I would like to propose a simple experiment.
Two set-ups, one with four sub-woofers on center points around a room, another with three sub-woofers along the front wall on center and thirds
For the 4 driver setup
phase align however you want.
For the three driver setup
If the AVR has mono-LFE, side drivers phase delayed ~15 degrees.
If the AVR has dual LFE outputs then set the center sub to out 1 and the sides to out 2, set the delay to the side drivers -1ms.
Listen to whatever you want
Hi Jason, thanks again for your thoughts!
I have thought about these issues as well. My stated priority is to try to reduce what can be large variations in seat to seat response. For these, the steady state (pink noise as you call it) response corresponds well to what we hear, in my experience. And, this is the "500 pound gorilla" in the room, as opposed to more subtle phase coherency issues. I'm not convinced that such a "coherent" wavefront ever exists, especially if there are multiple seats where phase between different sources won't add up the same anyway. In my full paper, there is also another metric called Variance of Spatial Average, with all the same analysis done for it as was done for the MSV and MOLn metrics. The VSA metric is all about flatness of response as opposed to seat to seat variance. I omitted it from the Audioholics article, since it was more of an afterthought anyway (the seat to seat variance always comes first for me). Anyway, if one wanted to try to maxmize "phase coherency", perhaps one would want to include the VSA metric to try to get flatter overall responses. It would complicate matters though. Aside: I may have forgotten to include it in the Audioholics article, but I am assuming that someone serious about optimizing their subwoofers would also be using at least a couple of bands of parametric eq to clean up any out of control resonances, and tighten up the sound. Yes, I believe in deft use of minimum phase correction filters.
What do you mean by "turbulent sound field"?
As for using steady state transfer functions, yes it is true that for example, a narrow resonance that looks bad in the SS response is not that audible with normal music program. I think that in the statistical analysis used in my modelling, those types of narrow band issues do not have much influence. In any case if not a SS analysis, what do you propose instead? Anything more complicated may tend to have too many degrees of freedom to be useful.
All this aside, I really like your suggestion to do a test. In order to make it relevant to the conditions I am assuming, one would need to do multiple seats though, and include a couple of bands of parametric eq for either option. Also, your proposed settings for the frontal sub setup begs a question. Would your 15 degree phase adjustment create "phase coherency", since that phase delay results in different time delays at different frequencies? Anyway, back to your test. To get definitive results, one would need to get at least several "subjects", double blind listening, multiple different programs, and other methods to control biases (this is one of the things I have been doing at Harman for the last 15 years). I could suggest a simpler test. You set up the two sub configurations in a room. Measured each sub to each of, say, 6 seats. Send me the measurements (impulse responses please). And I will calculate the results from your configuration and for the 4 sub setup (use corners please for this one). For the 4 sub setup, I'll use Sound Field Management to calculate the delays. Then i will apply some basic parametric eq to mine (and yours if you want). Then we just look at the calculated impulse responses to see if there is a clear difference in coherence, as judged by the width of the impulse response. Note that this will not be a simulation, but will represent the simple summation of the individual impulse responses, with appropriate delays and global eq. This will be exactly what you would measure in the room, since it is a linear system. What do you think?
