Hi again, ventilator!

I'm experiencing some problems using small entities:
I'm using the meter-to-quant conversions you set in your example. I'm simulating a sphere falling down on a surface (treecollision). According to the conversion, the sphere is about 3.7cm in diameter (~1.2 quants). I want the sizes to be as close as possible to the real world, so that the newton simulation is as realistic as possible.

The problem is that, falling from only 20cm high the sphere will pass through the surface. If I time slice it so that newton is updated at least ~45 times per second the sphere will stand on the surface, but fall through when it rolls across an edge of a treecollision polygon. In addition, spheres will not collide with each other nor with other physics entities, even with an absurdly high frequency of update (1024 Hz, which is the maximum my laptop will handle...).

However:

If I set METERTOQUANT and QUANTTOMETER both to 1 I will get the expected result (except for size related differences, such as the need to increase gravitational acceleration). On the other hand, If I maintain the 32 quant per meter ratio, but increase the size of the models, I will get the expected results (same size related differences as above). If I increase model size, but also increase QUANTTOMETER accordingly, the passthroughs will happen again.

Is this a known problem or do you believe it is a bug in my implementation? Maybe centimeter scale simulation is not accurate enough... In that case is it valid to use larger models/acceleration constants/forces... or will it result in unrealistic behaviour?

Thank you!