Measurements were taken at 2 different days.
First graph ACW spins and is showing in milliseconds time differences in between rotations. It wasn’t the same as CW.
Graph 3 showing differences and the yellow line is an average.
Blue lines is how the FFA would work because it learns wheel separately for each ball direction. Yellow line is what the FFZ would use.
Due to system design it wouldn’t make big difference. 10ms would create errors from 0-2.5 pockets dependently where in time we took sample.
Difference in between CW and ACW is responsible for different offsets for each direction.
Interesting is green line. Next day the ball was decelerating faster in both directions, no wonder why saved data didn’t work well and I had to reset the system.
Yes, my point was that there is enough difference between CW & ACW to justify the FF learning each of them separately instead of just using the average but also the fact that you are finding there are sometimes more “spinners/rollers” in one ball direction than the other suggests that it will be a worthwhile change in the FFA.
Was there any other obvious reason for the change in deceleration or do you think it was just atmospheric air pressure?
Clean the wheel and you may get blue line, after few hours it may be as green.
It could be different ball.
On both graphs CW and ACW the green line is about 12 ms higher so something definitely happened.
This is change per rotation and not per sec, and here is different view.
So if we look it to time reference then on graph 2 deceleration is not decreasing while at first graph it already start decreasing. Which means that the knee point is somewhere there but for CW direction is coming a bit later. It could be because of deformation that causes ball to exit at same diamond only in one direction.
I wouldn’t say that there is much improvement in recording both directions.
But who knows maybe with other scenarios it may be different. We win a bit but we lose because the system needs few more spins to lern.
I do it when the ball hits the rotor but it isn’t so important because we do not need exact point of ball hitting the rotor. We need constant distance. +or €“ 0.5 sec wouldn’t make prediction wrong it will only shift offset. We can not have exact point if we do not know how much the ball made in last revolution from clocking point. During redesign I was thinking about that a lot but due to so many problems it would be so complicated and still close to imposable. With new changes on tilt prediction, we could be wrong by more then one sec but since that time is only for rotor calculation it doesn’t matter much. If rotor changes from 10 pockets/s to 18p/s it would matter (by 8 pockets) but anyway then we need to review our prediction since it is huge change in rotor speed.
I was thinking more about ACW and CW separately recorded.
It my be more important then I explained. What if we play few spins in a row in one direction. The average would go close to that direction and the error may increase when we play another direction.
Saving system settings.
I think it may be useful on leveled wheel.
If parameters have changed we may get different offset which we can readjust but on tilted wheel we may get good spins unpredicted and bad ones predicted. Perhaps by getting ACC data from system we can get idea where to move clocking point to get it corrected.
I think in real game would be better to start from the start. 3-4 spins to tune it isn’t much.
[quote]I was thinking more about ACW and CW separately recorded.
It my be more important then I explained. What if we play few spins in a row in one direction. The average would go close to that direction and the error may increase when we play another direction.