I've also been curious about Trifield. As I recall it is essentially an equal-energy-vector Mid/Side upmixing scheme derived by Gerzon. The patent should be expired at this point, with parole-board release from commercial proprietary prison following. It shouldn't be very complex, maybe someone will make a plugin available for us to play around with outside of Meridian gear.
[edit- link to Trifield article describing the Meridian implementation, and discussing the Bell labs stereo experiments, Klipsch, Hafler.. http://www.meridian-audio.info/public/trifield%5b2563%5d.pdf] Similarly, I'd like to play around with Logic 7 without being tied to Harmon gear and gear-related specific implementations of it.
I had lots of fun years ago playing around with the two recorded channels feeding more than two speakers thing, using Klipsch/Hafler sum/difference wiring setups. Most often ended up using a pair of difference channel speakers (L-R & R-L) spliced into a standard 2-ch stereo setup bouncing off the walls. Later played with matrix surround modes (Dolby PLII, DTS neo, etc) which sometimes worked very well but often seemed overly tied to the specific implementations, varying from device to device, setup to setup, tweaking for different material. Three identical speakers in front helped a lot for arrangements trying to derive a center, as did using more speakers than ambience channels around the back.
The commercial surround matrices do a pretty good job of extracting ambience from a stereo difference signal, but I suspect something like Trifield would be far better at deriving a pseudo-discrete center channel in comparison to a simple (L+R) mono sum or the similar schemes used by the standard upmix matrices, which sometimes mess with the purity of the primary L/R channels too much.
In messing around with all that stuff, playing with the recording side of things, and playing around with what manipulations might be done in the processing phase in between I've put a lot of thought into the fundamental relationship between recording and playback. Ignoring multi-take/multitracked panorama/mixed material (the majority of modern music) and modern proprietary 3D object-based cinema rendering schemes for material recorded using a single microphone-array (of which taper stuff is subset), consider that there are the following points at which the number of channels can vary:
The number of microphones used
The number of channels recorded
The number of channels used for processing/mixing
The number of channels delivered
The number of speakers used
The Bell Labs guys doing initial stereo research in the 1930s explored various combinations of that in a simplified form:
Number of mics
Number of transmission channels
Number of speakers
[edit- see the article linked above which has some description of the Bell experiments with regards to this]Simple parity is the obvious straightforward solution: 2 mics > 2 channels > 2 speakers (or 3>3>3; 5>5>5; 7>7>7..)
This is standard stereo taping, and also me plugging my 4-8 ch multichannel recorder directly into the 5/7ch analog DVD inputs of a home theater receiver for playback.
But the asymmetric configurations get interesting:
~Recording few microphones, yet playing back through more speakers. ex: 2ch stereo > upmixed multichannel playback via Trifield, matrix surround upmixing, etc.
~Recording more microphones, yet playback using a lesser number of speakers. ex: OCT/OMT/SBD matrix > 2ch stereo.
^
The first is standard 2ch taping played back via a matrix surround mode, Trifield, Hafler, etc. The second is mostly what this thread is concerned with, likely to be the focus of most tapers reading it, and what you were posting about above concerning the suitability of OCT mixed down to 2ch.
~As does an "hourglass shaped" (many channels > fewer channels > many channels) symmetrical arrangement.
^
This one has always interested me in that it fits standard 2-ch stereo distribution, making it universally compatible. Likely not as good as a pure symmetric (many > many > many) arrangement, yet could it be made good enough to be acceptable (to me) if I could find an appropriate way to matrix encode it? I could then mix to a single 2-ch master delivery format regardless of playback. Everybody wins. I've not been able to answer this yet.