On 2010-11-05, Franz Zotter wrote:
I don't know how rigorous your "rE should be constant" constraint is.
I don't want to incite another war, but... rE and rV are just metrics evaluated over the full sphere of directions. They could attain pretty much any value given a certain speaker setup and a certain decoding algorithm. Because both the setup and the decoder figure into the determination, they don't really talk about either directly.
So when we have a certain rig, every decoding algorithm will give a different profile of rE and rV over the sphere. We do want to optimize either of them at first. But even optimizing one of them is pretty difficult, because we need to decide what the optimization criterion really is.
In the classical, symmetric case, a square norm optimization pretty much did the trick. In Vienna, we went with judiciously balanced ellipsoidal equivalue surfaces. All of Trinnov, Bruce's tabu, Trifield, Vienna and even MAG's HDTV work seem to have gone with something different entirely (and MAG's work being the only one to consider complex phase).
What precisely is it that we're trying to optimize for? I think MAG really spelled that out loud and clear in the early work, but now I think we need some further guidance from somewhere. Especially given uneven speaker spacings.
-- Sampo Syreeni, aka decoy - [email protected], http://decoy.iki.fi/front +358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2 _______________________________________________ Sursound mailing list [email protected] https://mail.music.vt.edu/mailman/listinfo/sursound
