I started writing this "History Of SQ" a few years ago for a website I was working on, but various things got me side-tracked. I've decided to start working on it again - but think I'll just 'publish' it here on QQ (if that's OK) as I write each section. This first post is what I've already written but I might go back at some point and re-write this. So, here it is - if anyone has any additional info or differing info, or questions, please don't hesitate to post or question what I've written because it should be obvious, I love discussing and debating this stuff.
I still have to upload the figures to Photobucket so I can place them in the article and correct the formatting - it was in Word originally. So, please bear with me as I put it together.
BTW, all my information comes from technical documents, press statements/releases, magazine articles and interviews as well as personal correspondence with some of those involved with the various systems.
____
Quadraphonic Beginnings
Although multichannel sound got its public start in the 1930’s, with Bell Lab’s 3-channel experiments and Walt Disney’s experimental Fantasound, it didn’t become practical until the advent of magentic tape and widescreen motion picture systems in the 1950’s. The term “stereo” itself means “solid” and makes no explicit reference to the number of channels present, although in modern usage most people understand it to mean 2 channel. For the home listener, the original stereo recordings of the 50’s were typically three tracks, Left, Center, Right, and it was only due to the LP’s inability to hold more than two tracks that limited stereo to two channels - and the disc format was an absolute requirement from a cost standpoint. The form multichannel stereo took in the 1970’s, that of Quadraphonic 4-Channel, got its start in 1968/69 when Vanguard Records publicly demonstrated some experimental four-channel ‘surround sound’ recordings on discrete 4-channel open-reel tape. Utilizing an extra pair of stereo speakers placed in the back corners of the room, Vanguard demonstrated both ambience-type recordings as well as ‘full-surround’ with instruments placed behind the listener. This setup, which became typical of Quad, was not based on any psychoacoustic research or scientific studies into how we hear directionality from behind and to the sides; no, the 2/2 front-back layout was used simply because it seemed the easiest to achieve in the home and was quick to implement. The press reports of this demonstration set off a huge amount of interest from both audiophiles and record companies and discrete-four-channel open reel tapes were soon offered to the public. Surprisingly, RCA Records, who was not known for being an innovator in the audio industry, quickly introduced the discrete Q8 cartridge format. It was based on the standard 8-track tape but with two of the four "programs" used for the back pair of channels - thus, playing time was cut in half unless double length tape was used. The Q8 format had severe drawbacks however; while it was a discrete format and did replicate the orignal four-channel performance, its fidelity aspects - in frequency response, dynamic range, signal-to-noise ratio, distortion and wow and flutter, were very poor. The Dolby B-Type noise reduction system, which had recently been introduced to consumers, was used to try and help lessen noise somewhat, but strangely, it was only applied to the front channels. Perhaps the biggest drawback of the Q8 cartridge was its incompatability with existing 8-track units; a listener with a standard 2-channel 8-Track player would hear only two of the channels from a Q8 cartridge. A discrete 4-4-4 'compatibility matrix' could have easily been used to make the Q8 format 100% stereo (and mono) compatible, but no one seems to have done this at the time. Because the Long Play record was the highest fidelity mass-produced consumer audio format then available, the record companies quickly looked for methods of delivering four channels of sound from the standard LP. At this same time, magazine articles started appearing about the so-called “Hafler” method of surround sound reproduction. By wiring a second set of speakers in a specific way with the main speakers, any out-of-phase, reverberant, information contained naturally in most two-channel recordings could be reproduced at the back of the room – giving an open, spacious surround-sound effect. Some recordings, especially those made with the Blumelin M-S microphone technique, naturally captured a large amount of reverberant information that could be beautifully reproduced from the Hafler speaker array. The Hafler array could be fatiguing to listen to in the long run though due to the misaligned phase relationships between the front and back speakers. However, the seed had been planted and this started engineers and inventors down the path of what would soon come to be known as Matrixing.
The first inventor to take advantage of this new field of signal-conversion theory was Peter Scheiber, and he filed the earliest patents for matrix encoding/decoding methods in 1970. His patents contained something truly new however - an extra circuit that monitored channel dominance and altered the output amplifier gains to enhance that directionality of the dominant signal. His first patent was for what is now known as the ‘diamond’ layout and Dolby's MP Matrix (Dolby Surround) was originally licensed under this patent.
(As a side-note, all the original Scheiber/CBS/Sansui/Tate patents are now expired) In Scheiber's diamond layout, speakers were placed to the right and left of the listener and in front of and behind, making the speaker assignment Left Side, Center-Front, Right Side, Center-Back. This was strictly a polarity matrix, with no phase encoding being used; Left and Right were the normal stereo channels, while Center was the sum of left and right and Center-Back was the reversed polarity difference between left and right. Channel separation was quite poor – Left/Right and Center/Surround both offered infinite separation, but between L/C/R or L/S/R, separation was only 3db, which is akin to no separation at all. As mentioned, Scheiber used a gain-riding enhancement system that could transfer output power between diagonal pairs of speakers - in other words, it increased signal levels in the wanted channel by +3db while, at the same time, decreasing levels in the unwanted channels - by using a Log/Ratio detector he hoped to measure the separation needed in actual db and thereby alter gain to maintain total output power in the room so that a listener would never hear the individual channels level ramping up and down - this was never achieved. A short time later he filed a patent with the more standard ‘square’ speaker layout – this became known as the Scheiber “Square Resistive Matrix” and like the Diamond Matrix, no phase encoding was used, only channel blending, reverse polarity signals and amplitude (level) differences. Sansui's QS Matrix was a variation of the Square Resistive Matrix and Sansui was the first company to enter into a matrix encoding/decoding license agreement with Peter Scheiber. CBS licensing of Scheiber's patents soon followed, although CBS felt that their SQ matrix was both distinct from Scheiber's and non-obvious; the patent office felt otherwise, saying that Scheiber's original engineering work and patent claims were so good that they encompassed virtually every modification of the technology that could be thought of. A lot of people would be owing Peter Scheiber a lot of money in licensing fees!
(for various reasons however, that never happened)
From Scheiber's original idea and early work sprang the Quadraphonic matrix systems of the 70’s and the Surround Sound motion picture & music systems of today. The most notable systems were: QS of Sansui, RM-Regular Matrix, Electro-Voice EV-4, Hafler DynaQuad, BBC Matrix-H, Michael Gerzon's Matrix-45J, Denon’s UMX, NRDC’s UHJ, SRS Labs Circle Surround, Dolby Lab’s MP-Matrix, Lexicon Logic-7, Shure Stereosurround and many, many others. But the most successful, compatible and well-known four-channel matrix system of the 1970’s was CBS Labs SQ Matrix, which stood for Stereo-Quadraphonic.
CBS Labs – 4-Channel Sound From A 2-Channel Record.
In 1969, Columbia Records was one of America’s largest record companies and their research arm, CBS Labs, one of the worlds preeminent electronics laboratories . From CBS Labs came the 33.3 Long Playing (LP) Microgroove record, the Field-Sequential Color Television System, EVR-Electronic Video Recording, CX Noise Reduction, FMX and numerous other consumer products. Once Columbia was aware of the Vanguard quadraphonic demonstrations, they knew they had to have their own quad system for LP. Not just for use by their own artists, but something they could license to bring a continuing royalty. In the early 70’s, unlike today, most American entertainment companies had their own research divisions to create new products, so instead of looking at outside quadraphonic systems, Columbia directed their top audio researcher, Dr. Benjamin B. Bauer, to take on the task of creating a four-channel LP system.
Ben Bauer had been interested in recording more than two channels of sound on a standard LP for some time and his first work was directed at the so-called “carrier” type disc. The back channels of the quadraphonic recording could be modulated on pair of high-frequency AM carriers. But while this might provide the four separate channels desired, CBS felt it would involve extensive modifications of the entire LP recording/pressing process and demand expensive, precision playback equipment in a consumers home. Worse, since the modulation itself would take up physical space on the record, playing time and fidelity would be reduced. Finally, a discrete carrier-type LP system was incompatible with FM broadcasting techniques. Columbia would only sponsor a quad system that was fully compatible with all existing playback formats, was low in cost and didn’t complicate record production, so, the discrete-carrier quadraphonic LP was quickly set aside in favor of the new, and cutting-edge, "matrix" theories of audio signal conversion.
Before beginning work on the matrix systems, Ben Bauer and his co-inventors, Dan Gravereaux and Authur Gust, came up with a list of performance characteristics they and CBS felt a successful quadraphonic system should include:
1. High fidelity – the system should not lower fidelity or decrease playing time. Full LP performance standards must be maintained.
2. Compatibility – the quadraphonic recording must be playable in mono or stereo, with full and complete reproduction of all sounds encoded on the record. In stereo or mono playback, the consumer should not be aware that he is listening to anything but a normal recording: this would lower costs by facilitating single-inventory release.
3. Full left-to-right separation must be maintained. Both in stereo and quadraphonic playback.
4. Quadraphonic accuracy: The system should provide full and accurate 360° directionality with all sounds reproduced at the same level and phase and from the same direction they were originally recorded, even if the localization was somewhat blurred and 'close' to the listener.
5. Simple or advanced decoding – low-cost decoders should deliver correct directionality (even if somewhat diffuse) while more expensive decoders should deliver a precise representation of the master tape. Thus, the signal encoding should be as such to provide the maximum amount of information about the soundfield and prevent phase or amplitude relationships from occuring that would confuse an advanced decoder.
Ben Bauer’s first matrix system was the diamond layout, essentially the same as Peter Scheiber’s first patent. An attempt was made to provide Center-Back transmission in mono reproduction by applying large phase shifts (+ or - 135°) to the Cb component in the LT/RT channels and a more sophisticated form of logic control was patented with it. However, the diamond matrix was never seriously tested and was quickly abandoned because it did not provide the ‘square’ of speakers required for a commercial four-channel system. Ben Bauer also felt that listeners would be uncomfortable with sounds coming from directly behind them.
New Orleans Code #1 & #2
The so-called New Orleans Code, named because Ben Bauer developed it while visiting New Orleans, was attempt to provide a matrix capable of monophonic transmission and correct decoding of Center Back signals with full stereo compatibility and directional fidelity. It used Scheiber’s ‘Square-Resistive’ encoding equations but augmented them with quadrature (90°) phase shifts between the signal components. The encoding equations were:
LT = 0.92Lf + 0.38ej90°Rf + 0.92Lb + 0.38e-j90°Rb.
RT = 0.38e-j90°Lf + 0.92Rf + 0.38ej90°Lb + 0.92Rb
This was New Orleans Code #1 and had the desired features of monophonic compatability and 360° directional fidelity. Since the signal components in LT and RT were in quadrature and Cf and Cb had 45° phase angles, Lf, Rf, Rb and Lb were all transmitted at the same level in mono or stereo; Center Left and Center Right (Cl-Cr) were increased 2.8db in mono, while Cf and Cb were augmented by 3db in mono; thus in terms of mono compatibility, the New Orleans Code #1 was satisfactory. When decoded via a simple matrix, separation between channels was 3db with diagonally opposite channels producing no output. If signal-adaptive (Logic) decoding were implemented, the New Orleans Code would require side-to-side and front-to-back enhancement. Side-to-side enhancement is undesirable since we are most sensitive to stereo image shifts when they occur from left to right; we are less sensitive, by a factor of 10, to front-to-back shifts. Therefore, any logic action should be on the front-to-back axis.
In Bauer’s early work with matrix coding, the New Orleans Code #1 was exhaustively tested and given serious consideration due to some of its desirable characteristics. While its front channel separation was only -7.7db in stereo playback, the corresponding LT and RT signals were in quadrature which made the limited separation less audible than if they had an in-phase relationship. However, the Cf signal components were at a 45° phase angle which caused spreading of the Center-Front image. That problem was remedied with New Orleans Code #2.
Bauer and his co-workers felt they were on to something good with the New Orleans code, so in an effort to fix the problems of the #1 Code, a second New Orleans Code (#2) was devised and tested. The #2 encoding equations were:
LT = 0.92e-j22.5°Lf + 0.38ej67.5°Rf + 0.92e-j22.5°Lb + 0.38e-112.5°Rb
RT = 0.38e-j67.5°Lf +0.92ej22.5°Rf + 0.38ej112.5°Lb + 0.92ej22.5°Rb
The new #2 code had the same –7.7db seperation in stereo as Code #1, but undesirably, this was more audible due to the fact that LT and RT had a 45° phase relationship to each other. Cf imaging was improved with a 0° phase shift, while Cb was at 90° and, in mono, reduced by only –1.7db. Unfortunately the improved Center-Front phantom imaging came at a price: Lb and Rb were reduced by -4.8db in mono. As a result, New Orleans Code #2 was judged different, but no better than, Code #1.
The combined defects of poor Center-Front imaging, limited stereo compatability, reduced rear channel transmission in mono and undesirable adjacent channel crosstalk caused Ben Bauer to set aside the New Orleans codes in favor of the Stereo-Quadraphonic (SQ) code.
The SQ 4:2:4 Stereo-Quadraphonic Matrix
The SQ Matrix came about by abiding strictly to the list of desired features for a compatible matrix quadraphonic system. Because the front channels were obviously the most important (assuming a forward-facing listener) and would be required to present the main stage in stereo playback, it was decided to encode all front sounds in precisely the same way as standard stereo with no phase shifts or cross-blending between Lf and Rf. Center-Front would then image sharply in stereo or quadraphonic playback and increase by +3db in mono, which again, is exactly the same as standard stereo. The back channels would be encoded with broadband 90° phase shifts. Thus, the CBS SQ Code used Left-to-Right amplitude differences for encoding the front channels and used the phase-space between Left and Right to encode the back channels; hence, the name Stereo-Quadraphonic, or SQ. The resulting SQ 4/2, also known as “Basic SQ,” encoding equations were:
LT = Lf + 0.71Lb + 0.71e-j90°Rb
RT = Rf + 0.71e-j90°Lb + 0.71Rb
Ben Bauer depicted SQ encoding and decoding by using “phasors”, or vectors (arrows), which show both the amplitude and the phase of the coded signals. For a reference, or 0° phase, a vector will point to the right and move counterclockwise for leading phase – i.e., a phasor pointing up will have a leading 90° phase shift relative to one pointing to the right. A –3db level reduction is seen as a phasor that is 75% as long when compared to a full-amplitude phasor.
As can be seen from the Basic SQ phasors shown in Fig. 1, the (a) Left Total (LT) channel contains:
1. The Left Front signal recorded at reference level (1.0) and 0° reference phase.
2. The Left Back signal recorded at –3db relative level (.707) and -90° relative phase.
3. The Right Back signal recorded at -3db relative level and 0° relative phase.
The (b) Right Total (RT) channel contains:
1. The Right Front signal recorded at reference level and 0° reference phase.
2. The Right Back signal at –3db reference level and +90° relative phase.
3. The Left Back signal recorded at –3db reference level and 180° relative phase.
Notice that Left Front appears only in Left Total and Right Front appears only in Right Total. Left Back and Right Back appear in both LT and RT at equal levels with a –3db reduction but in leading or lagging quadrature (90°) relationships to each other. This simple phase-only encoding ensured the 360° directional fidelity that Ben Bauer desired and was the foundation of the SQ system with the first demonstration records and commercial releases encoded using these equations. If reproduced in stereo, the SQ-to-Stereo channel-fold would appear (image) to the listener like this:
Left Front, Center-Front and Right Front appear in exactly the same positions as they would during quadraphonic playback – and, they are precisely in-phase; any sounds panned between them will stay in phase and, as a result, image clearly; no other matrix quadraphonic system shared this attribute. Left Back and Right Back image near the center line of the stereo array, but because they share leading and lagging quadrature relationships, have an audible sense of separation and appear displaced slightly towards the leading channel. The minor spreading of their respective images suggests distance, helping the stereo listener to differentiate the front channel sounds from the back channels. Center Back is 180° out-of-phase and thus, has no definite image, making it appear separate from the front-channel sounds; it dissapears completely in mono. Center Left and Center Right are not components of the Basic SQ modulations and will be dealt with later in this article. Notice, however, that a full 75% of the phase-space between the Left and Right channels has been given to the side (Cl-Cr) and back (Lb-Rb) channels – this mimicks the real life distribution of ambience in a concert hall, where the total sound energy typically consists of up to 75% reverberant sounds. Tests by the National Quadraphonic Radio Committtee found that SQ had the best quad-to-stereo fold of any matrix or discrete system tested and thus, the best compatability.
SQ Matrix Decoding
The first SQ decoders were straightforward inversions of the encoding process – a simple SQ decoder block diagram (Fig.3) looks like this.
CBS encoders and prototype consumer decoders used precision aligned 10-Pole phase shift networks that were accurate +1° over a 20-20kHz bandwidth. This was required because accurate SQ encoding and decoding can only occur over the bandwidth of the phase shifting. Such precise performance was not to be found in consumer decoding equipment – which will be dealt with later. The results of a basic SQ decoder, again displayed as phasors, are shown in Fig. 4.
As can be seen, each of the resulting four signals contains a mixture of three of the original four signals, with one being dominant. In both the front and back pairs of channels, Left and Right are completely separated from each other. Offsetting this excellent side-to-side separation is the front-to-back contamination of the originally discrete signal components, and is quite severe. With such signal contamination, or crosstalk, a basic SQ decoder gives the separation shown in Fig.5.
The 20db noted between the Left and Right pairs is actually infinite, but was limited by the LP/cartridge combination, which was typically 20db. The worst seperation occurs between Center-Front and Center-Back, where it is effectively 0db; center placed vocalists will appear to come from overhead in the center of the room. Nevertheless, CBS discovered that if a lister was positioned in exactly the center of the speaker array, they quickly learned to properly localise the various sounds. In addition, Ben Bauer felt that most quadraphonic recordings would be of the “ambience” type where such reduced separation was not of much consiquence – at least in classical recordings. In CBS’ experiments with matrix decoding, many listeners expressed the feeling that the basic SQ matrix was successful in retaining directional fidelity but with an apparent ‘shrinking’ of the room towards the listener.
To Market: Introducing The CBS SQ Quadraphonic System
Ben Bauer, CBS Labs and Columbia Records officially introduced the SQ Matrix system on June 10, 1971 in Montreux, Switzerland. At the same time, external SQ decoders, from CBS’ equipment division Masterworks, Sony and others, were released to the public as well as 100 SQ encoded albums. The first AES Paper, titled “A Compatible Stereo-Quadraphonic (SQ) Record System” appeared in the September, 1971 issue of the Journal of the AES. Ad’s from CBS trumpeted the system as the “perfect” matrix, needing no improvement – they claimed “full” separation, without noting they meant only side-to-side separation. Causing further confusion, instead of talking about the system in terms of actual signal components, they referred to SQ as encoding horizontal, vertical and two kinds of helical modulations on an LP record, giving the impression that discrete signals were present and could be recovered from an SQ disc.
Almost immediately, complaints from customers and reviewers started coming in – there wasn’t much surround-sound effect, in fact, the system sounded just like double-stereo. The decoded crosstalk signals and their various phase shifts coming from every speaker also caused severe listening fatigue. These defects had been noticed during development, and even before the public launch Ben Bauer had decided that perhaps some of the side-to-side separation, which had at first seemed so important, might be sacrificed slightly in favor of more front-to-back separation. The solution came from decoder work done on the New Orleans code – a slight crossblend between output channels in the SQ decoder. Left and Right Front would be blended by 10%, reducing the technically infinite separation to no more than 20db (which wasn’t any kind of drawback,); the back L/R channels would be blended together by 40%, again reducing their separation from infinite to a mere 8db. This decoding modification increased Center-Front to Center-Back separation from 0db to 6db. While not impressive, it did at least provide for a slightly more audible surround-sound effect. The 10-40 blend, as it came to be called, quickly replaced the straight decoding in low priced SQ decoders. In fact, very, very few SQ decoders ever used the basic matrix with no cross-channel blending.
Even with the 10-40 Blend, because of poorly engineered and spec’d equipment, consumers were still not experiencing good SQ decoding results. In the early 1970’s, the so-called “consumer movement” had not yet started and most consumer electronics manufacturers didn’t feel a need to make particularly accurate decoders. 10% tolerance resistors and sloppy 3-Pole phase shifters lead to licensed decoders with 20° phase errors, 5db channel imbalances and appalling frequency response, all of which caused havoc with the precisely encoded SQ records and their necessity for accurate decoding. For unknown reasons, CBS allowed 20° phase errors in licensed decoders and only called for a minimum of 2-pole phase shifting. If a licensee asked for technical help, CBS was always there to provide it, but the final design, specs and performance of the SQ decoder was left up to the individual manufacturer. This lead a small American company called Audionics of Oregon (in partnership with TATE Audio, Ltd.) to create an “audiophile” SQ decoder using accurate 6-pole phase shifters, which produced improved decoding quality.
Another difficulty with the decoders was a lack of any sort of input balance/level control to compensate for misaligned phono cartridges. CBS never even mentioned that it was a problem, stating that ANY stereo phonograph system was good enough to playback SQ encoded albums. Due to the inherent limitations of the LP system in maintaining phase and amplitude alignment, precise stylus and tonearm calibration were required – no public discussion of this ever took place, not even in consumer publications like Stereo Review or High-Fidelity.
