.

Much to the pleasure of music lovers everywhere, the development of the 4-channel approach to sound reproduction has brought a new dimension to high fidelity.

It makes possible the re-creation of the acoustical conditions of a "live" performance, and brings all the dynamics and presence of music into play in the listener's own home.

Sansui's QS 4-channel system is a most practical way to enjoy such effects. It utilizes conventional 2-channel stereo transmission media (including records, tapes and FM multiplex) and thereby cuts manufacturing costs of both software and hardware.

These media are also playable on conventional stereo (or mono) equipment, making this system compatible - an essential point in the change over from stereo to 4-channel in the software and hardware industries and in the thinking of the consumer.

• Anmerkung : Diese Quadro- zu Stereo-Kompatibilität stimmt nur sehr begrenzt und nur bedingt bezüglich der Hifi- und Stereo-Qualität im Vergleich zu echten Stereoplatten mit der gleichen Aufnahme.

 
And now, thanks to Sansui's QS vario-matrix circuit, the 4-channel reproduction of QS*, SQ** and other matrix-encoded 4-channel media - as well as synthesized 2-channel sources - can be performed with infinite inter-channel separation.

Sansui's QS 4-channel approach has become a worldwide standard. And with Sansui's professional encoder/decoder units (QSE-4 and QSD-4) being used by record manufacturers and FM broadcasting stations more and more frequently, the availability of QS-encoded 4-channel software is growing rapidly.

This booklet is intended to provide all the technical information on would require to understand the QS 4-channel system in its basic and applied concepts. Please let us know if we can provide further details.

*QS is a trademark of Sansui Electric Co., Ltd.
**SQ is a trademark of CBS Inc.
.

The illustration at the left shows how a QS 4-channel record is made and reproduced using the QS regular matrix 4-channel process. The advantages of this system are explained below.

.

• # Live musical performances can be recorded and/or broadcast in 4-channel sound.
• # Multi-track tapes can be encoded.
• jf Discrete 4-channel tapes can be encoded.
• # The 360° sound field can be encoded without loss or deterioration of any part of the original audio information through the use of the + j phase shifters (patents pending).
• # Only one additional unit is necessary to perform the above, and that is the Sansui QSE-4 professional encoder.

.

In conventional stereo playback, all QS-encoded 4-channel records have a wide-dispersion 2-channel reproduction pattern. The stylus-tip movements are as shown in Figure 2.

Since all QS-encoded 4-channel records are cut in the standard 45°/45° groove format on standard lacquer discs - and since they utilize no special high-frequency carrier or other such devices, no special cutting procedure is required.

Such QS-encoded 4-channel records can be reproduced in stereo (and mono) with no loss of sound quality.
.

• Anmerkung : QS-Platten bei Mono oder Stereo ohne Qualitätsverluste ?? Das stimmt nicht. Und sie wußten, daß Hifi nicht bei 13 kHz begrenzt werden darf, weil sich sonst die Phasen verheddern. Somit ist das dreist gelogen.

.
In other words, they have perfect "2-4" compatibility. Such recordings, and original or pre-recorded 4-channel material, can be broadcast over an FM multiplex wave as if they were 2-channel. The QS decoder in the consumer's unit will decode them into 4-channel. (See below.)
.

.

• # QS-encoded 4-channel sources can be reproduced with the same tonal quality and inter-channel separation as 4-channel discrete tapes, thanks to the new Sansui QS vario-matrix circuit.
• # An application of this QS vario-matrix circuit makes possible the 4-channel reproduction of CBS SQ records and SQ-encoded FM broadcasts. This phase-matrix application provides better results, particularly in front/ rear separation, than that achieved by the CBS SQ basic decoder.
• # As another application of the QS vario-matrix circuit, conventional 2-channel sources can be reproduced in 4-channel sound with either "Hall" or "Surround" effects.
• # No special additional equipment (such as a special 4-channel disc cartridge pickup) nor special attention is necessary.
• # The QS 4-channel system can be applied (and is available now) in various circuit construction forms from an inexpensive decoder to a top-quality professional model. Here are a few examples: QS basic decoder; QS decoder with front/rear QS vario-matrix circuit; QS decoder with independently-controllable QS vario-matrix circuit; QS decoder with split-band QS vario-matrix circuit, etc.
• # Thanks to the development of the QS IC "chip," these QS circuits are now simplified and are available at economical cost.

.

Since June 1969 when Sansui introduced the world's first consumer 4-channel equipment at the Consumer Electronics Show in Chicago (CES), we have been continuing development of 4-channel equipment in many forms.

An example of the commitment Sansui has made in this field is the development of the QS vario-matrix circuit. This circuit, which drew astonished praise from the industry when it was introduced at the Second Central AES in Munich in March of 1970, makes it possible to achieve from QS-encoded records the kind of inter-channel separation that was hitherto thought possible only in discrete 4-channel tape recordings.

With these and subsequent developments, Sansui's QS "regular matrix" with QS vario-matrix system has gained and is continuing to reenforce its strong position in the world hi-fi market.

The following technical papers prepared by Sansui have been prepared for and presented to the top industry societies and associations over the last few years.

They contain theoretical and practical concepts of the QS 4-channel system and may be helpful in your understanding of the truly revolutionary aspects of this approach to 4-channel sound.
.

.

.

There are two ways to reproduce multichannel sound by "matrixing". One is a so called 2-2-4 system which "synthesizes" conventional two-channel program sources.

The other is a 4-2-4 system which encodes four-channel program sources into a two-channel format and then decodes them back to four.

This process is visually explained in terms of disc cutting vectors (or stylus motion) in Fig. 1.

• Fig 1-a) illustrates a disc cutting method of an earlier period when only in-phase signals were cut in a disc groove.
• Fig 1-b) illustrates a modern disc cutting technique which mixes in- and out-of-phase signals together to obtain better presence. When these disc records are "synthesized," it is possible to better approximate the original sound field.
• Fig 1-c) shows encoded four-channel disc cuttings which utilizes the out-of- phase area in the vector diagram to cut back channels information. Thus 360° sounds can be cut in ordinary stereo discs and be reproduced as four-channel stereo.

.

There are four basic types of coding system four-channel matrixing, as shown in Fig. 3. All four-channel coding systems announced so far by various manufactures - Scheiber-Audiodata, Feldman-EV, Hafler-Dynaco, J.O. K.E., Columbia (CBS)-Sony (this system will be further discussed later because it is slightly different from and is not fully compatible with other matrix systems) and others - are classified into the four basic types.

Fig 2

Whichever system is standardized (which we believe should be done as soon as possible), it must satisfy certain basic requirements. Mr. P. Scheiber lists these requirements in three categories in his AES article* with which we completely agree.

They are as follow:

.

• a) The ability to record sounds occurring at any point in 360°, and to reproduce each sound from the correct location in playback;
• b) Nondegradation of signai quality, including noise, frequency and nonlinear distortion as consistent with highest standards in the state of the art.

.

.

• a) four-channel compatibility: nonobsolescence of playback equipment, using each standard component and construction wherever possible;
• b) stereo compatibility: the ability to reproduce the four-channel program on all standard two-channel "stereo" equipment, with all sounds in the four-channel program heard in their proper left-right positions;
• c) mono compatibility: monaural playback possible on all standard equipment, without losing, or altering the relative level of, any sound in the four-channel program.

.

.

• a) adaptability to standard practices for software manufacture;
• b) full playing time within a given format, as compared with the equivalent stereo recording;
• c) usable with all major recording media and, preferably, broadcast.

.
* Journal of the Audio Engineering Society, April 1971 "Four Channels and Compatibility" by Mr. P.Scheiber.

Among the requirements Mr. P. Scheiber enumerates above, the special emphasis must be placed on 1) a), which points out the importance of correct localization of original surround sounds in the reproduced sound field.

As will be analyzed hereunder, four-channel matrixing causes a loss or miss localization of information in its encoding / decoding process mainly because of the existence of out-of-phase sound components.

To simplify the explanation using the equations in Fig. 3 a)~d), the vector angle 0l and 02 are set at identical values. (Of course, both 01 and 02 could be independent and be of any angular values. The E-V Decoder has different 01 and 02 angular values.)

.

The equations .... show that there occurs a complete cancellation of out-of-phase components in the left and right back channels, so that all the resultant left and right channel encoder outputs are composed of in-phase sounds only.

In other words, there occurs a loss of information and miss localization of sound sources during the encoding process. This also indicates that it is practically impossible to encode simultanious four-channel signals of identical levels and phases.

Then, equations ... show that the left and right back are in completely reverse phase. This means that any sound sources located in the back in a four-channel program would sound very unnatural and unclear from the lack of directionality and localization of sounds even if the encoding is correctly done.

In the same manner, it is easily proved that the same phenomena also occur in other types of matrixing shown in Fig. 3.

Unfortunately, all the four-channel encoders introduced so far have been incapable of converting true four-channel information to two channels faithfully and reconverting them to four because of the cancellation of certain information in program sources in their encoding processes.

Here, it must be concluded that a simple matrixing system to encode and decode four-channel programs does not seem to overcome this vital defect, and that a new technique must be added to augment it.

Anmerkung : Hier wird ganz deutlich ausgesprochen, daß alle bisherigen Matrix Decoder (die eigenen wie die des Wettbewerbs) Müll waren, es konnte so gar nicht richtig funktionieren, wenn hinten nicht das raus kam, das vorne codiert worden war.

From the very beginning of its research and development work in this area, Sansui tackled these difficulties, and has recently completed, for the first time ever in this field, a revolutionary four-channel encoding/ decoding system which completely does away with them.
.

.

The Sansui Reqular Matrix System completely satisfies the basic requirements previously mentioned. It offers these exclusive features:
.

• a) An ability to encode original 360° sounds without any loss and miss localization of sound sources, and to decode them faithfully in playback as four-channel stereo.
• b) Nondegradation of sound quality by adding two back channels to ordinary stereo. No colouration or artificiality.
• c) Compatibility with stereo (in two-way) and monaural (for practical use) and also with other matrix four-channel systems (except the CBS system as they claim).
• d) Adaptability to present standards of discs, tapes and broadcasts, and thus nonobsolescence of existing hardware and software.
• e) Very light extra-expenses for consumers to quad-ralize their existing equipment and no increase of cost to purchase encoded four-channel program sources.
• f) Very light investment for the recording and broadcasting industries to quadralize the existing systems.

These have become a reality by the adoption of ±90° phase shifters (patents pending) and by setting disc cutting vector angles (0) among four-channels at 22.5° in an ordinary disc groove. Fig. 4 shows a block diagram of the Sansui QS Regular Matrix System.

Unlike conventional four-channel matrixing circuitry, the Sansui QS Encoder phase-shifts left and right back channels by ±90°, instead of using the usual 180° phase inverting method to achieve a reverse-phase (180°) relationship between the left and right back channels.

This puts the four encoded channels in an ideal phase relationship as illustrated in Fig. 6-a). Signals are no longer cancelled in the encoder, and any information from any direction in the original sound field can be faithfully encoded.

On the decoder side, back channels are phase-shifted in a manner contrary to the encoder; namely, left back is shifted by -90° and right back by +90°. Thus the reverse-phase relationship between the back channels is reconverted to an in-phase one (see Fig 6-b).

This function is well explained in terms of vector angles in the disc groove (see Fig. 7-b). Now, the encoder outputs will be:

Formel

The above equations ® prove that no loss of information is caused in the encoding process by the adoption of j phase (±90°) shifters.

As the above equations ... indicate, the four-channel input signals fed to the QS Encoder are fully brought out in their entirety by the QS Decoder through any two-channel media onto the reproduced sound field.
.

In determining the value of the inter-channel blending coefficient for a matrix coding system, careful consideration must be given to the types of program sources available.

Even conventional 2-channel stereo sources are available in variety, including:
.

• 1) those utilizing only two (left and right) mono channels independently (no intermediate phantoms).
• 2) those with sound images localized at three mono channels i.e., left, center and right.
• 3) multiple-track sources with sound images localized at multiple points on the line between left and right channels.
• 4) sound field recordings without any distinctive sound images such as those of a church organ.

Four-channel program sources can also be classified in a similar manner, and this fact must be carefully weighed in determining the blending coefficient. The blending quantity among four channels of information in the encoding process must in no way restrict the flexibility.
.

Fig. 8 (A) illustrates the localization of sound images by the Sansui QS Regular Matrix System. L1 and R1 symbolize the input signals representing sound sources located to the front of the listener, while L2 and R2 symbolize those representing sound sources located to the back of the same listener.

Sound images between L1 and R1 can be localized in exactly the same manner as when recording a conventional 2-channel stereo program source. It is therefore possible to localize as many sound images as desired. The same is also true of the localization of sound images between L2 and R2. For example, suppose, in Fig. 8 (A), a microphone located at point M2 picked up a sound pressure E1.

• Formeln über Formeln, nicht nachvollziehbar

As is formularized above, the most suitable psycho-acoustic sound presure is obtained when <pi=<p2. Fig. 12 on page Tl-11 represents a relative crosstalk characteristics. Localization of surround sound images shown in equations (§) and © may be thus accomplished either by electrically varying the blending coefficient through the manipulation of the panpots of the mixing console, or by encoding with a pre-determined blending coefficient with respect to the "primary sound images"* which should be consistent throughout the "encode-decode" process. Most important, it should be such a value as produces a "zero" output in the diagonally opposite channel. In the Sansui QS Regular Matrix System, this value is set at -^- = 22.5°, rendering the "encode-decode" matrix square, for the reasons to be explained later.

• Das alles muß man nicht mehr verstehen.

.

In a four-channel stereo system, the information contained in each channel must be treated equally. As is clear from Fig. 7-a), this can be accomplished only when the vector angles among the four channels are identical, i.e., when they are all .......

In this condition.
.

• a) the crosstalks among adjacent channels are equally -3dB. Thus, the four channels are reproduced uniformly to obtain a square sound field (see Fig. 10);
• b) equal volume balance is attained among the four channels, so that distinct sound images can be positioned in any direction inside the square sound field (see Fig. 11);
• c) the encoder and decoder can be allowed to blend an identical quantity of information into adjacent channels;
• d) programs encoded by an encoder in which (9=22.5°, can be decoded by a decoder with a different vector angular value without losing much of their four-channel effect.
• e) conversely, a decoder in which #=22.5°, is able to reproduce programs encoded by an encoder with a different vector angular value without losing much of their four-channel effect.

.
To clarify the points made above, Figs. 11,12 and 13 show the sound pressure response in different channels to visualize how the sound images are formulated in reproduction.

Fig. 11 illustrates eight visual patterns of the phase, crosstalk and directional relationship among the channels. As can be seen in these figures, the QS Decoder allows distinct psycho-acoustic images to be formed in the same directions as those in the original sound field.

Fig. 12 showing the sound pressure response of a sound source located in the left front channel, clearly indicates the reproduced sound image is shaped as a symmetrical pattern accurately directed from the crossing of polar coordinates toward the point where the sound source is located.
.

Let us now consider for a minute what would happen if the vector angle 0 were not 22.5°.

First, if it were smaller than 22.5°: As can be seen in Fig. 13-a), the separation between the left front and back, and that between the right front and back, would be impaired. This would result in a vertically short oblong sound field, making it impossible to obtain equal sepatations among the four channels.

If it were larger than 22.5°:
The separation between the left and right front and that between the left and right back would deteriorate, impairing the volume balance among the-channels (see Fig. 13-b).
Thus it is clarified that the value of the vector angle 0 has a serious effect on the directional characteristics of the reproduced sound images and the shape of the reproduced sound field. This is relatively inconsequential in a 2-2-4 conversion, but if the vector angle were different in the encoder and decoder, it is obvious that the original sound field pattern would not be faithfully reproduced.

As outlined in the preceding paragraphs, the Sansui QS Regular Matrix System thus successfully attains the true aims of four-channel reproduction with its versatile compatibility with two-channel stereo and monaural through an encoding/decoding process. We are convinced that, measured by any standards, it is the most advantageous, most authentic coding system that facilitates the original recording, 4-2 conversion, transmission, 2-4 conversion and reproduction processes. The QS Encoder, an essential part of that coding system, thus makes it possible to take maximum advantage of conventional two-channel audio equipment to reproduce the original sound field in the listener's room.

.

Reasons are given why a matrix must be rotationally symmetrical to encode and decode a sound field. Then a new technique is introduced to achieve an exceptional sense of naturalness and greatly enhanced separation between any pair of the four decoded channels. It involves the use of a rotationally symmetrical variable decoding matrix(QS Vario-Matrix) and controlling it with the phases of the two encoded channels.

A few fundamental requirements of a 4-channel matrix are described below.
.

Crosstalk is an unavoidable consequence of any matrixing system of 4-channel stereo whereby four channels of information are matrixed or encoded into two channels, stored in 2-channel media, and then dematrixed or decoded back into four channels.

The question then is how to best distribute such crosstalk. It is mathematically obvious that the maximum separation among the four channels in any 4-2-4 matrix system (such as described above) is OdB, -3dB, -3dB and -oodB.

The question is thus further narrowed down to how best to exploit this mathematical ultimatum.

To locate a real sound image correctly, the ideal distribution of the crosstalk is as illustrated in Fig. 1. In this arrangement, the -3dB crosstalk is allowed in the two adjacent channels, x and y, of the primary channel A. The phantom sound image formed by the crosstalk components will then be located in the primary channel.

All this means that a phantom sound image resulting from crosstalk components coincides (überlagert) with a real sound image only if the speakers reproducing the crosstalk components are placed symmetrically on both sides of the subject speaker. The same symmetry is also required for any phantom sound image located between any pair of speakers.

This is another important requirement. Namely, if identical signals were fed simultaneously to the two input terminals of the encoder which are directional by phase 1 and phase 2, respectively, they must be superposed upon each other inside the encoder, without cancelling each other to any extent, and be encoded into an output signal which is directional exactly halfway between phase 1 and phase 2. This is particularly important to locate a phantom sound image correctly.

Since we are encoding into two channels, the matrix must be constructed so that the two encoder output signals will gradually come to represent a full circle as the difference between phase 1 and phase 2 approaches pi.

Wer hat das verstanden ?

To reconstruct a complete sound field out of four speakers, the above requirements need be fulfilled with respect to signals on a complete circle. It means that a rotationally symmetrical matrix is required for both encoding and decoding.

However, this is much easier said than done. Symmetry and superposition of input signals are realized very well in given directions if the encoding matrix always treats them in phase with one another. Doing so, however, would produce points of discontinuity in the reproduced sound field.

This is a basic dilemma in trying to embody a full 360° directionality in two channels only. The rotational symmetry, input signal superposition and circular continuity are important characteristics of the QS Regular Matrix System, and are also essential foundations for the variable decoding matrix described in the second half of this paper.

A matrix as described below achieves the symmetry, superposition and circular continuity discussed previously.

Fig 2. Basic QS Regular Matrix

Visualize a sound field where microphones Mi~Mn are placed as in Figure 2. M1 is placed at a counterclockwise angle of phase1 from the right, and its output voltage is assumed to be E1. Then, assuming all the succeeding microphones up to the Mnth one are placed in corresponding counterclockwise directions, the optimal encoding matrix is expressed as

Jede Menge Formeln

Such a matrix provides for "panning" a complete circle. With two differently directional microphones feeding signals from the same sound source, their cutting vectors will be superposed.

As the difference in angles (0) grows and approaches pi, the locus of the resultant composite vector will approach a full circle to represent sounds on a full circle. It will be rotationlly symmetrical, and will always be cut in the direction of phase1/2 from the R axis.

In practice, the exponential term j of the term e in equation (1) need not be varied continuously, but can be substituted by several fixed angles. It is for this reason that we have been proposing a matrix which puts the front two channels in phase while shifting the back two channels by ±90° from the front channels.

This matrix, which we term the QS Regular Matrix, is, for encoding purposes, expressed as

Formeln .....

Then, for decoding, the matrix is given by

Formeln .....

where (0) is the counterclockwise angle of the decoder matrix vector from the R axis, F' (0) the decoded output signal to be reproduced in the front, and B' id) the one to be reproduced in the back.

Fig. 3 shows a block diagram of the encoder employing such encoding matrix.
Fig 3. Block Diagram of QS Encoder

Das alles muß man nicht mehr verstehen, es sind ausgetüftelete absolute Krampflösungen mit jeder Menge an Fehlern.

The QS Regular Matrix System is based on the encoding and decoding matrices previously described. Because of their rotational symmetry, superposition and circular continuity properties, the coding system is able to offer these advantages:

1. 1. There is no directional error. As the matrix gives rotational symmetry and circular continuity, it is possible to reproduce sound sources in a full 360° correctly.
2. 2. No loss of the input information occurs. Input signals are superposed inside the encoder and create continuity; no part of the input information is cancelled.
3. 3. As the matrix is rotationally symmetrical and permits superposition of input signals, it is possible to locate a sound image at the dead center of the sound field.
4. 4. Since superposition takes place in a rotationally symmetrical manner, simultaneous multiple input signals are processed normally.
5. 5. It allows rotationally symmetrical coding of input signals, both electrically and acoustically.
6. 6. The front two channels are encoded in phase with each other, lending themselves to 2-channel stereo playback without affecting the sound image positions.
7. 7. The back two channels are encoded out of phase with each other by exactly 180°; in stereo playback they will be located outside of the two speakers or at no definite positions, merely enhancing the stereo effect.
8. 8. For the same reason, even the simplest form of matrix, such as a speaker matrix, will be able to provide sufficient front-back separation.
9. 9. On regular 2-channel stereo discs, reverberation components are often recorded with mutually reverse phases in the left and right channels. This permits the QS decoding matrix to work as a "synthesizer", expanding the two channels to 4-channels.
10. 10. Mixing down to 2-channels by a matrix does not degrade the current standards of hi-fi stereo playback, including those pertaining to frequency response, dynamic range, distortion, etc.
11. 11. It affords excellent compatibility with 2-channel stereo playback. Because of its symmetrical and continuous properties, the left and right speaker positions in 2-channel stereo playback may be merely regarded as two points to the straight left and right of the listener in 4-channel stereo playback. It is possible to locate sounds anywhere between these two points, so there is no problem of compatibility with stereo playback.

.

It is a mathematical fact that the maximum possible inter-channel separation of any m-2-n (bei uns hier 4-2-4) matrix system is a combination of two -3dB channels and one -oodB channel.

It is also true that such separation does not permit enjoyment of 4-channel stereo playback with sufficient directional resolution. There are only two courses of action to get around this - giving some psycho-acoustic treatment to the encoded two channels, or exploiting the redundancy in the two channels of stereo media.

There is a certain matrix system which places almost undivided emphasis on the left-right separation. It offers excellent separation in the left-right direction by arranging the -oodB channel in that direction.

Such a matrix, however, loses the rotational symmetry, superposing capability and resultant circular continuity that we discussed before. As a result, the front-back separation by such a matrix is practically nul.

This is almost analogous to connecting the two left-hand speakers and the two right-hand ones separately in parallel and listening to two separate stereo playback performances. It takes us away from our original aim which is to obtain directional resolution in 4 or more directions from two channels.

The reason why we have brought this argument up here is because the symmetry and resultant continuity are vital prerequisites of the new technique that we have developed to improve the inter-channel separation in 4-channel playback.

A symmetrical matrix provides for controlling its separation characteristic by a special technique which we propose to call a "variable matrix" or "Vario-Matrix", instead of resorting to the usual logic circuit which gives an apparent increase in separation by controlling the decoder gain.
.

Das ist so komplex, das ersparen wir uns.
.

Matrixing systems of 4-channel stereo offer one outstanding advantage that they require relatively simple hardware, but have invariably had one basic fault, i.e., poor inter-channel separation.

If we empoy a rotationally symmetrical matrix, however, it has been found that practically sufficient separation is obtainable by controlling the matrix itself.
.

.

This paper was presented by Sansui engineers at the March, 1973 Meeting of The Society for Electro-acoustics, Japan.
.

The greatest advantage of 4-channel stereo lies in the fact that it is able to reproduce sound field information in all 360° directions around the listener (Anmerkung : So die Theorie !!!).

On one hand, this can be accomplisned by utilizing four signal transmission channels. On the other hand, the question was raised whether it could also be done with the present 2-channel stereo system, without requiring a fundamental modification of it.

This would be achieved by specially treating the phases of the 360° sound field signals and encoding them into 2-channel signals, then decoding them into 4-channel signals in reproduction.

The latter approach, called matrix 4-channel, has one disadvantage: It is unable to position the signals as chearly as when four signal transmission channels are employed. This paper describes an improved encode-decode system which overcomes this fault, and reports one practical application of that system.

Scientific analyses to date have made it an accepted practice to regard a sound field as a circle around the listener on which sound sources are distributed.

In some actual program source recordings, however, sound sources inside the circle are also picked up and are a very important element.

Theoretically, the transmission of the entire sound field
information to a listener's room demands an infinite number of channels for transmission, and an infinite number of speakers for reproduction.

However, to design a practical system suitable for consumer applications, it is necessary to sample the sound field information at several points on its circle.

A 4-channel stereo system is a system which samples it as four points on the recording end, and utilizes four speakers to reproduce those samplings.

Such a system may either employ four transmission channels, or employ only two and still reproduce the samplings at four points. The latter is called a matrix 4-channel stereo system.

The first system, which employs four transmission channels, has the sound pressure response pattern as shown in Fig. 2. It reveals the fact that the conventional 2-channel stereo system only transmits and produces a portion of the sound field within ±45 degrees from the center of the listener.

Several system have been proposed until today to transmit all the sound sources on the circle through only two transmission channels. One basic idea is to extend the 2-channel stereo system and double the signal distribution to the two channels (see Fig. 4).

The so-called Regular Matrix system provides ±j phase shifts to the rear signals in order to distinguish them from the front signals.
.

The Regular Matrix standard specifies the use of an encoder as shown in Fig. 6 to encode n number of sound sources on a circle into two total signals.

It is presumed in this case that the n number of sound sources are directly encoded into two total signals. In practical applications, however, a so-called n-4-2-4 conversion process is often more apropriate where the n number of sound sources are once 'mixed down' to four channels, which are then encoded into two total signals. In such a case, the encoder as shown in Fig. 6 would be inadequate.

An encoder having four input terminals is therefore more generally used. It shows a somewhat different sound pressure response pattern from a n-2 encoder. It is nevertheless preferrable for the purpose of improving the inter-channel separation by a technique described later.

Only such a 4-2 encoder provides for a proper encoder-decoder relationship in terms of signal level.

As can be seen from Fig. 5, the inter-channel separation in a matrix 4-channel stereo system is basically a combination of OdB, -3dB, -3dB and -oodB. Since this seems inadequate to position sound images clearly, it is necessary to improve it. We have discovered that the improvement can be accomplished by taking advantage of two facts:
.
(1) Music and other sound field information contains redundancy in sound source directionalities;
(2) When a primary sound source (usually one that is higher in level) and other sound sources (lower in level) exist simultaneously, the human ear is less sensitive to the directionalities of those other sound sources than it is to the primary source. This masking of directionalities, however, is different from the so-called loudness masking, in which the lower-level sound sources are masked by a high-level one and become inaudible. (Studies of the directional masking phenomenon are yet incomplete, and it is expected that future psychological experiments will bring us more meaningful discoveries.) When applied to 4-channel stereo, the above two facts mean that, when one of the four speakers is producing a loud sound, the listener's ears are momentarily less sensitive to the directionalities of the other three speakers. If this is true, then it seems possible to make clear the directionality of the primary sound source while broadening the directionalities of others, and attain psychoacoustically the same degree of inter-channel separation as when four transmission channels are employed.
.

With a symmetrical matrix configuration such as the one stipulated by the Regular Matrix system, it is possible to control the matrix itself and improve the inter-channel separation so as to sharpen the directionality of the primary sound source at each moment.

Another way to accomplish the same sharpening of directionality would be to employ a logic circuit which controls the levels of the decoder output signals to improve the apparent Inter-channel separation.

When a loud sound enters the LF input terminal of an encoder and a weak sound enters the LB input terminal, such a logic circuit would work to increase the decoder gain in the LF' channel while reducing that in the LB' and RF' channels, and if necessary, RB' channel as well.

As a result, the inter-channel separation would definitely increase with respect to the loud sound. But since the decoder gain in the LB' channel is reduced, the weak sound that entered the LB channel would no longer be delivered from the LB' channel, and only its crosstalks would be produced from other speakers, creating an unnatural sound field.

Thus, the gain control method and the matrix control method provide similar improvements in the inter-channel separation for a single sound source, yet they are quite different when it comes to establishing the directionalities of two or more simultaneously existing sound sources or the distribution of their acoustic energy.

Suppose that a sound source located in the pi1 direction in Fig. 1 is encoded, and then decoded in the pi2 direction. Then, from Eqs. 1 and 2, the encoder outputs would be

Formel

while the required decoder output would be

Formel

The leakage or crosstalk of D(0') can be changed by controlling 0' in the decoder. In other words, when $-0 =tt, the crosstalk of S(fa) becomes zero. For example, when a signal enters the LF input of the encoder and is decoded in a given direction by the decoder, its sound pressure response pattern is as shown by the real line in Fig. 7.

Fig. 7 Controlling matrix

Now, if the signal entering the LF input is of a high level and the one entering the LB input is of a low level, the LB' decoder matrix can be controlled in the direction of RB', so that the crosstalk of the LF signal contained in the LB' decoder output will decrease. When the decoder matrix finally coincides with the RB' matrix, the crosstalk becomes -codB. Then the level of the weak signal originally present in the LB input will decrease by - 3dB.

Similarly, if the RF'decoder matrix is controlled in direction of RB', the crosstalk of the LF signal will gradually decrease until it is finally -codB. At the same time, however, the crosstalk of the LB signal contained in the RF' decoder output will eventually rise by -3dB of the original LB' signal level, thus keeping the total power intact.

Furthermore, the matrix control as described above is performed while keeping on one single directional axis the sound image formed by the crosstalks and the real sound image formed by the primary signal. Therefore it does not affect the directionality of the final sound image, but only helps to increase the clarity of the directionality.

AnNmerkung : Der ganz Salmon verdeckt, daß dieses System anfängt zu pumpen und das wurde einfach verschwiegen. Und wenn dazu ein Orchester auf allen 4 Kanälen gleich laut spielt, funktioniert diese mathematische Theorie auch nicht mehr.

One practical application will be discussed here of the encoding and decoding equations given in the foregoing section to improve the inter-channel separation and increase the directional resolution of the reproduced primary signal.

Although it is possible to encode multiple original sound sources directly into two total signals, it is more popular to mix down the multiple signals to four signals once. Hence the practical application described herein also adopts this n-4-2-4 system.

Jetzt wiede eine Menge Theorie, die aber in der Praxis eben doch nicht funktioniert hatte.
.

2-1. How to detect control signals
The decoder incorporates the matrix control technique as described previously.
The signals for controlling the decoder matrix are derived from the encoded signals by detecting by means of phase discrimination whether the signals are more densely distributed in the front or rear, in the left-hand side or the right-hand side.

Our system utilizes the coefficients of the control signals as shown in Fig. 10 over a 360° sound field compass. And when such control signals are obtained, the decoder outputs for a single input signal will demonstrate the same sound pressure response pattern as when four transmission channels are employed. In order to obtain such control signals, the front and rear control signals are produced by phase-discriminating the LT and RT signals. At the same time, the left and right control signals are produced by means of L+R /45° and L-R /45°.

Such a phase discriminating system offers the advantage of allowing a wide dynamic range (lmV~10V) for input signals. Another way to detect control signals would be to use a logarithmic amplifier, but it is often difficult to obtain constant control signals when input signals fluctuate. Still another way to detect the signal distribution is to take advantage of the fact that - oodB inter-channel separation is achieved between the diagonally opposite channels - LF' and RB', and RF' and LB'.

Hier brechen wird diese Ausführungen ab, es war reine Theorie.

It is not definitely decided yet at the present time how many transismission channels are really required to capture, transmit and reproduce the total live sound field information.

The various 4-channel stereo systems presently available either utilize four or two transmission channels. We have been of the opinion that a system utilizing two transmission channels is one that is the most suitable for popular consumer usage without degrading the quality of sound reproduction. Based on this belief, we have developed a system actually utilizing only two transmission channels to transmit and reproduce 360° sound field information.

At first, there was an obstacle. When only two transmission channels are employed to transmit the entire 360° sound field information, the resultant encoded program is only able to position the sound sources half as clearly as a system utilizing four transmission channels.

But then, we discovered two interesting facts - that music and other sound field information contain considerable redundancy about the directionalities of sound sources, and that the mechanism of human hearing is subject to a masking phenomenon with regard to the directionalites of sound sources.

On the basis of these two facts, we have devised a way to detect the positional information about the sound sources and a technique to enhance the directional resolution of the primary signal at each moment while broadening the directionalities of the lower-levels signals existing concurrently.

It was reasoned that such a technique would produce a 4-channel sound field whose directional resolution to the human ear is equivalent to a system utilizing four transmission channels.

A practical encode-decode system was developed to incorporate this system and described in this paper. Repeated auditions have proved that the system was indeed almost identical to a system utilizing for transmission channels in terms of directional resolution.

Finally it must be pointed out that psycho-acoustic studies of 4-channel stereo listening are still gravely underdeveloped and need future efforts. When the importance of such studies is more widely recognized, it is expected that they will contribute to the advance of 4-channel stereo systems.

• Anmerkung : Hier steht  mit vielen umschreibenden Worten, es ist eine unausgereifte rein theoretische Möglichkeit, 4 echte Kanäle in 2 Informationen unterzubrigen. Und damit ist dieses System komplett gestorben.

.

.

• 1. "Proposed Universal Encoding Standards for Compatible Four-Channel Matrixing" by Itoh, present at 41 st AES Convention.
• 2. "The Sansui QS Coding System and a New Technique to Improve its Inter-channel Seporation Characteristic" by Takahashi and Itoh, presented at the 42nd AES Convention.
• 3. "Masking of Directional Information" by Ebata, presented at the Spring Convention of The Acoustical Society of Japan, 1972.
• 4. "Characteristic of Sansui QS Vario-Matrix Based on a Psychoacoustic Study of Sound Sources in Four-Channel Stereo" by Takahashi, Itoh, et al, presented at 43rd AES Convention.
• 5. "RM Four-Channel Decoder System Improved In Separation" by Ohkawa, et al, journal of The Society for Electro-acoustics, December, 1972.
• 6. Technical Standard for Regular Matrix 4-channel Sound Reproduction System, Engineering Sub-committee, The Electronic Industry Association of Japan.

.

.

By now you should have a thorough understanding of the basic theories and functions of the QS vario-matrix circuit, and its resultant effects. Presented in the following is a theoretical investigation of the conceptual and functional aspects of two important additional applications of the QS vario-matrix: the QS Synthesizer (with SURROUND and HALL modes) for synthesizing a 4-channl sound field from 2-channel sources, and a Sansui-exclusive Phase Matrix for reproducing CBS SQ*-encoded program sources.

The Sansui QS vario-matrix circuit is intended for high-fidelity 4-channel reproduction, providing enhanced inter-channel separation on a par with the discrete 4-channel method. With it, Sansui's QS system has now the advantages of both matrix and discrete 4-channel audio reproduction systems.

QS vario-matrix has greatly improved the reproduction of not only QS-encoded program sources (i.e. records, tapes and FM multiplex 4-channel broadcasts), but also the synthesization of conventional 2-channel programs into 4-channel sound.

And an important additional advantage: it can be applied as a phase matrix circuit to achieve optimum 4-channel reproduction of program sources encoded with the CBS SQ* 4-channel system (a phase matrix technique) as well.

There are two possible solutions to the need to increase the quantity of audio information to be transmitted through a limited number of channels. (In this case we are discussing the transmission of increased information through only two channels.)

One is the use of an inaudible carrier upon which additional information is conveyed, later to be converted back into audible sound. The other is the compression of the increased number of channels by exploiting the characteristics of that additional information itself, for later expansion prior to playback.

The matrix system is based on the latter concept. By exploiting redundancies in audio information, the signals in the original 360° sound field are compressed(encoded) during the recording process into a 180° spread. Then, in the reproduction process, are expanded (decoded) back into their original 360° spread.

First, the front signals of the original sound field are distributed in phase, and the back signals distributed in reverse phase, between the two composite channels (LT for left total and RT for right total).

The signals located in the front half of the sound circle (i.e. all sounds located between L, R and F for front in relation to C for center in Fig. la) will be compressed in phase into the quadrant of the circle as defined by C, L and R in Fig. lb. Thus the groove modulation of the signals located anywhere in the front half will be more or less horizontal, as indicated in Fig. 2.

Likewise, the back signals (i.e. all sounds located between L, R and B for back in relation to C in Fig. 1a, excluding the actual L and R positions) are cut more or less in the vertical direction as shown in Fig. 2.

Conversely, during the expansion or decoding process just prior to amplification for playback, the front signals distributed in phase into the LT and RT are discriminated electronically and localized for each specific position. The reverse phase back signals contained in LT and RT are likewise localized properly.

The human ear has the ability to spot the position of a sound and the direction from which it comes by sensing the difference in volume (level), or difference in phase, or by sensing the interaction of both factors.
.
Localization by difference in level is distinctive; localization by difference in phase is less so. This phenomenon as it relates to human hearing is yet to be fully explored; we only know that it occurs and can be exploited. The method of LT-RT discrimination generally called "Amplitude Phase Matrix" (Figs. 4a and 4b) is based on differences in level. This method is used in the Sansui QS and most other matrix systems. The method called "Phase Matrix" such as is used in the SQ and New Scheider matrix systems, is based on differences in phase.
.

Er bringt vom Verständnis nur Verwirrung.
.

Perhaps you are already acquainted with the fact that the QS 4-channel decoder with QS vario-matrix reproduces QS-encoded records with excellent inter-channel separation, comparable only to that of discrete 4-channel tape reproduction.

It thus offers a natural and effective reproduction of the original music. This same QS vario-matrix can also enhance the reproduction of conventional 2-channel stereo records in 4-channel. We call this function the new QS Synthesizer. It processes conventional 2-channel stereo records so that they will sound more "alive" in the new and exciting 4-channel format.

.

The QS Decoder, by virtue of the QS vario-matrix circuit, makes it possible for QS- and SQ-encoded records to be processed for high-quality reproduction with almost infinite inter-channel separation on a par with discrete 4-channel tape recordings.

An additional feature of the QS Decoder is that it can process (synthesize) conventional 2-channel stereo records into realistic 4-channel sound with remarkable separation among all four channels.

The introduction of three newly-developed QS vario-matrix IC's (Integrated Circuits), operating in conjunction with one another, now allows the three aforementioned functions to be performed with the highest performance efficiency.

Also, it will simplify alignments and also cut down the production costs of high-quality 4-channel stereo equipment.

Details of each of these three multi-function, high-performance QS vario-matrix IC's and their specifications will be discussed in the following paragraphs.

.

• a. Buffer Amplifier: Produces signals to be fed into an Input Stage Buffer Amplifier and 45-degree Shifters.
• b. Limiter Amplifier: Amplifies input signals ranging from those that are very feeble to those rated at maximum-strength, and converts them into square waves, so that the Phase Discriminator circuit operates properly.
• c. Phase Discriminator circuit: Produces an output control voltage (D.C.) in accordance with the phase difference between the LT and RT square waves delivered by the Limiter Amplifier.

.

Dual inline/16 P/Plastic

Hier kommen jetzt 3 Tabellen mit den technischen Werten. Aber beachten wir die Überschrift - (Temporary) - die drei ICs sind brandeneu - im Januar 1975.
.

.

• a. Input Stage Amplifier: Amplifies the input signals LT and RT to a specified level and, at the same time, produces the sum (LT+RT) and difference signals (LT-RT) signals from them.
• b. Control Amplifier: Controls the gains of the four signals LT, RT, LT+RT and LT-RT supplied by the Input Stage Amplifier according to the control voltage provided by the Phase Discriminator.
• c. Processor: Produces the 4-channel output signals LF', RF', LB' and RB' from the signals controlled by the Control Amplifier.
• d. Output Stage Amplifier: Amplifies the 4-channel output signals LF', RF', LB' and RB' to a specified level..

.

.

• Variable Resistor: Controls the gains of the Control Amplifier stage in HA1328 in accordance with the control voltage supplied by the Phase Discriminator.

.

By utilization of the QS vario-matrix IC's, various types of decoders can be designed and assembled to meet almost any need that may arise in the industry. These decoders range from simplified types to more complex and sophisticated versions.
Figs. 1 and 3 illustrate two typical examples of QS decoders utilizing these new IC's.

• (1) Example of a simplified QS Decoder
  Figure 1 illustrates a schematic of a simplified, low-cost QS decoder (using one HA1327, HA1328 and HD3103P each) which controls the decoding matrix parameter only in two directions - detecting whether the main weight of the signals is in the front or back at any given moment. The block diagram is as shown in Figure 2.
• (2) Example of a sophisticated QS Decoder
  This QS decoder, whose schematic is shown in Figure 3, controls the matrix parameter in the front and back, and left and right directions to achieve better inter-channel separation and tonal quality than that aforementioned low-cost decoder. This decoder comprises two HA1327's one HA1328 and one HD3103P. The block diagram of the decoder is shown in Figure 4.

Figure 1: Schematic of a Simplified QS Vario-matrix Decoder
Figure 2: IC Block Diagram of a Simplified QS Vario-matrix Decoder
Figure 3: Schematic of a Sophisticated QS Vario-matrix Decoder
Figure 4: IC Block Diagram of a Sophisticated QS Valio-matrix Decoder
.

Und das wars dann auf Seite 51. Sie brauchen da nicht mehr anzurufen, es ist seit 30 Jahren keiner mehr da.



If you have any INQUIRIES ABOUT THE QS SYSTEM (technical questions on the system itself or the QS vario-matrix IC's or questions about licensing contracts for the system), please contact one of the following addresses.

SANSUI ELECTRIC CO., LTD.
14-1, 2-CHOME, IZUMI, SUGINAMI-KU, TOKYO 168, JAPAN TELEX: 232-2076/TELEPHONE: 323-1111

SANSUI ELECTRONICS CORPORATION
55-11 QUEENS BLVD. WOODSIDE, NEW YORK 11377, U.S.A. TELEX: NEW YORK 422633 SEC UI/TELEPHONE: (212)779-5300

SANSUI AUDIO EUROPE S.A.
DIACEM BUILDING, VESTINGSTRAAT 53-55, 2000 ANTWERP, BELGIUM TELEX: ANTWERP 33538/TELEPHONE: 315663