Is it correct to say that the early ElectroVoice EVX-4 "Stereo 4" decoder did not use phase modulation?
The EVX was a 'real' decoder for the EV-4 matrix, not just a synthesizer for non-encoded recordings, as the QS-1 was. I'm not sure which "version" of the EV-4 matrix the unit was for though - see, at first, ElectroVoice EV-4 encoding/decoding was very similar to the "regular" matrix systems such as QS and RM, but it didn't use any phase-shifting to encode the rear channels - only reverse polarity and L/R blending. So it couldn't encode a "Center Back" signal, only a 270 degree horseshoe. For various reasons, to encode the full 360 compass, quadrature networks (phase shifters) are required, and the 'first' version of EV-4 didn't use them. The quadrature networks used in QS/SQ/UHJ and Dolby Surround all shift the relative phase between the channels by
+90 degrees, which is very different than the phase modulation used in the Sansui QS-1.
Anyway, after a short time on the market, CBS and ElectroVoice joined forces and EV-4 was changed (the sequel) to a phase-matrix that used phase shifters, with points on the Energy Sphere that were very close to CBS SQ's encoding/decoding... it wasn't an exact duplicate, but close enough to work and it could now encode the full 360 compass. Unlike SQ's basic, non-logic matrix, EV-4 (the Sequel) used non-complimentary decoding matrices - in other words, the encoding and decoding points were different from each other - so, instead of the -3db separation between Lf and Rf that QS had when decoded, EV-4 gave -6db L/R front separation but it was reduced to only 1db between the rear channels. (in 4:2:4 matrix decoding, whenever separation is increased between a pair of channels, the separation is decreased between the opposite pair of channels. The maximum 'all around' separation you can
ever achieve without a logic decoder is 4.7db, which Peter Scheiber's 'high-separation' Tetrahedral Matrix did)
CBS SQ put all the separation in the Left/Right axis, in both the front and back channels - so when decoded with a non-logic decoder, Left or Right Back had infinite separation from the opposite back channel, while Left and Right Front had infinite separation between themselves too; but there was only -3db of separation between Left Front and Left Back/Right Back and Right Front and Right Back/Left Back and vise-versa (in SQ all crosstalk between channels was into the two opposing speakers, i.e. Lf into Lb & Rb) Center Front to Center Back had NO separation at all. As everyone on QQ probably well knows, decoded without logic, basic-matrix SQ sounds like the front and back channels are blended completely together - in other words, double stereo, which is basically what it was. So ElectroVoice was attempting to get compatibility with the most popular and widely licensed matrix system, SQ, while giving more of a front-to-back 'surround sound' effect when decoded.
BTW, QS's 'infinite' separation is between diagonals - so instead of infinite Lf-Rf separation as SQ has, QS has infinite Lf-Rb and Rf-Lb separation - the crosstalk from, say Left Front, is decoded into Right Front and Left Back with only -3db separation between them, so without Vario-Matrix, QS sounds almost like mono since it has basically no L/R separation - which makes it very, very difficult to keep the action of the Vario-Matrix below audibility - you can often hear the soundstage shift from left to right, or a CF vocalist pull to either side as a dominant sound comes in from the side. It's one of the reasons Sansui lets the Vario-Matrix pull mono vocals to the center of the room - to try and mask the side-to-side shifting.
For a GREAT introduction to matrixing, two good papers to start with, which should be available for everyone here to read shortly, I hope, are the July/August 1971 paper "
Multichannel Matrix Systems Overview" by John Eargle and another of his papers from December 1972, "
4-2-4 Matrix Systems: Standards, Practice & Interchangeability."