On Sun, May 01, 2011 at 09:39:55PM +0200, Jörn Nettingsmeier wrote: > paul's first order recordings are lovely, but kind of easy on the > format, since they have a frontal soundstage for the most part, which > you "tune in" to, and simply disbelieve any spuriousness from the rear. > but take the funky recording of the VoiCE trio where the singers > surround the microphone - with a horizontal square of speakers, i easily > lose track of single voices every once in a while. it gets a lot worse > in a cube.
That is indeed a lovely recording, and I can only confirm your impression about it when reproduced in 3D. Simple fact is that periphonic 1st order has an rE of around 0.6, while simple stereo (60 degrees angle between the speakers) has 0.87 for a center image (worst case). To get up to that value for 3D Ambisonics you need 3rd order. And don't underestimate the rE metric - it's not at all specific to Ambisonics but just a measure of how 'concentrated' in the source direction the directional information is, and consequently how stable imaging will be if you move away from the sweet spot. > i guess a better question is: at what order do ambisonic systems stop > falling flat on their faces with "hostile" content? That is indeed the right question. And the answer seems to be that starting at 3rd order things seem to work. > today, you have a director or composer, he demands something, you've got > to deliver. s/he certainly doesn't want to discuss how what you're > failing to deliver is still "good enough". Exactly. With all due respect, I can't help but feeling that at least some of the 'founders' generation of Ambisionics practicioners fail to understand what higher order is about. It's *not* about 'more precise' localisation. IMHO there are two aspects that set HOA apart from POA: * It extends the listening area, and it's not difficult to see why it does. Higher order decoding will concentrate the signal in the speakers close to the source direction, and have much lower levels in the others compared to 1st order. In the limit it approaches pair-wise or triple-wise (AKA VBAP) panning, except that it won't provide preferential treatment for directions corresponding to a speaker. A more abstract view of the same is that HOA considers the radial dimension as well as the angular one - it extends the range over which the Fourier-Bessel expansion is valid. This is an aspect that has been and still is ignored in many texts about Ambisonics, in particular those dealing with first order only. * It resolves ambiguities inherent in first order encoding. Take for example the Gregorian choir I mentioned in a previous post. The first order encoding of such a source distribution is highly ambiguous: the same relations between W,X,Y,Z could be produced by many and very different source distributions. Which is why I don't believe that systems like Harpex will handle it correctly. It's actually a tribute to the psycho-acoustic qualities of first order that it can deal with this quite well - it allows the hard work to be done by the listener's brain. Dealing with that in any algorithmic way will require something more sophisticated than separating the field into a sum of two plane waves. Ciao, -- FA _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
