Like the first article entitled Benchmark 1. this article seeks to chart the performance of Pachyderm by applying it to an existing benchmark. This time, the second Round Robin will be used.

For more information on this historic test, please see the PTB website at the following address:

PTB Round Robin II

The room used for the benchmark was Elmia Hall, in Sweden. This room was carefully modeled according to the instructions of the PTB. Absorption and Scattering coefficients were provided. Source and Receiver locations were specified. The results from this Round Robin are less frequently used for benchmarks. However, it is my opinion that this benchmark provides an interesting opportunity to understand the software, as long as the results are not taken as a hard benchmark to strive for.

In other words, as long as we are aware of the difficulties, it can't hurt to study it...

It should be noted that PTB 2 was a very controversial event. There was little agreement between the various programs used and the measured results. Numerous theories have been proposed to explain why the results of the benchmark turned out so inconclusive. Still, the PTB Round Robin 2 is yet another testing ground that can be used to fine tune acoustical simulation programs. For more information on why simulations and measurements of this particular room are difficult to reconcile, this article is quite enlightening:

Elmia Hall Private Studies

Applying this benchmark to Pachyderm allows us an ambitious opportunity to improve the software, though it is unlikely that any software can hope to match measured values, and it is not clear whether they ever should match them perfectly, due to the exhibited uncertainties in the measurements.

This simulation was executed with 2 orders of image source, 50000 rays and a ray cut off time of 3 seconds.

For purposes of brevity, the results of simulations from source 1 and receiver 4 are shown below. The rest are available in a spreadsheet for download (see the downloads section of this website):

Directional parameters such as Inter Aural Cross Correlation Coefficient (IACC) or Lateral Fraction (LF) are not included above, because Pachyderm does not implement them yet.

We hope that the results show that the software performs on par with the industry standard as of the time of the PTB Round Robin shown here. It is true that simulation program results do not match the measurement exactly. For this test, the error has been minimized, but it is uncertain whether this benchmark is the proper benchmark to be used as a hard line to tune the algorithm with. Still, it gives some indication that Pachyderm performs similarly to other programs of its kind.

This page will be updated as future versions of the software, or algorithm design scenarios are benchmarked.