On 2012-07-23, Richard wrote:
OK, so, i've been using software to decode UHJ, but can anyone tell me what i need to do to re-create the 'super-stereo' mode on the hardware decoders?
Isn't that supposed to be just that you BHJ decode a stereo recording which wasn't originally BHJ encoded, for two frontal speakers instead of the minimum of four speakers needed for full BHJ pantophony? I.e. the dual of listening to a BHJ encoded recording via a discrete, pass-through stereo setup? The existing controls on a normal domestic ambisonic decoder (mostly "width" which is for setting the aspect ratio of the quad geometry), and especially the fully array of controls on a professional ones (including "dominance"/"zoom" and "forward preference") help optimize the experience if you have a full four speaker geometry around you, but even if you have only two, the decoder process can be engaged and it will lead to more "envelopment"; but then it needs some extra processing to fold the soundfield onto just two frontal speakers, which is engaged by the "superstereo switch". (The tech is the same which underlies both Gerzon's work with compatible HDTV systems, to be distinguished from the Vienna decoder theory, and then Trifield, whose details and patents should be attributed to Geoffrey Barton.)
If I don't remember it wrong altogether, the separate kind of width control for the super-stereo effect alone just attenuated the delta ("side") channel within a BHJ decoder which was formulated in the delta/sigma (mid/side; no connection with d-s-digital converters) framework, like most of the early analog pieces like the Integrex work were. So, Richard, if I remember half right, the most well-known ambisonic super-stereo implementation was essentially just a BHJ-decoder, combined with a simple mid/side pan pot for controlling the width of the spreading introduced by the decoder; even if other, more complicated circuit topologies and perceptual effects have been referred to by the same term.
No guarantees here, though, because I haven't kept refreshing my ambisonic know(how) of late like I should have. :)
I know it's a mixture of two of the three signal, possibly W & X, but at what values, etc.
It's a mixture of all of them, because you can never extract but two of them from a stereo L/R transmission signal, you always need some components of all three to get the directions at least half-right, and so whatever you end up with is a mixture of the three at best. You just have to design for and/or hope that the mixture is something sensible.
UHJ, and as its special case BHJ for two channels, aimed from the start at discrete stereo compatibility. The analysis which lead to it not only looked at the primary use case where something was first BHJ encoded and then symmetrically BHJ decoded. It also took note of other standards and use scenarios, like a BHJ encoded transmission (from the BBC...) which somebody listened to without decoding it. So for that kind of asymmetrical use it already works by design.
The symmetrical mode of analysis used to derive the system happily ended up with working the other way around, too: the system *does* work pretty well with many different kinds of matrix encodings with an UHJ decoder used only at the receiving end. At least cosine panpotted discrete stereo was analysed even in the earliest papers, over the Scheiber sphere, and then empirically validated. So were the quad systems of the time which were mostly found to be compatible, either by design or by convergence in stereo compatibility. In fact in the early BBC empirical work where Matrix H (pretty much the H in UHJ) was found to be the best, they tested whether an active version of H might do some good; they found it could do good for quad-like material, while at the same time the passive, linear version of H retained envelopment better. Which leads me to my next point... ;)
The one fun theoretical prediction from that time, coming via technological heritage, which I've never had the chance of hearing via my own ears, is that I'd actually expect even early hardware BHJ to sound better as a Dolby Surround decoder than Dolby's own hardware ever did before Pro Logic II. Why?
Because Dolby Surround encoded the stuff just dandy wrt the analysis underlying BHJ, only with a less refined encoding locus. Once you got the frontal image right, the mathematics pretty much guaranteed that the signal would decode semi-right over BHJ, both as pantophony and as simply-frontal superstereo.
At the same time their decoder relied on a simple subtraction of the channels to derive the back surround channel, which they fed back in-phase to the back and then all-over via the following Pro Logic active active matrix. That is just fucking wrong if you calculate what the Gerzon theory tells you about the scheme. (That's a straight forward matrix calculation my TI-86 can do. Try it out; no math needed, just a bit of matrix algebra with complex coefficients.) The real quantities work out pretty well, but the complex, phase quantities are totally fucked up. Which I think is pretty much the explanation why Dolby Surround/MP sounds so dull and unenveloping and badly imaging, especially to the back.
They then added all kinds of weird shit to the back channel for perhaps a decade, not to mention the multiple incarnations of the active matrix. Resulting with such things as the THX recommendation that side back channels should be of the dipole, dispersive type with no direct sound produced. (Usable in some regards, as in DirAC, but theoretically speaking pretty much just ad hoc, and with Dolby's design adding insult to injury.)
Yet the information embedded by the encoder is still there, to be retrieved by a proper decoder; and a BHJ one is much closer to the proper, tightly imaging, optimal decoder even for a Dolby Surround encoder than Dolby's original one, because it spatially decodes the signal as out of phase as it should, and includes phychoacoustical optimization which Dolby never had.
The funkiest thing then is that Dolby Pro Logic II(x) decoders finally got this. Their marketing speech stresses the "revolutionary" "balanced feedback active matrix" "with rapid adaptation". But in actuality, if you take a close look at it, that constitutes a *very* minor design change as opposed to the original Pro Logic decoder. What *really* changed was that at least in IIx they finally got their phase factors (the imaginary, phase part) in line with the ambisonic decoding equations. Because they had to: with 5 or 7 channels some of whom are behind you, you just necessarily bump into the problem which BHJ already solved.
What then finally cracks the whole case into laughter is that part of the patents within Pro Logic II was "music mode" or what was it. Which basically means that the decoder backs towards zero steering. I.e. a passive martix of the H kind, and with pretty similar complex coefficients for each speaker no less. "Because of envelopment and how much more realistic them choppers coming from the back suddenly sounded."
You can't make this shit up. ;) -- Sampo Syreeni, aka decoy - de...@iki.fi, http://decoy.iki.fi/front +358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2 _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
