What determines wheel zippiness?

[Aneta]

3 hours ago, Mike Sacristan said:

@Aneta i'm beginning to wonder if you really own a wheel or if you just are here to discuss physics. 
And @Mono has a very mesmerising avatar which I can't stop looking.

When climbing a slope the contact patch is further toward the front of the tyre. This puts the bias of our weight behind the center of the wheel.
This has been calculated and illustrated on this forum before.
To overcome it you need more forward lean. Tilting the pedals backwards will cause a dorsiflexion penalty making it very physically demanding so that only a few people will be able to keep their heels on the pedals. Olympic squatters come to mind.

I made figure 3 just for laughs to illustrate forward mass distribution with pedals tilted backwards.

Like I said in the other thread.. I understand the misunderstanding because when I tilted my pedals back many months ago I did it in the belief that it would emulate falling to neutral and hence I would get "effortless acceleration" because I had an X degree advantage (because gravity you know). As it has now been explained by others backward tilting simply changes the position of the pedals and the whole wheel and for our bodies a pedal decline worsens the lever making forward movement more demanding but braking easier. This is easy enough to try in real life by standing on your heels and then squatting vs standing on the balls of your feet and squatting. Or do it on a -7 + 7 decline/incline or a wooden board.

I wonder if there's a fundamental difference between how KS and RW implement the balancing algorithm, more specific, perhaps Kingsong uses angular velocity as throttle and Rockwheel uses angle? This would explain the difference between Mike's results and mine. KS tries to maintain the pedal angle constant no matter what, while RW only maintains the angle corresponding to required throttle?

@Chriull, how do we know that throttle!=alpha, throttle=omega? (Jesus, I think I heard these words, "I'm alpha and omega" before.) Can you think of any experiment that we can do to establish this fact with certainty?

Edited December 13, 2019 by Aneta

[mrelwood]

8 hours ago, Mono said:

I don't understand the point. The maximum acceleration we can only get at zero speed. That is, we talk about "time it takes to reach 1km/h" instead of reaching 24km/h?

Yes, the wheel’s maximum acceleration is also not linear. I’m trying to point out that zippiness and maximum acceleration are different measures. Like the acceleration of a car vs. the travel length of the gas pedal.

I once drove a car X that was a tiresome drive since the gas pedal reacted instantly already to a tiny press. Last summer I drove a car Y that needed a long push to the gas pedal in order to keep up with city traffic. Which car would accelerate faster to 60mph?

The car X was a 100hp Toyota family wagon (12sec 0-60mph). Car Y was a 270hp Audi TT S (5sec 0-60mph)

6 hours ago, mike_bike_kite said:

I'm afraid I still disagree. We can measure 1/4 mile times on motorbikes

Have you timed how fast you can accelerate with your EUC?

Motorbikes and all other vehicles except self-balancing ones can be used to accelerate with maximum acceleration. Self-balancing vehicles can’t. Sure, cars still get slightly different 1/4 mile times with different rider skills, but a with a car you can just floor the gas pedal and experience the maximum acceleration.

A self-balancing vehicle doesn’t behave like that at all. When we accelerate, we take a forward lean and hope the wheel has enough power to catch us. If it didn’t, we crashed. There wouldn’t be many acceleration tests with cars if they ejected the driver at the instant the maximum power was reached.

Also, since this thread is about wheel zippyness and your calculations clearly prefer small diameter wheels, it seems that the maximum acceleration is not what you are going for after all. 

Quote

given enough data you could make a reasonable stab at which wheel was faster accelerating.

I’m sure you can’t. It might have a slight tendency to favor wheels that require the least effort to accelerate, but even that would require an assumption that at least some people will try to accelerate as fast as possible.

 

Quote

perhaps Kingsong uses angular velocity as throttle and Rockwheel uses angle?

I’m sure they use the same fundamental mechanisms.

Quote

Can you think of any experiment that we can do to establish this fact with certainty?

Switching to the hardest riding modes might give more similiar results. Then go on a group ride and repeat the experiment with a few people on a few different wheels.

Edited December 13, 2019 by mrelwood

[UniVehje]

This is an extremely interesting discussion. It is important to realize that ultimately we are discussing perceived acceleration, how it feels, what feels zippy. Absolute maximum acceleration of each wheel might be academically interesting topic but in reality that is not how we ride. What we experience is the effort it takes to gain speed. I think It’s probably some combination of power, wheel diameter, pedal softness/hardness and firmware algorithms. 

Here are my observations: 

1. My 800W V8 feels zippier than my 2000W 18XL. This will always be true even though the 18XL might be able to accelerate faster in absolute terms if taken to the very edge (just before face planting). It is true even if they actually did accelerate just as fast when measured in lab conditions. The bigger wheel will require more leaning and that feels like a bigger rider input. 

2. My gut feeling is that I could reach first 5 meters or 10 km/h quicker with my V8 but 100 meters or 30 km/h speed quicker with the 18XL. Simply because I would be more confident leaning heavily on the 18XL. And I would still think the slower to 30 km/h V8 would feel zippier and less effort. 

3. You can only test different wheels side by side using robots or lab conditions. I can now accelerate faster with my 18XL than I could when I first got it. I can also now ride up hills that I could not ride when it was new to me. If you just jump on a new wheel it will feel different and sluggish because your brain is not yet tuned in. Just changing the riding mode to hard, medium or soft will feel wrong initially and takes about 50-100 km to get used to. 

4. Wheel diameter seems to be the biggest factor but firmware is not insignificant. Inmotion updated V10F firmware to include a setting that felt zippier. Same happened to my 18XL with 1.13 firmware. It probably didn’t get quicker in absolute terms but definitely felt zippier and more responsive after that. If we could find out what exactly they did to the firmware that would help us understand this topic better. 

 

[Aneta]

1 hour ago, mrelwood said:

I’m sure they use the same fundamental mechanisms.

We could be sure about this if this mathematical problem had only one way of solving it. But many math problems have multiple ways of solving them. Finding roots of function (y(x) = 0), finding minima/maxima, etc. etc. Even proving Pythagorean theorem can be done in gazillion ways.

At this point, I don't see why the balancing cannot be done by using angle as throttle. If the balanced position is a root of some function, the bisection method (https://en.wikipedia.org/wiki/Bisection_method) can be used to find it, by basically "boxing" the root by smaller and smaller intervals. The motor "overshoots" the point of balance (throttle applied is more than needed), then undershoots back by half the amount, overshoots again by half of that half, etc., until it's "good enough". Rinse, repeat for the next moment of time. Since this is done hundreds of times per second, we don't notice it. However, GT16 is known for its famous "high voltage power lines" buzz when idle (some people even thought these were sparks in the controller or the motor), which is absent in many other wheels, so I wonder if it's an indication of a different algorithm, and perhaps even exactly the "boxing" method I just described.

Edited December 13, 2019 by Aneta

[Aneta]

7 hours ago, Mike Sacristan said:

I made figure 3 just for laughs to illustrate forward mass distribution with pedals tilted backwards.

Like I said in the other thread.. I understand the misunderstanding because when I tilted my pedals back many months ago I did it in the belief that it would emulate falling to neutral and hence I would get "effortless acceleration" because I had an X degree advantage (because gravity you know). As it has now been explained by others backward tilting simply changes the position of the pedals and the whole wheel and for our bodies a pedal decline worsens the lever making forward movement more demanding but braking easier. This is easy enough to try in real life by standing on your heels and then squatting vs standing on the balls of your feet and squatting. Or do it on a -7 + 7 decline/incline or a wooden board.

To make it clear, in my experiment your drawing #3 was when riding on level surface with the pedals calibrated with tilt up - weird, uncomfortable. Once I was on the steep slope, it was exactly like #4 - effortless, just standing on a magical invisible escalator. I absolutely cannot mistaken this feeling of effortlessness #4 with the strain of #3. If I understood correctly, #3 is how you were riding the slope in your experiment years ago (what wheel was that back then?). Hence, I'm completely puzzled as to why such dramatic difference, and have a suspicion that Rockwheel's and Kingsong's (if it was KS) algorithms are different at the very core.

[mrelwood]

 

Quote

I don't see why the balancing cannot be done by using angle as throttle.

Perhaps it could. But while I’m also eager to know stuff, I’m not expecting you and me to resolve the operational gyro sensor algorithm guidelines for self-balancing vehicles at a discussion forum. Instead, the most common or simple answer is most often correct. Which is why I’d pick my combination of explanations from the following:

- A nameless person at a discussion forum senses things to happen differently than they actually do. (Very common, especially with self-balancing vehicles, which operate on witchcraft anyway.) Especially probable if the nameless person would for example describe riding uphill without leaning forward. (Against the law of physics.)

- A factor that either person didn’t remember to, or come to mention. For example things that would’ve been clear without mentioning for the other person, while new or forgotten to the other. (Quite common in discussion forums.)

- People are not very good in explaining their thoughts and findings in writing. (I’m not.)

- People from different countries are not equally good at writing in English, so small misunderstandings may bundle up to an actual failure in delivering a point across correctly.

 

Any (combination) of the above is highly more probable than a random person who’s somewhat interested in physics being the first to figure out that his/her (first?) wheel (which was introduced 2 years ago) responds fundamentally differently to gyro sensor data than all other manufacturers’ wheels.

Actually, I believe advanced lifeforms existing on another planet to be much more probable.

Edited December 13, 2019 by mrelwood

[Chriull]

7 hours ago, Aneta said:

I wonder if there's a fundamental difference between how KS and RW implement the balancing algorithm, more specific, perhaps Kingsong uses angular velocity as throttle and Rockwheel uses angle? This would explain the difference between Mike's results and mine. KS tries to maintain the pedal angle constant no matter what, while RW only maintains the angle corresponding to required throttle?

@Chriull, how do we know that throttle!=alpha, throttle=omega? (Jesus, I think I heard these words, "I'm alpha and omega" before.) Can you think of any experiment that we can do to establish this fact with certainty?

For the theory of EUC riding no pedal tilt angle or angular velocity is needed. In a perfect model the pedal stays horizontal all the time - the pressure/force of the rider on the pedal is the only neede input.

As not the force on the pedal is taken as input to the controller but the quantized angle/angle change one is one step further of the "ideal model".

In our real world every control loop has delays and by this the pedal tilts after applying a force until the motor can create the countertorque. How the pedal tilts is "just a side effect".

Additionally there are (overall) control loops corner cases like current limiting/motor limit.

Some riders like soft mode for riding as this delayed reaction is even offered in "amplified" versions

As human brains work very good/?mostly? as pattern matcher - we try to find correlations - like once we notice that the higher burden the is we ask from the wheel (acceleration/incline) the more the pedal tilts we turn the causality relation to the more we tilt the pedal the more power the EUC delivers. This conclusion/observation is not wrong,  but this tilt is not the root cause but an (more or less wanted) side effect.

[Chriull]

12 hours ago, mike_bike_kite said:

We can measure 1/4 mile times

Sounds like a great idea for a new EUC competition!

Could be that the only feasable wheels for this kind of race are GWs...

And it will be more something like a 15-30 m race...

But should lead to interesting challenges:

- how to get the initial force on the pedal for acceleration for a quick start (without frying the mosfets )

- to keep the "maximum" acceleration before the "high speeds" are reached (without melting cables/frying mosfets (again...))

- slightly reduce the acceleration before one reaches the motor limit and overleans

Hopefully mosfets are "stable" enough for this - but drag racing is an expensive sport anyhow
Good upholstery and "skidpads" should keep injuries assessable! (for the daring ones...)

[Mike Sacristan]

5 hours ago, UniVehje said:

This is an extremely interesting discussion. It is important to realize that ultimately we are discussing perceived acceleration, how it feels, what feels zippy. Absolute maximum acceleration of each wheel might be academically interesting topic but in reality that is not how we ride. What we experience is the effort it takes to gain speed. I think It’s probably some combination of power, wheel diameter, pedal softness/hardness and firmware algorithms. 

Here are my observations: 

1. My 800W V8 feels zippier than my 2000W 18XL. This will always be true even though the 18XL might be able to accelerate faster in absolute terms if taken to the very edge (just before face planting). It is true even if they actually did accelerate just as fast when measured in lab conditions. The bigger wheel will require more leaning and that feels like a bigger rider input. 

2. My gut feeling is that I could reach first 5 meters or 10 km/h quicker with my V8 but 100 meters or 30 km/h speed quicker with the 18XL. Simply because I would be more confident leaning heavily on the 18XL. And I would still think the slower to 30 km/h V8 would feel zippier and less effort. 

3. You can only test different wheels side by side using robots or lab conditions. I can now accelerate faster with my 18XL than I could when I first got it. I can also now ride up hills that I could not ride when it was new to me. If you just jump on a new wheel it will feel different and sluggish because your brain is not yet tuned in. Just changing the riding mode to hard, medium or soft will feel wrong initially and takes about 50-100 km to get used to. 

4. Wheel diameter seems to be the biggest factor but firmware is not insignificant. Inmotion updated V10F firmware to include a setting that felt zippier. Same happened to my 18XL with 1.13 firmware. It probably didn’t get quicker in absolute terms but definitely felt zippier and more responsive after that. If we could find out what exactly they did to the firmware that would help us understand this topic better. 

 

Very well said!

In the case of my 16X I went for a ride yesterday after catching up on this thread.
I changed pedal mode to the medium mode and went for a 20 km cruise. There was much less effort involved.
At first I tried the soft mode but for some reason it didn't feel as zippy as the medium mode.
16X pedals have a certain flex to them depending on pedal mode. Medium seems to have the most flex.
When I lean and thus put pressure on the pedals either accelerating or braking the pedals will yield and ever so slightly change angle.
The FW 1.07 is harder and with less flex than FW 1.05. I can only feel the flex when hitting a curb, roots or rocks and only if I do so unexpectedly.
When going back to the MSX and the hard mode there I instantly feel that there is pretty much zero flex.

As the 16X pedals are larger than the MSX pedals I have been riding with my feet a bit more forward compared to the MSX.
I have been using the front of my feet / balls of my feet to determine my position on the pedals. I am now using my heels instead.
While being a bit too far forward on the pedals I have always had the sensation that the 16X is "running away from me" and also hard to slow down.
When riding in a more centered position yesterday on the hard mode I had to work quite a bit to get to 40 kmh.
In the medium mode I just got to 45 kmh with very little effort and could stay there until I hit soft tilt-back. Soft tilt-back neutralised my effort forward and slowed me down to below 45 kmh upon which I put a tiny bit of more forward effort into the wheel to maintain the 45 kmh so I could beep along (45 kmh is where I set my alarm) for a few hundred meters. At one point I did press a tiny bit too hard and got a 4 beep alarm.

[Mike Sacristan]

4 hours ago, Aneta said:

To make it clear, in my experiment your drawing #3 was when riding on level surface with the pedals calibrated with tilt up - weird, uncomfortable. Once I was on the steep slope, it was exactly like #4 - effortless, just standing on a magical invisible escalator. I absolutely cannot mistaken this feeling of effortlessness #4 with the strain of #3. If I understood correctly, #3 is how you were riding the slope in your experiment years ago (what wheel was that back then?). Hence, I'm completely puzzled as to why such dramatic difference, and have a suspicion that Rockwheel's and Kingsong's (if it was KS) algorithms are different at the very core.

Yes I understand exactly. I am a professional understander. 
I've been married for 9 years.
I discovered many years ago that it is impossible for me to be fully understood so instead I focus on understanding others.
I hate the world a tiny bit less now and it is a work in progress albeit a slow one.

The only thing I can think of is that the GT16 would sense the incline and thus straighten out the pedals from -7 degrees to 0 degrees to facilitate the climb.
This would be quite a lot of pedal flex but if the gyro is aware that you are riding at -7 and allows for flex to reference point of 0 it could be possible.
It would be interesting to test the opposite scenario with +7 degrees forward incline or perhaps something a little less dramatic so that you don't slide off the pedals.

The experiment was done in March on my MSX on all pedal modes.
I have also tried on the Tesla, 16X, 18XL, V10, Nikola, V8 and Ninebot One E+.

[Mike Sacristan]

42 minutes ago, Chriull said:

Sounds like a great idea for a new EUC competition!

Could be that the only feasable wheels for this kind of race are GWs...

And it will be more something like a 15-30 m race...

But should lead to interesting challenges:

- how to get the initial force on the pedal for acceleration for a quick start (without frying the mosfets )

- to keep the "maximum" acceleration before the "high speeds" are reached (without melting cables/frying mosfets (again...))

- slightly reduce the acceleration before one reaches the motor limit and overleans

Hopefully mosfets are "stable" enough for this - but drag racing is an expensive sport anyhow
Good upholstery and "skidpads" should keep injuries assessable! (for the daring ones...)

Something similar was done here.
And at the end the MSX rider actually started to reach the limits of the wheel upon which it started to dump him forward. He was able to pull back and recover though.

 

[Mono]

4 hours ago, Aneta said:

At this point, I don't see why the balancing cannot be done by using angle as throttle.

I don't see how the controller can not use a monotonous transformation of the pedal angle as throttle input. It may not be the only input, but I don't see that it is possible in practice to keep the pedals calibrated for any longer period of time without using the angle.

[Mono]

19 hours ago, Chriull said:

Or the rider "accelerates" himself forward against the pedal. This would need some additional torque/acceleration from the wheel to keep the pedals tilted.

Right, I didn't think about it, but the rider could just make a tiny step forward on the pedal  instead of slowing down the wheel

19 hours ago, Chriull said:

If the whole force on the pedals is removed the wheel will not accelerate anymore...

Indeed, good point! The acceleration is proportional to (applied weight) / (applied weight plus wheel weight). Just releasing weight is even counterproductive in the static equation! The important aspect of the knee-bending maneuver is hence to suspend the horizontal component of pushing the riders weight (for a short moment), so that the weight will disappear in the denominator, meanwhile keep pushing on the pedals vertically (dynamically?). Nicely spotted. Makes perfectly sense now  The bending the knees procedure is more delicate to model than what I had imagined.

 

Edited December 13, 2019 by Mono

[Aneta]

9 hours ago, Mike Sacristan said:

Yes I understand exactly. I am a professional understander. 
I've been married for 9 years.
I discovered many years ago that it is impossible for me to be fully understood so instead I focus on understanding others.
I hate the world a tiny bit less now and it is a work in progress albeit a slow one.

The only thing I can think of is that the GT16 would sense the incline and thus straighten out the pedals from -7 degrees to 0 degrees to facilitate the climb.
This would be quite a lot of pedal flex but if the gyro is aware that you are riding at -7 and allows for flex to reference point of 0 it could be possible.
It would be interesting to test the opposite scenario with +7 degrees forward incline or perhaps something a little less dramatic so that you don't slide off the pedals.

The experiment was done in March on my MSX on all pedal modes.
I have also tried on the Tesla, 16X, 18XL, V10, Nikola, V8 and Ninebot One E+.

There's no way for EUC to sense the incline. It only has an accelerometer and a gyro, which only measure the angle of the motherboard (read: pedals) relative to apparent gravity vector, and the rate of rotation.

(Maybe in the future when Skydio hops on our train and starts building segwheels, it will build the surrounding 3D world with cameras like their drone does and know it's on an incline... but then I woke up - garage-level Chinese EUC manufacturers are here to stay for years.)

[Aneta]

9 hours ago, Mono said:

I don't see how the controller can not use a monotonous transformation of the pedal angle as throttle input. It may not be the only input, but I don't see that it is possible in practice to keep the pedals calibrated for any longer period of time without using the angle.

This. It definitely uses the angle of accelerometer/pedals in maintaining the balance, but the angle is not the only input, the rate of rotation of pedals (intent for more or less speed) is the second one. These are the two and the only inputs that the balancing algorithm can have.

[mike_bike_kite]

4 minutes ago, Aneta said:

There's no way for EUC to sense the incline. It only has an accelerometer and a gyro, which only measure the angle of the motherboard (read: pedals) relative to apparent gravity vector, and the rate of rotation.

Couldn't it estimate the incline from the amount of power required to maintain a steady speed (assuming you're riding at a steady speed)? The weight of the rider obviously comes into the calculation but given enough samples it should be able to work that out too.

[Aneta]

9 minutes ago, mike_bike_kite said:

Couldn't it estimate the incline from the amount of power required to maintain a steady speed (assuming you're riding at a steady speed)? The weight of the rider obviously comes into the calculation but given enough samples it should be able to work that out too.

There's simply no need for balancing algorithm to know what's the nature of the increased resistance - incline, stiff headwind, plowing through deep snow, etc. It just tries to keep the wheel balanced no matter what.

[mike_bike_kite]

Yep you're right, stiff head winds and deep snow would affect things, but under ordinary circumstances could it tell the incline from the speed, the amps being used and the voltage?

[Mono]

16 minutes ago, mrelwood said:

I think this exactly is one example of simplifying too much.

I didn't realize how simple it is to compute and visualize the effective leverage of the weight force. I added white bars in the figure of the original post. If we are not mistaken, the leverage is, somewhat surprisingly, not (at all) related to the pedal position or tilt, which is great news (simple). We also know that torque and counter-torque must be the same or otherwise the pedal tilt changes instantly.

From there, it seems that we can simply determine the torque (as the product of the rider weight and the length of the white bar), which, divided by half the wheel diameter, translates directly to forward thrust (up to pedal dip). What happens inside the wheel (FW, motor, battery) is for these considerations entirely irrelevant!

    forward thrust [kg] = weight x forward displacement / wheel radius

That's it. Under acceleration, the weight vector tilts forward and the forward displacement is measured at 90º to the tilted vector.

That means, AFAICS, that differences in zippiness can only be explained by

• the riders stance, stability and comfort (standing on or close to the pedal tip is just not practical or effortless) due to individual habits and the geometry of the wheel
• deviation of the controller from the exact torque demand inducing pedal dip, that is, changes due to FW updates etc. must affect the pedal dip characteristics to cause a perceivable change on the ride

Of course it could also well be that we are still missing something important 

[Mono]

51 minutes ago, mike_bike_kite said:

Yep you're right, stiff head winds and deep snow would affect things, but under ordinary circumstances could it tell the incline from the speed, the amps being used and the voltage?

Yes, but to what avail? It would make the incline an output variable, not an input variable  and superfluous as it was derived from the internal state.

You seem to suggest that the wheel should check current and speed and do something if the latter is lower than expected (that's basically how we can diagnose incline, or headwind...).

Edited December 13, 2019 by Mono

[mike_bike_kite]

It's to no avail at all. Anita just mentioned that "There's no way for EUC to sense the incline" and I just wondered whether there might be. Saying that, Anita did point out a few things that would stop the algorithm from working. Knowing the incline could be useful though for pointing headlamps but that's all I can think of.

[Chriull]

40 minutes ago, Mono said:

That's it. Under acceleration, the weight vector tilts forward and the forward displacement is measured at 90º to the tilted vector.

And the weight vector gets longer, as it's the resulting vector of the gravitational force (staying constant all the time) and the changing  "acceleration force". To the later force one could add air drag, too.

If one cancels the restriction for the "acceleration force" to be horizontal it will also work for inclines.

[Aneta]

27 minutes ago, mike_bike_kite said:

Anita just mentioned that "There's no way for EUC to sense the incline" and I just wondered whether there might be.

I meant that there's no direct way for wheel to measure the incline in the same fashion as you can measure it using accelerometer in your smartphone, or using a bubble level. It can be derived, but will involve so many inaccurate characteristics (such as torque curve of the wheel, variable tire pressure, rider's weight, including variable weight backpack and clothes, aerodynamic drag that varies with posture and clothes, head/tailwind) that the end result will be very inaccurate. For example, it will say it's a 20% grade, while it is actually 10. If exact incline is of interest, better to use a clinometer or Google Earth.

Edited December 13, 2019 by Aneta

[Aneta]

10 minutes ago, Chriull said:

And the weight vector gets longer, as it's the resulting vector of the gravitational force (staying constant all the time) and the changing  "acceleration force". To the later force one could add air drag, too.

If one cancels the restriction for the "acceleration force" to be horizontal it will also work for inclines.

To add to this, accelerometer - the essential tentpole of self-balancing vehicle, along with gyro - can only measure apparent gravity (vector sum of g and -a, where a is forward acceleration; air drag has no effect on apparent gravity), it cannot tell it from real gravity (except for its absolute value being different from 1g, but even that is not always - apparent gravity can be 1g, too).

Edited December 13, 2019 by Aneta

[Mike Sacristan]

 

 

17 hours ago, Aneta said:

To make it clear, in my experiment your drawing #3 was when riding on level surface with the pedals calibrated with tilt up - weird, uncomfortable. Once I was on the steep slope, it was exactly like #4 - effortless, just standing on a magical invisible escalator. I absolutely cannot mistaken this feeling of effortlessness #4 with the strain of #3. If I understood correctly, #3 is how you were riding the slope in your experiment years ago (what wheel was that back then?). Hence, I'm completely puzzled as to why such dramatic difference, and have a suspicion that Rockwheel's and Kingsong's (if it was KS) algorithms are different at the very core.

This would require the backward tilt calibration to allow tolerance towards level.
However it did not work that way when you were on level ground but it did when you were on the incline.  Maybe you weren't going fast/hard enough on level ground.

12 hours ago, Mike Sacristan said:

The only thing I can think of is that the GT16 would sense the incline and thus straighten out the pedals from -7 degrees to 0 degrees to facilitate the climb.
This would be quite a lot of pedal flex but if the gyro is aware that you are riding at -7 and allows for flex to reference point of 0 it could be possible.
It would be interesting to test the opposite scenario with +7 degrees forward incline or perhaps something a little less dramatic so that you don't slide off the pedals.

Yes I am aware of that I was just trying to comfort you after the unlikely paradox you presented. 

3 hours ago, Aneta said:

There's no way for EUC to sense the incline. It only has an accelerometer and a gyro, which only measure the angle of the motherboard (read: pedals) relative to apparent gravity vector, and the rate of rotation.

So basically you are saying that what you did was impossible.
And that is 100% fine by me because that's the way I see it too.