I have a working Bessler wheel in my simulation !

[hdeasy]

Well, on the balls in my system: they are firmly attached to the arms. The only degree of freedom they have apart from the main wheel rotation is the bearing I indicate in the sketch by a small circle. This can be seen as a sort of hinge about which the ball-arm can rotate except when it's prevented from doing so by the radial arm.


So when the wheel rotating CW rounds the bend at the top, the support of the radial arm falls away and the weight can now swing on the hinge bearing - oh yes, one point maybe not obvious from my description is that only the extension of the radial arm beyond the bearing can block the ball arm- it projects further out than the lower arm. So in the image at the top the ball-arm is free to oscillate almost 360° as the lower radial arm does not impede its motion.


Now on unforeseen mechanical effects - I know that if a resonance occurs between the main wheel and the pendulum oscillation, it can come to stand-still etc. But with 1cm arms, the period of oscillation is about 0.2s. That's 5hz or 300 RPM. As long as the wheel stays under that period, all should be good. In fact it could come close to it but cannot exceed that speed. But even at 120 RPM that would be phenomenal. The fact that the arms are small to reduce gravitational differential also helps decouple them from the main wheel motion.

[mlpmlg]

The original post of Stefan's device does not work. I tried putting it into algodoo. Looking at the device, intuitively there does not seem to be any reason why it would work in the first place. Algodoo doesn't seem to have "pixel perfect" accuracy and doesn't seem to set the position to 0, but rather some floating point of 0.00001 or something like that. That being said, if the device needs a precision of 0.00001 to work, it will probably never be built in real life, and also not be able to provide any usable energy more than 0.00001 energy.

There is another post, a device by Ghazanfar_Ali. In the video of Algodoo, the device is clearly tunneling between the two objects. If built correctly without relying on glitches of Algodoo, the device does not work in Algodoo or real life. One would expect the Algodoo file posted alongside to prove it works but the download is nowhere to be found.

[a_user]

Seen this: https://youtu.be/6tKy0jFD_e4

[hdeasy]

I analyzed that video frame by frame and was able to confirm my suspicion that the wheel was slowing down in the video. I attach my measurements with time in video on the x axis and wheel rotation time on the y axis, found by following a feature round.

[hdeasy]

I forgot to mention - both x and y axes are in seconds on that plot.