Comment of : Control of a number of particles

This was due to the algorithm used to determine the number of superparticles to be injected at each time step. It used to impose to inject at least one particle per triangle per time time step. There were two drawbacks to this method:

• the number of superparticles could not drop below a given value if avParNbPerCell was further reduced (what you noticed)
• the number of superparticles was time step dependent (what you did not).

The advantage was that it maintained a good statitics (but in an uncontrolled way).

This was changed and now the number of superparticle injected per time step can be smaller than 1 (in average, so sometimes 0 sometimes 1,with correct randomisation). So now you can control the total number of superparticles down to any small value (it is your responsibility to ensure good enough statistics). This change is available in the current development version, and will be in next release.

NB: of course in both cases the weight of the superparticles was/is adapted so that the number of physical particles was/is correct!

posted by Jean-Francois Roussel at Jun 30, 2005 3:50 PM