Hey, Lynn! I've been out of touch with everything quad related for quite a while - my sister died, then a month later my mom died - life has been a mess for the past year!
While I haven't talked with Martin Willcocks for a while, he did tell me a lot about the DES - and clarified some of the misconceptions about it that got spread by people like Bill Sommerwerk, such as the Tate only sensing and decoding the 6 cardinal directions. In the final "Fosgate" version of the DES, it analyzed and enhanced 360 degrees and could enhance 3 directions simultaneously. Here's some excerpts from Martin in his emails to me about his design and performance of the Tate DES:
"The DES was, as you rightly perceived, not a variable matrix decoder in the sense of VarioMatrix, CBS Paramatrix or ShadowVector decoders.... I thought out the idea of making the decision of where the signal was to operate a combination of gain changing and blending by using a superposition of the six "blend" or B matrices with the standard fixed matrix decoding, so that the performance could be similar whatever the source direction was. That was basically my inspiration for the idea of a holistic decoder rather than a piecemeal one. Everything would be considered at once and the best modifying matrix for the particular mix of sounds would result. There were several different approaches that I could have chosen, but it seems that I chose the right option - the six directional matrices would be weighted by the detection of sound levels in each of the six principal directions, and added together, along with the standard fixed decoding matrix to enhance the separation of sounds from any one of these directions. For the SQ system, a linear superposition of three of these weighted matrices onto the fixed matrix works quite well, but it actually does not work so well when applied to QS. As a bonus for SQ, it turned out that the separation of simultaneous sources in up to three axes could be enhanced simultaneously - the front-back, left front-right front, and left back-right back axes were all orthogonal on the energy sphere, so a 3-axis control system was feasible, with six direction control signals and six "blend" matrices. Although I added the center-left/center right (side) axes both for normal and SQ position encoding in the patent, we never really tried out whether the three signals together would work to the best effect. It just happens that they do, allowing separation enhancement of about 9dB for two signals or 6.7dB when three signals are acting together. Since the enhancement is instantaneous (1ms), you never hear that reduced separation. Three axis control comes out of the fact that the pan locus of SQ traverses the energy sphere passing through all three geometric axes, front-back, left-right, and top-bottom, which translate to CF-CB, LF-RF and LB-RB axes in terms of sound locations in the listening area of interest. This is not true of the QS matrix, nor of Ambisonics, which have essentially great circle pan loci, with some distortion of the locations between these axes. So only the Tate SQ DES is capable of three-axis control; the best you can get for other systems is 2-axis control. Three-axis control doesn't work with systems where the pan locus is essentially great circle, which includes QS, UHJ, BBC-H etc. You can use two-axis control, but then the variable matrixing systems do something like that anyway. That's why QS with Variomatrix worked reasonably well, but not as well as SQ/DES. At the 1977 Summer CES, referred to below, Lynn Olson was also present, without the Shadow Vector. I only heard the latter when Charles Wood allowed me to audition it in my Los Angeles home for a few weeks, and to try to capture the schematics from the ptototype circuit boards, which Lynn had built. I think I may still have some of that information in my office, but it's years since I last looked at it, and I may even have prints of the NSC Tate chipset schematics as well.
I have tried a two-band quad setup using two Tate DES's and an active crossover filter to separate the highs from the lows. It's only marginally better than the full band decoder, but I think multiple octave band decoders might be an improvement - just very expensive! There really isn't much need to decode stuff below 200Hz and above 3kHz as the directionality is mainly contained in the central region of the spectrum. Psychoacoustics takes care of much of the rest.
Paramatrix used (axis-crossing) phase detectors rather than amplitude detection as in the Tate. The theory is that the Hilbert Transform converts amplitude to phase and vice versa, which allows zero-crossing detectors to provide information about directional content. I never quite bought into that, as the phase conversion really needs more bandwidth. Still, Dan Gravereaux and Jerry Budelman swore by it and wondered why I had used a complicated agc system and amplitude detection. I think the answer must be that it worked better!
The ADC (Automatic Dimension Control) was only used in the Fosgate DES, but was hinted at by one of the CBS engineers and they had something like it in Paramatrix, but they didn't write a patent on it. Mine being amplitude-based was quite different, anyway. As it turned out (see the ADC patent) the result was that in the side quadrant the two active control signals needed to total exactly 1 unit, while in the front and back quadrants it needed to change upwards slightly, up to about 15% higher. In practice the ADC was about halfway in between these ideal states, so could achieve better than 20dB separation around the entire 360 degrees soundstage for a single panned signal while keeping all other signals in their properly encoded locations. The DES used true analog multipliers in its circuitry. The Fosgate/Scheiber units used FETs as gain control elements, which must have involved some compromises. Im not sure about Paramatrix, as I never saw a schematic. I think the Scheiber unit used FETs as well. If I were to redo the Tate these days it would be a DSP, sub-band decoder using my 16-pole digital phase quadrature networks and incorporating volume, L-R and F-B balance, stereo width and the variable rate and dimension controls, and would have all the other bells and whistles that people like you would want. The patents are all public domain now, long since expired. A project like that would take significant funding and a young keen DSP expert audio engineer.
"
I copied/pasted from Martin Willcocks email so spelling errors and such are his - don't blame me!
I'm interested in comments on what he told me.
BTW, Lynn, until your email of a week or so ago about DD+, I hadn't realized the phase problem with DD - I'll fill you in soon on what Roger Dressler told me back in the AOL days (when Dolby was only on AOL) about DD's phase performance when coupling channels.
My best to all of you,
Ty Chamberlain