The video explicitly states that it's not the Bernoulli effect that keeps the ball up and gives arguments to support the claim. In short, they say that the impact, adhesion, and change in direction of the water lift the ball and force it towards the stream.
Nice work. 4500 RPM at 240 Hz works out to 3.2 ticks per revolution. I'm not sure what the minimum ticks per rev is, but once you hit that, then the only way to higher revs is to increase the sim.frequency.
Statement: I did notice a problem which I'm pretty certain is due to the poor precision of the Algodoo polygon generator.
Response: Yes, Algodoo automatically reduces the resolution of large polygons.
S: I generated a large gear which had 106 teeth, and then I cloned it and carefully meshed the two gears with axles at their exact centers.
R: The gear generator automatically puts the axles at the exact center of the gears and also creates pitch circle to aid in alignment.
S: It was almost impossible to mesh the two gears so that they moved very smoothly.
R: I was able to get them to mesh relatively smoothly (speed set at 60 RPM (6.28 rad/sec) resulted in speed variation of 5.7 to 6.7 rad/sec). Motor torque was set at 8000 Nm. Gear diameter ~ 10m. 106 teeth. Sim.frequency = 240 Hz. 60 RPM is 1 rev/sec or 106 teeth/sec. If 2 contacts per tooth then 212 Hz minimum sim frequency.
S: The driven gear kept jamming along certain sections of groups of teeth.
R: An off-center hinge could cause this.
S: When zoomed way in to view the area where the teeth meshed, I was able to see the poorly formed teeth.
R: The teeth are not quite as bad as they look. Cut out a tooth and look at it. You will see the tooth's true form.
If anybody knows how to increase polygon resolution, then please explain how to do it below.
P.S. I cut each gear into eight pieces and then glued them back together. The angular velocity variation was a lot less. The speed was 6.280 to 6.283 rad/sec.
Statement: This barely holds back anything! The water is mostly a box!
Response: The force on a dam depends only on the depth of the water against the dam and not the volume of water that is held back by the dam. The boxes can be converted to water and the pressure against the dam will not change. I used boxes and circles as fillers in order to minimize lag.
panda's "8>5>3 while 8 is filling 5" method does work. I think that his solution is the fastest and most creative. This two step method can be viewed two ways, either panda is breaking the rules, or Xray's scene does not work as well at it should.