RC24 is out, and with it we have new convergence algorithms. Finally, there is a way to run your simulation without having to worry about whether or not you used enough rays.

There are two different convergence algorithms - Minimum Convergence and Detailed Convergence.

Minimum Convergence: As you know, every 3 seconds, Pachyderm checks the simulation status and provides the user feedback. When you choose minimum convergence, instead of giving you an estimate of the time remaining, it checks up to two receivers for convergence. Convergence, in this case, is defined as a minimum amount of change for the time interval from 0 to 50 ms, from 50 to 80 ms., and from 80 to inf. ms. When the impulse response has been changed less than our pre-defined minimum for a minimum of 10 checks in a row, the simulation is considered converged, and concludes. This should be enough to get a reasonable estimate of most of your acoustical parameter values, often within just a few minutes. It generates this cool feedback display:

Detailed Convergence: Using detailed Convergence, every three seconds, the algorithm will check up to two receivers at every 1 ms. interval for a minimum amount of change. When the impulse response has been changed less than our pre-defined minimum for a minimum of 10 checks in a row, the simulation is considered converged, and concludes. This should be enough to get a very realistic auralization - with the caveat that this simulation may take a while. Try leaving it running overnight. It generates this cool feedback display:

This is very new technology, so if you find any problems or think of something that I need to rethink, please let me know. If it works for you (which it ought to) then I hope you enjoy the peace of mind associated with knowing that the computer has your ray-count covered. After all, it can see many things that we can't. Who better to figure this out?

--Arthur