View Single Post
Old 4 weeks ago
Lives for gear
David Rick's Avatar
Stereo Array Design as an Imperfect Art

The various tools available for designing stereo microphone arrays do indeed produce differing results. That's because the underlying psychoacoustic data are scattered and inconsistent. The results depend on the type of program material used (clicks, tone bursts, instruments, etc.) and the conditions under which those data were acquired. I've excerpted a couple a graphs from Wittek & Theile's 2002 convention paper that illustrate this clearly. The complete paper is well worth reading; you can find it here. Not only are the experimental curves different from one another, but they are all non-linear. The authors point out that the worst-case deviations occur at the outside edges of the stereo image, but that there is much more consistency if only the middle 75% of the image width is considered because the localization curves are much more linear there.

There is also confusion between the ~1 ms ITD found by many researchers to produce full left (right) imaging and the ~ 1.5 ms ICTD assumed in the Neumann and Sengpiel apps. To further confuse matters, charts for omni arrays based on the 1 ms assumption appear elsewhere on the Sengpiel website! I believe the ICTD number to be what's required for playback over typically-arranged loudspeakers, while the ITD limit was derived in headphone testing.

In actual field practice, I find that the Neumann Recording Tools app produces more reliable results for me than the Williams curves. In the years that I used the latter, a second, post audition tweak was almost always required to give good results. The fact that the Neumann app runs on my phone is a big advantage.

I use the Schoeps Image Assistant app when I need to design LCR arrays. Its prediction of the central "flat spot" in the localization curve of an OCT array matches well with my experience. I use it less often in the field, because it requires web access.

The MARRS app, written at the University of Huddersfield, is also available. It is noteworthy in that it allows you to have the mic array at a different height than the ensemble, and to accommodate wide playback baselines. The underlying theory is explained here.

David L. Rick
Seventh String Recording
Attached Files