btBvhTriangleMeshShape tunneling

[dangerdaveCS]

I have some capsules with mass 0.001, which have been scaled up 500x, giving capsules of length 10 units and radius 2.5 units. I have also scaled up gravity by 500x to suit.

I then have a btBvhTriangleMeshShape, which is just a flat surface, tesselated so that each trangle is approx 1 unit along a side.

I then drop capsules from a height of about 25 units above the surface.

I find that unless I use a fixedTimeStep size of around 1/1000, most of the capsules fall through the mesh, and the ones that don't 'jitter' wildly. A fixedTimeStep so small means very very poor performance.

I've tried all sorts and can't figure out a good way around this issue. CCD functions have no effect, probably because btBvhTriangleMeshShape isn't convex.

Any help very much appreciated?

[sparkprime]

Gravity: 5000m/s/s
Radius: 2.5m
Height: 25m

sqrt(25*2/5000) = 0.1 seconds to hit the ground
0.1 * 5000 = 500m/s when it hits the ground

In order for something going at 500m/s to move in increments of < 2.5m then you have to clock at 500/2.5 = 200hz

[dangerdaveCS]

200Hz doesn't work, most capsules fall through and the ones that don't 'jitter'. Up to 700Hz the same problem occurs (though it is less pronounced). At 1000Hz there is no problem, but it runs really slow.

[sparkprime]

If it doesn't work then things aren't working out the way you described.