For some time I have had the suspicion that something limited the top speed of soda models. I have come to this suspicion because different models have similar top speeds for the same tracks when the height is allowed to change.
After some experiments I have come to the conclusion that the maximum speed of individual masses in soda is 20 pixels/frame.
I have done the following experiment: in Sodaconstructor beta, I have dropped a mass from a height and saved the XML file just before it hit the ground. The gravity was set to maximum and the friction to 0. I found that whatever height I dropped the mass from, the highest velocity was always 20 pixels/frame.
This means that the maximum speed for rolling models (amoebas, cars, turbines, etc.) is 10 pixels/frame. This is because the top of the wheel is travelling at double the speed of the wheel and the bottom of the wheel is not moving.
I have checked this maximum speed with many races in which I have competed and found that no real models have actually been able to travel faster than this speed. I have also found that many of the GA optimized models come quite close to this limit.
This means that if a racetrack is 3000 pixels wide, the theoretically fastest time is 3000/10 = 300 frames. This of course only applies to a flat track, because when the models go over bumps the effective travelled distance will be larger.
My question to Ed is: would it be possible to increase this maximum speed? Why is it built in? Is it because of numerical instability that the velocity needs to be limited? There is no physical equivalent to this except for the speed of light, but that is much much larger than the typical speed normal objects travel at.
It would be very interesting to see what the real top speed of models is when this artificial limit is removed.
|
