Is it just me, is this in slow motion or are things big?

[JVene]

I searched but find little directly on this question....

I started with an example demo (just a static plane for the ground), fashioned about 20 'planks' (rectangular solid shapes), set on their end in a row like a kid setting up dominoes for a "kinetic chain reaction".

That worked as expected.

The models are 10 x 2 x 1 (visual object size)...the btBoxShape is created with 5, 1, 0.5...using a 'debug' view that fits exactly around the visible object (I'm using Ogre).

What strikes me, and seems true in most of the bullet packaged demos, is that these objects are very large. I believe I've read the unit is meters, so my "dominoes" (not quite to the aspect of real dominoes I know)...they're all about 10 meters tall. If that's the case, these are falling at about the right speed...but....

I read that objects should not be made very small, so I'm inquiring....

How is it typically managed that one simulates the sense of scale for small objects, like real dominoes? It seems that 'real' dominoes are too small for the engine to accurately handle well (unless I misunderstand the documentation).

[info128]

I'm fairly new to Bullet, and just starting to experiment. I did a bit of research into this subject, as the documentation recommends objects should not be much smaller than 10cm or so. This presents a problem for what I would assume to be the majority of physics simulations in games, which require small and light objects (say, less than a few centimeters in size). I found a lot of threads in this forum which helped a lot, and the following wiki entry gives the most succinct recommendation:

http://www.bulletphysics.org/mediawiki- ... _The_World

Basically, as far as I understand, it is common to scale your average sized object to around 1m, and then scale the world up. The pool ball example given indicates a scaling factor of 100. It seems that it is then necessary to scale other values including positions, velocities, gravity, impulses, inertias, etc. this keeps all size values within the recommended range of [0.05, 10]. Apparently this is required due to the hard-coded internal margins which are set to 0.04.

Erwin also noted that using a smaller internal timestep will help, stepSimulation(dt,10,1./240.) will step the simulation at 240Hz. He also recommends to generate single-shot multiple contacts for convex-convex, by calling m_collisionConfiguration->setConvexConvexMultipointIterations(3); This is recommended almost a year ago, so it may not be pertinent now. (http://code.google.com/p/bullet/issues/detail?id=45)



Other forum threads on the subject:

http://bulletphysics.org/Bullet/phpBB3/ ... ale#p15526

http://bulletphysics.org/Bullet/phpBB3/ ... ale#p18424

http://www.bulletphysics.org/Bullet/php ... 2421#p9522

http://bulletphysics.org/Bullet/phpBB3/ ... fast+scale

[JVene]

Fantastic response...many thanks!