cooked when accelerating strongly from low speed or when pushing backwards against a wall without accelerating at all? Something is getting warm. My observation is that you can't do these things for a very long time. I suspect there are heat sensors which then limit the power. 

[Chris Westland]

Sorry to hear about the crash, but the data is fascinating.  Your interpretations sound very reasonable to me, and this provides a lot of insight into what is going on in the control circuitry inside an EUC.

[Chris Westland]

On 12/5/2016 at 4:25 AM, Paco Gorina said:

I had a little time and have made a comparison with a normal braking (one that ended well) from the same tests. I have got one very similar.

As you can see in the graphic (dark color = Crash, lighter colors = OK) they are very similar. In fact speed and current are a litthle over in the crash but not by much.

Big difference is the excursion of current into negative values is able to stop pitch increase in the correct one but not in the crash. Also between 0,7 and 0,8 s the controller decides to change current and as a consequence pitch goes up and I go down. From here all is done.

Other posible differences are :

Sampling (s/100)	OK	Crash
Speed	1,83	4,52
Current	1,81	4,41
Pitch	1,93	5,17
Voltage(v)	59,14	58,68

So we have a lower voltage (may have affected) but curiosly a big increase in sampling period.

I don't know if it is an artifact of the sampling and transmitting to 9BMetrics but linked to the crazy current reversal of the controller, one may think that the controller really choose a wrong way, was working with excesive delay and got messed.

Another interesting think is that maximum current when braking seems to go about 15A. I have not seen a lower value.

 

 

Paco, do you have any idea what is the microcontroller being used in the Ninebot.  I'm curious as to how fast it can crunch the numbers, and whether there is the potential for 'whiplash' effects that may cause the idiosyncrasies in behavior that your suggest in your analysis...

[Frode]

1 hour ago, Mono said:

One can only see that -15 amps lead to more deceleration than +50 amps

There is no + or - amps (except from between the battery and the mosfets). In the motor the current have increased from 15 amps to 50 amps - both goes in the same direction through the motor windings (if plug breaking has happened). What has happened is that instead of the motor driving current into the battery, the terminals (in effect) are reversed so that the battery and the motor (the back-emf that drove the current into the battery) now both drives the current in the same direction. If you reverse the current in the motor windings, the direction of the torque will inevitably reverse - and the result is that motor will no longer brake (which is actually what I suggest happened). From the motors perspective current from regenerative braking and plug  braking is indistinguishable (except that you are able to drive much more current through the motor in the later case because battery work with the motor instead of against it).

 

 

1 hour ago, Mono said:

which does not necessarily imply that it generated more torque

I have always learned that the torque from the motor and current through the motor are proportional (for a motor like this, so that for example 30 amps give twice the torque of 15 amps). If more torque is not generated, the current from the battery have to have been directed to another place. This is why I don't think this is plugging braking.

 

There is also something strange about the idea of draining extra energy from a battery when you need to brake. Due to the law of conservation of energy, the result will be that the system need to loose even more energy, the original energy from the moving driver/EUC + the energy you throw in from the battery. You cannot burn energy away with energy. They add up instead.

[Mono]

1 hour ago, Frode said:

1 hour ago, Frode said:

There is no + or - amps

I just used the numbers with sign as they appear in the figures, which also makes it easier to see to which number I was referring to. Why do you think we see positive and negative numbers for current if "there is no + or - amps"? 

 

1 hour ago, Frode said:

If more torque is not generated, the current from the battery have to have been directed to another place. This is why I don't think this is plugging braking.

Right, we are only in disagreement as to whether more torque was actually generated. I think it was. If there wasn't, then the shown current would be very very strange and needed indeed some explanation. 

1 hour ago, Frode said:

There is also something strange about the idea of draining extra energy from a battery when you need to brake.

Draining energy from a battery to create a force/torque doesn't sound that strange to me. 

1 hour ago, Frode said:

Due to the law of conservation of energy, the result will be that the system need to loose even more energy, the original energy from the moving driver/EUC + the energy you throw in from the battery.

Agreed. 

1 hour ago, Frode said:

You cannot burn energy away with energy.

Somewhat you can. It's called heat. 

 

[Frode]

3 hours ago, Mono said:

Why do you think we see positive and negative numbers for current if "there is no + or - amps"? 

What I meant to clarify was that it does not show the current through the motor, only through the battery (+ for draining, - for charging). From what you say I believe you agree to that. Current through the motor increases from 15 amps to 30 amps (which should mean much more braking force) at the same time as the tilt makes a sudden skip backwards. This is counterintuitive. Intuitively it should at least have started to slow down the slow but steady increase just before this.

 

10 hours ago, Mono said:

Interesting question in any case when we see high currents. I guess the answer is similar to where are they cooked when accelerating strongly from low speed

Energy is not cooked when you accelerate. It is much like charging a battery, but instead of converting it to chemical energy in a battery you convert it to speed (E = 1/2 m v^2). And if you go upwards you convert it to potential energy (E = m*g*h).

 

5 hours ago, Mono said:

Somewhat you can. It's called heat.

Yes, you have to convert it all to heat, the speed of the driver/EUC + the extra power drained from the battery.

 

5 hours ago, Mono said:

Draining energy from a battery to create a force/torque doesn't sound that strange to me. 

Yes, for a short time (like an emergency situation like this?) this might be possible without burning the motor controller and/or motor itself. Peak power is in this case 3100 W (when the amps is at the peak, seen in the Power.csv file). In the duration of the plugging braking (in this case) the system consumes a roughly estimated 175 Ws of power before it brakes down for some reason (not overheating I beleive). This is only a fraction (10%? 15%?) of the total energy needed to bring the driver + EUC down from 20 km/h to stand still. Could it be that plugging braking is just to hard for a wheel like this, so the battery (or something else) will get overloaded and the BMS (or something else) cuts/reduces the power? It could be an explanation.

But would this be a sound construction? Looking at these graphs (and assuming you are right about plugging  braking) I'm not sure if I like it... Humans don't react well to that fast changing conditions (< 1 second from everything is OK til wheel turns off).

BTW: I'm not sure if I disagree with you about what actually happened. But I definitely find these values disturbing. I'm not sure I like them any better just because the wheel might be designed to do what it did... To me, it seems like the trouble really starts when the current reverses...(there is however nothing there that clarifies wether the pitch angle caused the current or if it was the other way around)

Don't think there is much more to get out of this tough. If I was to brake hard, I mean really hard, I would apply braking force gradually. Remember that performed work (=energy) equals force times distance: W = F*s. When you travel fast you travel a longer distance in one second. This means that for a given braking force you must cook (convert to heat) more energy pr second when the speed is fast than when it is slow. This in turn means that if you want to cook a given amount of energy pr second while braking, you should gradually increase braking force while slowing down. Might be that this strategy is better for an EUC than hammering in the brakes.

[Mono]

2 minutes ago, Frode said:

Current through the motor increases from 15 amps to 30 amps (which should mean much more braking force) at the same time as the tilt makes a sudden skip backwards. This is counterintuitive. Intuitively it should at least have started to slow down the slow but steady increase just before this

Maybe it was reacting too little too late and already too close or over the lean limit, which would indeed partly be a controller issue.

6 minutes ago, Frode said:

Could it be that plugging braking is just to hard for a wheel like this, so the battery (or something else) will get overloaded and the BMS (or something else) cuts/reduces the power?

That is what I assume from all I know. When training emergency braking from rather moderate speeds, I cannot repeat this very often until I have the feeling the braking becomes much softer.

9 minutes ago, Frode said:

But would this be a sound construction? Looking at these graphs (and assuming you are right about plugging  braking) I'm not sure if I like it... Humans don't react well to that fast changing conditions (< 1 second from everything is OK til wheel turns off).

I don't see much alternatives on the horizon, as my assumption is that generative braking is not strong enough (I believe this is what an IPS 132 does, which has terribly weak brakes). I really like to have one emergency brake every 10 minutes at my disposal compared to none. In my experience there is the same problem when pushing the wheel strongly in any other way. It does change behavior and it is not easy to predict where the point of outlean lies at any point in time. In essence, you need to feel it and react timely and this usually work. Still troublesome, but what can you do...

I assume the main problem here is in the end that the motor is just a little too weak, even if a better controller software might have well avoided this particular incident. 

28 minutes ago, Frode said:

This means that for a given braking force you must cook (convert to heat) more energy pr second when the speed is fast than when it is slow. This in turn means that if you want to cook a given amount of energy pr second while braking, you should gradually increase braking force while slowing down. Might be that this strategy is better for an EUC than hammering in the brakes.

That's an interesting point. I asked myself a while ago what the best braking control strategy is to get the shortest braking distance and concluded that it must go to the limit of tire friction the entire time. Now given an additional constraint, namely a bound on the maximal energy which can be dissipated, what is the control solution to the shortest distance exactly?

[Frode]

14 hours ago, Mono said:

I really like to have one emergency brake every 10 minutes at my disposal compared to none

If we are seeing plugging braking (still not sure about that):
To me it looks like you will have it for only about half a second, and it will only be able to take away between 10% and 15% of the velocity energy before something brakes down. It is very strange that the pitch increases that fast when (if) the braking force was increased 2-3 times. At least the rider must have done something very different in the two cases (before he falls), so different that I believe the rider must have noticed that he did so. I mean, you have 2-3 times more braking power and just fall on your ass because you push a little on the wheel for a brief time (sub second). He must have pushed hard, much harder than in the other no-crash case.

As you can see in the other case, the rider manages to brake more than enough without it (from same initial speed). he is down to walking speed in about half a second and would probably have been able to stop to stand still before another half second. That is hard braking. I would also think that plugging braking would feel like the brakes are slammed on (the current in the motor suddenly increases from 15 amp to 30-50 amps in very short time). Maybe it amplifies, or even induces the instability that made the rider fall?