I second everything that Fons wrote, most especially that "(n)either method can be used 'blindly', in the sense that you just throw a set of speaker coordinates at it and get a guaranteed result"
This has been a very difficult problem for us. For regular arrays it's easy to say that a certain result (group of decoder coefficients) is optimum in some sense. When the array is irregular then you have to make a determination as to which direction is most important for you. Then you can decide on a trade-off between making the performance good in the better directions (directions with the greatest density of loudspeakers) without making it bad in the poor directions (directions with the lowest density of loudspeakers). From one point of view the solution would be to only use regular arrays, and that's fine for research institutions and facilities where the room can be built with the intention of housing the regular array. But for many or most applications there is a real limit to what one can practically do. The challenge is to do as well as possible within the real-world constraints. And that's what Aaron Heller, Richard Lee and I are trying to do. Our toolkit (really, it's Aaron's) will be available at about the time of the Linux Audio Conference and it uses open source tools that operate on a variety of computing platforms. But there's a lot more to be done, particularly in the area of providing indications as to which of several compromised solutions is better. Eric ----- Original Message ---- From: Fons Adriaensen <[email protected]> To: [email protected] Sent: Wed, February 29, 2012 6:28:55 AM Subject: Re: [Sursound] Decoding coefficients for non symmetrical setups On Wed, Feb 29, 2012 at 01:15:40PM +0000, Miguel Negrao wrote: > I’ve been a bit disconnected from the ambisonics world. From my past reading > on >this list it is my understanding that there isn’t an explicit formula for >decoding coefficients for non-symmetrical setups. Are there currently >available >tools to generate decoding coefficients for non-symmetrical setups or an >iterative algorithm that I could implement on my own ? Two possible methods have been published. The first is Franz Zotter's work, using an 'almost regular' virtual layout based on t-designs for which a decoder can be designed in closed form. Then the virtual set of speakers is mapped onto the real one using VBAP. As far as I know there is no 'ready to use' toolset for this method. The second is the subject of BLaH paper #4, and is based on using a non-linear optimisation toolset to find a solution, given the error constraints and weights for a set of performance metrics. I developed more or less the same (even using the same freely available non-linear library), but wasn't allowed to publish the results, so I can confirm this can work rather well. What looks like a practical implementation (by B,L & H) of this method is scheduled for presentation at the next Linux Audio Conference at CCRMA. I don't think either method can be used 'blindly', in the sense that you just throw a set of speaker coordinates at it and get a guaranteed result. You will have to evaluate the results, and you may have to decide on tradeoffs and iterate the procedure using modified parameter sets. At this point you will need some familiarity with the internals in order to be able to proceed. Ciao, -- FA Vor uns liegt ein weites Tal, die Sonne scheint - ein Glitzerstrahl. _______________________________________________ Sursound mailing list [email protected] https://mail.music.vt.edu/mailman/listinfo/sursound _______________________________________________ Sursound mailing list [email protected] https://mail.music.vt.edu/mailman/listinfo/sursound
