This chapter considers the special problems of reproducing and copying sound recordings which have a “spatial” element - stereophonic, quadraphonic, and so on.

In case you didn’t know, I should make it clear that no perfect system of directional sound reproduction will ever be achieved. The plain engineering reason is that (for loudspeaker listening, at least) the amount of information we need to reproduce is many orders of magnitude greater than we need for high-definition TV, let alone audio. But there are more subtle reasons as well. Human beings use many sound localisation mechanisms, which are learnt and correlated in early childhood. Different individuals give different weights to the various factors, and the sense of hearing works in conjunction with other senses (especially for this purpose the senses of sight, balance, and the sensation our muscles give us as they move). And there are other factors you wouldn’t think would make a difference (such as the relative importances of reverberation in the recording and playback environments, the degree of curvature of sound waves, etc.).

No recording can contain all the information needed for accurate spatial reproduction. In the meantime, recording engineers have made compromises, managed performances, and used undocumented “trade secrets” to achieve their ends. So this chapter mainly helps you reproduce the intended original sound, not restore the original sound faithfully. It will guide you through the processes of conserving what directional information there is, nothing more. For an important but brief summary of the problem, I urge you to read Ref. 1.

Spatial recordings generally comprise two or more “channels” of audio-frequency information. “Channels” is rather a woolly term, because in some circumstances one channel can be split into two or two can be combined into one; so I shall use it to mean “intended separate audio signals.” This isn’t much better; but it does allow me to present you with the options in order of increasing complexity.

By this I mean recordings which consist of two separate sound recordings made in synchronism - two “channels.” The first difficulty occurs when the same subject matter appears both in single-channel and two-channel versions. You might go for the two-channel one, on the grounds that inherently it has a better power-bandwidth product; and ninety-five percent of the time, you’d be right. But always remember that occasionally this isn’t true. Usually, a stereo recording can be turned into a mono version by paralleling the two channels. You should remember that sounds appearing equally on both channels will appear on the mono version three decibels higher than sounds which appear on one channel alone. Therefore you should think carefully before abandoning the mono “equivalent” without checking it meets this specification. (A consistent exception to the rule will be encountered in section 10.4 below).

Another difficulty might be that the single-channel version has a better power-bandwidth product simply because it was done on a better recording machine, or because alternative restoration processes can be used (such as the one described in paragraph 7 of Box 4.17 – see section 4.17), or because the two versions have been edited differently. You’d expect me to recommend the two versions be treated like two separate recordings in these cases; but the rest of the time you can assume mono listeners will simply parallel the channels if they want to, and you need only do the job once.

Nowadays, the word “stereophonic” has settled down to mean a two-channel recording intended to be replayed on a matched pair of loudspeakers in front of the listener. But this has not always been the case; you should be aware that not all recordings conform to this definition. For example, the words “stereo” and “stereophonic” were used pretty indiscriminately before the mid-1950s to denote a single-channel recording which had a particularly clear “sense of space” around the performers. You will find the words in publicity-material and record reviews, and on the labels of at least one series of commercial mono discs (the Columbia “Stereo Seven” series of 7-inch 33rpm LPs of 1951-2, published in the USA). All these items are single-channel mono, and should be treated accordingly.

To confuse matters further, in 1960 the British Standards Institution advocated that the word “Stereophonic” should be printed on all disc record labels where the groove contained a vertical element. This was done with the best of intentions, to prevent consumers from damaging such records by playing them with lateral pickups; but it does not necessarily mean that the sound is “true stereo” (recorded with two separate channels of sound). Many such records are “fake-stereo” in some way. This is not the place to explore all the ways in which a single-channel recording can be turned into a fake-stereo one, but the techniques include frequency-division into two channels, inserting 90-degree phase-shifts between the two channels, manual panning, and adding two-channel reverberation to a single-channel original. To the restoration operator, the only question is “how to restore the original sound.” My opinion is that if you can’t get access to a single-channel original, you should transfer the two channels as they are, since that’s obviously what the second engineer intended; but document it as a fake where you know it for a fact, and leave the users to decide what to do about it (if anything)!

Sometimes the “fake-stereo” version may have a better power-bandwidth product than the original mono one. We must then neutralise the fake-stereo effect in order to follow the principles I mentioned in section 2.3. This will require the operator to discover which fake-stereo process was used, and reverse-engineer it where possible.

This leads us to the topic of “compatible stereo” records. Given a correctly-adjusted stereo pickup, most stereo disc records are “compatible” in the sense that satisfactory mono reproduction occurs when the two channels are mixed together. But a few discs were made with deliberately-reduced vertical modulation so they would not be damaged by a lateral pickup. This was done by restricting the amplitude of the “A minus B” signal - the difference between the two channels. This was helped because the human ear is less able to distinguish the direction of low-frequency sounds. The effect might be achieved acoustically (by using two directional microphones whose polar diagrams degenerated to omnidirectional at low frequencies), or electronically (by introducing a bass-cut and/or an automatic volume limiter into the difference signal).

Frequently, this type of compatibility would be achieved subliminally. In a multi-mike balance, for example, most engineers panned bass instruments to the centre, which resulted in lateral vibrations of the cutter (analogous to a mono disc). Certain resonances in a church or hall might be recognised by a balance-engineer as causing lots of (A-B), so he might re-position the microphone to make the effect sound similar for a mono listener, as well as making easily-playable stereo discs. And I know at least one record - for which I was responsible! - where my difference-signal was so great, the record company reduced my tape to mono and turned it into a “fake stereo” one.

In principle, at least, it is possible to reverse-engineer some of these artefacts, and restore “true stereo” when transferring the sound to a medium which is not an LP disc. But should this be done? My personal answer to this is “No - but document it!” In other words, we should not manipulate the (A-B) signal when doing the transfer, but to document when the original disc claimed to be “compatible stereo,” so users may perform the operation on the copy if they wish. And research whether there has been an unmodified re-issue on CD!

Another terminological difficulty concerns the words “stereo” and “binaural.” Nowadays, the former implies that the recording is meant to be heard on loudspeakers, and the latter on headphones. (We shall meet another meaning of “binaural” in section 10.5). Different microphone techniques are normally used for the different listening conditions, and it is the duty of the copying operator to ensure the labelling is transferred to the copy so that future listeners take the appropriate action. But in the 1950s the two words were often used interchangeably, and further investigation may be necessary to document how the sound was meant to be heard. Fortunately, only the documentation is affected; the transfer process will be the same in either case.

In 1993 Thorn EMI’s Central Research Labs introduced “Sensaura Audio Reality,” a process which seeks to modify separate tracks of sound in the digital domain to produce suitable effects for headphone listeners. The published descriptions are (perhaps deliberately) ambiguous about whether the process is meant to sound better on loudspeakers or on headphones (Refs. 2 and 3). We shall have to listen to examples of the same music processed in two different ways before we can decide; but the point remains that you should document the intended means of playback.

“Dummy-head” stereo is the principal form of “binaural” stereo (in the sense of headphone listening). The recordings are generally made using two omnidirectional microphones spaced about six inches apart with a baffle between them. The baffle might have the same shape and size as a human head. For obvious reasons, the technique of having a human head on top of a pole doesn’t often happen; much binaural recording is done with a flat transparent baffle between the microphones.

In the remainder of this section, I shall outline some two-channel systems which are supposed to be better than ordinary stereo or binaural ones. None of them will require special treatment when archive or service copies are made, although nearly all will require transfers to an uncommonly high standard. But the documentation should carry the name of the system, so users will know what to do with the transfer. So we start with some terminological matters.

“Holophonic” recordings are also intended for headphone listening. The published descriptions of the invention were surrounded by an enormous amount of hocus-pocus, and it is difficult to establish exactly what was involved (Ref. 4); but a conservative interpretation might be that a sophisticated “dummy-head” was used. I understand that the acoustic properties of human flesh and the detailed structure of the outer ear were simulated to help resolve ambiguities between front and rear images. Whatever the hocus-pocus may have meant, most people found it an improvement over conventional dummy-head stereo; but it proved remarkably difficult to maintain this improvement through mass-production processes, which shows how important it is that conversion to digital must be done to extraordinarily high standards, if at all.

We revert to loudspeaker listening for our remaining examples of two-channel terminology. “Ambisonic Stereo” is a special case of “Ambisonics,” a spatial-sound processing technique which has much wider applications than just stereo. Section 10.11 will give a fuller description.

The “Hafler System” takes the difference between two loudspeaker stereo signals and sends it to a third loudspeaker (or pair of loudspeakers) behind the listener’s seat. Certain recordings made using coincident bidirectional microphones contain out-of-phase information which represents sounds picked up from the sides of the array. The Hafler circuit takes this information and puts it round the back. It distorts the geometry of the original location, but it reproduces genuine sound from outside the “stereo stage.” The circuit can be applied to many stereo recordings, and some were marketed specifically for this treatment.

“Q-Sound” (Ref. 5 and Ref. 6) and “Roland Space Sound” (Ref. 7) are systems for loudspeaker listening dating from the early 1990s. They claim to take advantage of some psychoacoustic effects to permit reproduction rather wider than the “stereo stage.”

“Dolby stereo” isn’t really a stereo system at all; we shall leave it till section 10.8. Here again I must repeat an earlier point. Various digital compression systems which sound perfectly satisfactory on plain stereo have been known to give surprising side-effects when applied to Dolby stereo soundtracks. So strict rigour in the digitisation process is essential. The spatial information is often very vulnerable, and must be conserved.

In recent years, some workers have attempted to get “stereo” from two single-channel recordings of the same performance made through different microphones, the two recordings not having been synchronised in any way. This might happen when the same concert was being recorded commercially as well as being broadcast, or when a record company insured against breakdowns by recording with two independent sets of kit. The first documented example seems to have been Toscanini’s 1939 NBC broadcast of Copland’s El Salon Mexico (Ref. 8), where NBC had rigged microphones for networks in the United States, and another with Spanish announcements for NBC’s affiliates in South America. The difficulties are (a) identifying sessions done this way, and (b) synchronising the two recordings with sufficient accuracy (ideally a few microseconds).

One might also ask whether this process should be used, since it does not represent the wishes of the contemporary artists or engineers. In any case, the two microphones are extremely unlikely to have been positioned correctly for either stereo or binaural purposes. My view is that so long as the two single-channel versions are not corrupted by the synchronisation process, and there are no other side-effects of creating the “stereo version,” there is no harm in trying to get closer to the original sound; but obviously the archive copies must be left alone.

I should dearly love to be able to tell you about restoring such sounds, but personally I have never been able to get beyond stage (a) in the previous paragraph but one. The technique is simple enough, though, as described by its inventor Brad Kay (Ref. 9). The recordings are approximately synchronised, equalised, and run at almost the same speed and volume. Sooner or later they will crawl into synchronisation and out again. As they do, the effect is heard upon a pair of loudspeakers. What usually happens is that a double-mono effect occurs, which can be verified by switching with one of the items phase-reversed. If this gives a characteristic “phasing” effect, then
achieve synchronism which gives a stereo effect; but having never got this far, I cannot predict what should happen next. Personally, I would put the two recordings on a digital editor with “varispeed”, manipulate one so it synchronises approximately with the other, and reproduce the result through a Cedar “azimuth corrector” which is programmed to improve stereo using certain psychoacoustic algorithms (end of section 7.6). And clearly that must be confined to “service copies” only.

In section 2.6 I spoke about the duties of a professional discographer trained to locate several copies of a given recording. We will now have to ask him to check if they make stereo pairs. For 78rpm discs, visual inspection is the first sign. A few discographical clues exist (not enough to make any hard-and-fast rules). For example, recordings made by branches of the European “Gramophone Company” marked the two matrixes differently. Duplicates of takes were marked with an A after the take-number, and this practice was continued after the formation of EMI in 1931.

The presence of an A means “look out for plain takes as well.” But most of the time it seems the two machines recorded the same electrical signal, and there is no stereo to be had. Also when EMI started using tape recorders (about 1948), it became the practice to give a “plain take” to a direct-cut disc and an “A” to a tape-to-disc version of the same performance, so the idea is unlikely to work post-1948. If two versions of the same take were published, it is likely that one would be used in Europe and the other overseas (e.g. in America), so the discographer’s job isn’t going to be easy.

This is the place to record that the HMV “Artist’s Sheets” on microfilm at the British Library Sound Archive show coded details of engineering parameters for the years 1925-1931. All the evidence supports the theory that Columns 2 and 3 contain serial numbers of the cutting-amplifier and microphone respectively. If the “straight takes” and the “-A takes” have one piece of equipment in common, then the records must be “double-mono.” For pseudo-stereo, we must have pairs of recordings for which both serial numbers are different.

The English Decca Company is thought to have run its “ffrr” and non-“ffrr” cutters in parallel between 1941 and about 1945, calling the former “Take 2” and the latter “Take 1.” Decca did not document its matrixes very well, and did not export many, so only a test-pressing from those years will be likely to lead to success.

And in the 1930s there were numerous minor makes where the same material was recorded on discs of different diameters, or on conventional 78s and “microgroove” versions. These might all bear investigation (but I haven’t found anything myself).

As we saw in chapter 6, stereo recordings only work well when there is pretty tight synchronisation between the two channels. You should be aware that some digital recording systems (for example, the “EIAJ” system - used for the Sony PCM-F1 encoder) do not record the two channels synchronously. They economise by using use one analogue-to-digital converter, switched between the two channels at an ultrasonic rate. This doesn’t matter so long as the same process is used on replay - the sound comes out synchronously then - but if you are doing any digital processing in the meantime, or converting to another system (e.g. the Sony 1610) in the digital domain, you will have to take this into account. The time-difference between two channels sampled at 44.1kHz is eleven microseconds. This makes a barely-detectable shift in the stereo image-location (to the left), and causes a slight loss of treble (again barely-detectable) when the channels are paralleled for mono.

To make matters worse, one manufacturer of digital interface boxes did not know whether it was the left or the right channel which came first, and got it wrong in his design. Thus it is absolutely vital to check that centre-mono sources come out in synchronism from every digital recording, irrespective of the provenance!

Two-channel recordings also exist which are not always meant for reproduction on two loudspeakers or earpieces. Sounds may be “encoded” from three or four audio-frequency channels, and distributed on two audio-frequency channels, in anticipation that they will be “decoded” back into three or four loudspeakers. This principle is known as “matrixing,” and so far all the systems which have been commercially marketed are supposed to give compatible stereo results. They too can be stored in the form of two-channel “archive copies,” but we shall consider how they may be decoded (e.g. for “service copies”) in section 10.8.

For the time being, that is all I have to say about two-channel recordings. In the majority of cases, all we need to do is transfer them as well as we can, documenting any unusual features as we go. But when we come to systems with more than two channels, we may be obliged to do considerable research to discover how the sounds were meant to be heard. The remainder of this chapter basically comprises such evidence.

First, another piece of terminology. When a four-channel recording system starts off as four different channels (for example, four microphones), which are encoded into two channels of audio bandwidth, and then decoded back into four to be reproduced on four loudspeakers, the system is known as a 4-2-4 system. When a system comprises four original sound channels which are kept separate (or “discrete”) through all the distribution processes and end up on four separate loudspeakers, the result is known as a 4-4-4 system. This convention can be extended to most forms of spatial reproduction, and from now on I shall be forced to use it, starting with three-channel recordings.

It is a commonplace of the film industry that everything is stored in threes - “music,” “effects,” and “dialog.” But I shall not be dealing with this type of three-channel recording here; I shall be dealing with three-channel versions of the same sounds. Such recordings are more common in America than in Europe. They are confined to magnetic media, usually half-inch tape or 35mm sepmag film. They usually occurred for one of three reasons, which I shall now outline.

The first reason was to make a two-channel stereo recording using two spaced microphones, plus a third mono recording with a microphone midway between the two spaced microphones. This enabled stereo and mono versions to be accommodated on the same tape and edited in the same operation. It reduced the artefacts of phase-cancellation for the mono listeners. Also, if three tracks of equal dimensions were recorded, it gave each of the stereo tracks the same hiss-level as the mono track. This idea counts as a 2-2-2 system plus a 1-1-1 system on the same tape.

Pairs of spaced stereo mikes often gave faulty stereo imaging on playback - the notorious “hole-in-the-middle” effect. This could be reduced by using the third track for genuine three-channel stereo, which automatically filled the “hole-in-the-middle.” Three-channel stereo was favoured in cinema applications, because if there were only two loudspeakers situated at each side of a wide screen, any member of the audience sitting substantially off the centre-line would perceive the stereo image collapse into the nearer loudspeaker. A third loudspeaker mitigated this. This counts as a “3-3-3” system. The Warner “3-D” system (used for The House of Wax in 1953) had a 35mm sepmag soundtrack made to this philosophy, but it is comparatively rare to find it on complete final-mix film soundtracks.

For domestic listeners, who could be assumed to be sitting in the central “stereo seat” in preference to paying for a third channel, the third microphone could be used to create a stable phantom image between the stereo loudspeakers by using conventional panpot techniques at the studio. This was usually essential in (for example) a violin concerto. With only two spaced microphones picking up the orchestra satisfactorily, a soloist in front of the orchestra might disappear into the “hole-in-the-middle,” or any small movements he made might be ridiculously exaggerated as he moved closer to one of the two spaced mikes. These problems could be ameliorated by rigging a third spaced mike; but it was risky to combine it electrically and record the concerto on only two channels. Careful judgement was needed to balance the soloist (who was very close to the middle microphone) against the orchestra (which was not), and achieve a satisfactory compromise for both stereo and mono listeners. This could be done more satisfactorily after the three-track tape was edited and before it was transferred to the master disc. This counts as a 3-3-2 system; and the resulting two-channel tape was often called “binaural” to distinguish it from the three-channel one.

Thus we have three separate reasons for the existence of three-track stereo tapes. You will not be surprised to hear that I consider all three tracks should be transferred onto the “archive copy” using a multitrack digital medium; but for a satisfactory service copy, we need to delve into the reasons for the three-track tape. If a contemporary two-channel version exists, we shall have to use the three-channel original for its power-bandwidth product, and compare it with the two-channel version to establish how the centre track was used. Was it used only for the mono version? Or was it was used throughout the stereo version (and if so, at what level)? Or was it mixed “dynamically”? (We should certainly need to make a special “service copy” in the latter case, for it would be impracticable to remix it every time someone wanted to hear it!) Another difficulty arises if it had been a genuine three-channel stereo recording in the cinema, and we can only provide two channels for the listeners to our “service copy.” We may need to remix the three tracks into two using a Dolby Stereo Encoder to approach the original effect. And - document it!

First, some details of four-channel systems used in the cinema.

The first major three-dimensional picture, The House of Wax (1953), used two picture projectors for the left and right eyes, plus a third three-track 35mm sepmag film carrying sounds for three loudspeakers behind the screen, as we saw in section 10.5. There was, however, a fourth channel on one of the picture reels; it was intended for effects from behind the audience. I have no information about how well the fourth track worked, but I have my suspicions! In any case, Warners foresaw the difficulties that such a complex projection system would involve, and arranged for the optical soundtrack of the other picture to have a straightforward mono track. Most cinemas either showed the film with “one-eyed” mono or “two-eyed” mono from this comopt track. Since it wasn’t a wide-screen film, the disadvantages weren’t great.

Shortly afterwards, “CinemaScope” films were introduced specifically to compete with television, using a wide screen. The first manifestation in Britain was in November 1953, using three-track sepmag like The House of Wax. But Twentieth-Century Fox soon found a way of dispensing with the sepmag machine. Four magnetic stripes were attached to the 35mm film, which was modified by providing smaller sprocket-holes to allow the necessary space. Three of the tracks were 0.063 inches wide and gave reasonable quality; these were used for the loudspeakers behind the screen. The fourth track was intended for occasional effects behind the audience. It was only 0.029 inch wide, so it was rather hissy. When the fourth track was meant to be silent, a 12kHz gating-pulse was recorded, so the rear loudspeakers were muted instead of emitting hiss. (Ref. 10).

Now we return to sound-only media, and (I’m afraid) some more terminology.

First, “quadraphony” (there are variations in the spelling of this). This generally means a system for domestic listening in which four original signals (e.g. four microphones) are reproduced on four similar loudspeakers - preferably in a square - surrounding the listener in the horizontal plane. Proceeding round the listener clockwise, the individual loudspeakers are usually “left front,” “right front,” “right back,” and “left back.” Experienced listeners criticised this idea because the front speakers would then subtend an angle of ninety degrees at the listener, instead of the sixty degrees generally considered optimum for stereo; but I won’t bother with the various compromises designed to circumvent this difficulty. Instead, I shall concentrate on pure square reproduction whenever I use the word “quadraphonic.”

“Surround” (used as a noun or an adjective) usually implies a system which might be for domestic or auditorium reproduction in conjunction with pictures. It comprises a two-channel soundtrack which gives reasonable reproduction when played back on a pair of stereo loudspeakers to a listener in the ideal stereo seat. But a “surround-sound” soundtrack can be extended to give extra channels - for example, one midway between the stereo loudspeakers (to “anchor” dialogue, as with “Dolby Stereo” which we shall consider later), and one behind the listener (this may not be used continually, but for occasional special effects). A few audio-only versions have appeared on compact discs, but they never seem to work as well without the pictures, underlining that we need to remember there is interaction between our senses of hearing and of sight.

Few subjects have caused more confusion than the subject of “matrixing,” especially in the early days when it was used for squashing four channels into two for Quadraphonic Sound. No-one was able to agree on the aims of the process, let alone its basic implementation. Were the encoding and decoding matrixes supposed to be “transparent” to an ideal discrete four-channel master-tape, or was the master-tape supposed to have been pre-configured with the deficiencies of the subsequent matrix in mind? Was the aim to give perfect downwards-compatibility with stereo, or give degraded stereo at the expense of better multi-channel sound? Did the matrix have to cope with four loudspeakers that were not identical? or not in a square? Was the priority to give accurate reproduction in the front quadrant, or the front and sides (ignoring the back), or what? Were all the signals to be kept in correct phase relationships in (a) the encoded and (b) decoded versions, and if not, what compromises were acceptable? What about mono compatibility? or height information? or listeners off-centre? or listeners sitting on typists chairs who could turn and face any direction? And so on.

The confusion was made worse by the fact that no matrix system was perfect, and different manufacturers provided different “directional enhancement” techniques to improve their results. These often meant that a decoder made for System A’s records and actually used on System B’s records might sound better than System B’s decoder. I shall ignore these enhancement techniques; in an archival context, they constitute “subjective” interference. But if you’re interested in seeing how they worked, please see Ref. 11. The basic theory of a “linear matrix” (i.e. one which works the same at all signal volumes) is complex but fairly well known (see Ref. 12).

I shall now consider the principles behind the major systems used for pre-recorded Quadraphonic media.

The 4-2-4 quadraphonic systems and the 4-4-4 quadraphonic systems require different approaches from the archival point of view. We shall start with 4-2-4 systems, i.e. ones requiring a “matrix.”

Two 4-2-4 systems were promoted in the early 1970s (“QS” and “SQ”), and (“Dolby Stereo”) in 1976. All three systems claimed to be “downwards-compatible” so far as the stereo listener was concerned. That is to say, if the listener didn’t have a decoder, he would just get acceptable stereo, making a “4-2-2 system”. None of these systems was “mono-compatible;” there are always problems with sounds intended for reproduction on the rear pair of loudspeakers. So a fourth was developed by the BBC for broadcasting purposes, where both stereo and mono compatibility were important. (“Matrix H”).

An “archive copy” of a 4-2-4 recording needs only to be on two channels, since expansion back to four can be done at a later date. That’s the point of principle; I do not actually advocate the use of a decoder for making a four-channel “archive copy.” But I appreciate an institution may rebel against installing decoders whenever such a recording has to be played, and therefore I could understand its desire to make four-channel “service copies.”

So, although it isn’t essential for the purposes of this manual, I shall say something about how 4-2-4 matrix decoding may be achieved, so you will do justice to what the engineers intended. I shan’t give all the technical details, I shall only describe the outlines in words. If you seek mathematical definitions of the various matrixes, please see Ref. 12.

This was a development of “the Regular Matrix” invented by Scheiber (Refs. 13 and 14). In Britain about 400 “QS” records were published in the years 1972-4, mostly by the Pye group, and the system was used by the BBC Transcription Service (notably for the wedding of Princess Anne). Unfortunately, BBC policy prevented these recordings from being advertised with the name of the commercial process, so I cannot advise you which ones used it!

I shall first describe the “Regular Matrix,” because it was supplied on a great many domestic amplifiers, although as far as I know there were no commercial records published with it. A point-source sound would be processed by the encoder as follows. If centre-front, it would be sent equally on the two transmission channels, just like centre-front stereo. All the remaining angles from the centre-front line would be halved. In other words, a left-front quadraphonic signal would be treated as if it were half-left stereo, centre-left quadraphonic would become left-front stereo, and back-left quadraphonic would become left-front with an out-of-phase component. Thus all the sounds picked up quadraphonically would be reproduced on a pair of stereo loudspeakers, although not necessarily in phase. All sounds intended to be in the rear two quadrants for the quadraphonic listener would have an out-of-phase component, and would decrease in a mono mix. When decoded back onto four loudspeakers, a theoretical point-source resulted in output from at least two (usually three) loudspeakers. (In other words, there would be a large amount of crosstalk - actually -3dB - between neighbouring channels, although the crosstalk was perfect between diagonal channels).

The “QS” system comprised the above matrix with a modification proposed by Sansui to mitigate the crosstalk disadvantage. (Ref. 15). Before being matrixed, the two rear signals were subjected to phase-shifts of +90 and -90 degrees with reference to the front channels; QS decoders had circuits after the matrix to shift the rear outputs by -90 and +90 degrees to restore the status quo. In the meantime, some types of crosstalk were slightly reduced, at the cost of ninety-degree phase-shifts in the reproduction of sounds in the side quadrants.

In practice, two techniques were adopted to ameliorate the remaining crosstalk. The first depended on the fact that a “perfect quadraphonic microphone” didn’t exist. No four-channel microphone was ever made - or could be made - which would pick up sound from only one quadrant of the horizontal plane and give only one output. However, if the operator in the sound control room listened to the signals through an encoder and decoder, it was possible to juggle with practical microphones empirically to get improved results.

The other technique was incorporated in Sansui’s QS decoder. It recognised what the engineers were trying to do, and it adjusted the gains of the four outputs to enhance the desired directionality. It was a proprietary circuit taking advantage of some psychoacoustic precedence effects; it did not claim to “restore the original sound.” Despite this, the circuit was very successful on single instruments or small groups, but gave anomalous results on simultaneous sounds from all round the circle. In fact, it was Sansui’s circuit which made the system attractive to the critical ears of the BBC Transcription Service.

The QS system cannot be reverse-engineered to give the directional characteristics of the original; it can only give you what the engineers were hearing in the control room. Whether you use a Sansui QS Decoder for a better “service copy” or “objective copy” is up to you, but it has no part in making the “archive copy,” which must be the encoded two-channel recording.

This was also a “4-2-4 quadraphonic” system as defined above. It was invented by the makers of CBS (Columbia Broadcasting System) Records in America, and was used by the EMI group in Britain between 1972 and 1980. There were possibly a thousand issues in all. Unfortunately, Columbia was an EMI trademark in Britain, so CBS records were not allowed to be imported with that trade-name. The British market was insufficient to warrant printing new sleeves and labels, so quadraphonic CBS records were “grey imports” in Britain. Although a large catalogue of imports was offered, the discs themselves appeared with stickers over the sleeve and label logos.

Unlike QS, SQ left the left-front and right-front signals alone, so if the wanted sounds were confined to the front quadrant, SQ discs could also be sold to stereo customers and played on two loudspeakers without any perceptible difference. (The system altered the actual reverberation, however, and EMI only marketed the two versions as “compatible” from October 1975 onwards).

For full details of how the coders and decoders worked, please see Ref. 12 and Ref. 16. Briefly, the two rear channels were encoded as follows. A left-back quadraphonic signal was split between the left stereo and right stereo channels with 90 degrees of phase shift. A stereo disc cutter would respond to this by tracing a clockwise spiral as viewed from the front of the stylus. A right-back quadraphonic signal would have -90 degrees of phase shift, resulting in an anticlockwise spiral path.

Thus all the sounds picked up in the front quadrant would be reproduced correctly on a pair of stereo loudspeakers, but sounds from other quadrants would be reproduced on a stereo system with phase-reversed components. So long as such sounds comprised natural reverberation this wasn’t too serious. But for single-point sounds positioned at the sides or back of the quadraphonic image, sound-mixers had to listen to the effect through a coder/decoder system and make their own judgements to obtain a satisfactory sound. Sounds intended to be directly behind the quadraphonic listener would disappear in mono, so CBS advised recording engineers not to locate soloists in that position.

A system called the “Tate DES” (Directional Enhancement System) was applied to SQ decoders in 1982, but counts as a “subjective” system for our purposes. The Tate DES was later adapted for “Dolby Surround.”

The SQ system cannot be reverse-engineered to give the directional characteristics of all the original sound; it can only give you what the engineers were hearing in the control room. Your archive copy must be a two-channel version. Your objective copy must be a four-channel transfer through an SQ decoder. (Reference 17 gives details of a “do-it-yourself” SQ decoder. The performance of a SQ decoder may be checked with test disc CBS SQT1100). Your service copy may either be this or an undecoded two-channel transfer, depending on whether the subject matter was confined to the front quadrant or not; the implication of EMI’s announcement of October 1975 is that SQ recordings issued before that date will need to be on a four-channel medium.

This was a matrix system invented by the BBC in 1975-6 with the specific goal of being equally compatible for stereo and mono listeners. Since mono listeners were then in the majority, this was a laudable aim.

When the final reports by the BBC Research Department were completed, it was apparent that there had been several versions. In 1975 the “old boy network” repeatedly carried rumours that secret test transmissions were being contemplated. This was to see if the compatibility claim was met before the final version was announced. But in fact the first transmission was during a 1976 Promenade Concert, and after that one or two concerts were advertised in advance as being “Matrix H.”

The Hi-Fi Press castigated these for their “phasiness.” That is, there were complaints that stereo listeners were being fed an undesirable amount of phase-shift between the left-front and right-front channels (on top of that expected from ordinary stereo reproduction). It isn’t clear which version was being used for which transmission, but versions with inherent 55-degree and 35-degree shifts were tried.

With the collapse of the commercial quadraphonic market in 1977, “Matrix H” was quietly buried. But off-air stereo recordings of that period will have some quadraphonic information in them. It is debatable whether the engineers of the time would prefer us to decode them into four channels now! The last version of the Matrix H decoder (Ref. 18) was never marketed commercially, but relied on the QS Directional Enhancement process. When tested with skilled listeners, the processes gave only 77% success compared with discrete quadraphony.

Some history and some more terminology. Many early “wide-screen” films were on 70mm stock, which incorporated seven magnetic soundtracks. There were five intended for reproduction behind the screen and two “effects” (surround) channels - a “7-7-7” system. This was a pictures-only medium; but the next development became used in a sound-only context.

In 1974 Dolby Labs decided it was time to upgrade conventional (35mm) cinema sound, and they decided from Day One that they needed only three channels behind the screen and one “surround” channel. To get these four channels onto 35mm stock Dolby introduced a new type of matrix circuit. It was called (rather misleadingly) “Dolby Stereo”, although it was actually a 4-2-4 system, not a 2-2-2 system; for this reason, I shall always put the phrase between inverted commas.

Two optical channels on the release-print were located side-by-side and encoded with the Dolby A noise reduction system (section 9.4); but a cinema not equipped for “Dolby Stereo” could get acceptable results using mono projectors which played both channels, when it functioned as a 4-2-1 system. Simplified versions of the matrix were marketed for the public from about 1984, when the first pre-recorded videos using “Dolby Stereo” were issued to the public. The first domestic version was called “Dolby Surround,” and it had a linear matrix. This was upgraded to a “non-linear” one (with additional direction-enhancing circuits) and marketed with the name “Pro-Logic” in 1986; this, to all intents and purposes, duplicated what the professional “Dolby Stereo” matrixes were doing ten years earlier. A Pro-Logic decoder can also be used as a 4-2-3 system (dispensing with the rear channel) to improve television stereo sound; this was given the tradename “Dolby 3 stereo,” but nobody seems to use this name nowadays.

Instead of a square array of loudspeakers (as with Quadraphony), the four loudspeakers are intended to be in the geometrical shape known as a “kite.” (Definition: A kite is a plane figure bounded by four straight lines. Two of the lines which meet at a certain vertex are equal in length, and the two lines which meet at the opposite vertex are also equal in length). Generally, the system is meant to accompany pictures, so I shall assume this point from now on; but a few audio compact discs have been made using “Dolby Stereo” to take advantage of the user-base of Dolby Surround and Pro-Logic matrixes.

Ordinary stereo reproduction comes from loudspeakers L and R, while the viewer sits somewhere near the centre of the kite at V3 (Sorry, an ilustration for this and the next 3 paragraphs is missing.) A picture-display (whose width is denoted by AB) is placed between the stereo loudspeakers L and R. In the cinema AB is nearly as long as LR, but in domestic situations it is not usually practicable to get a video screen big enough for the angle LVR to be optimum for stereo (about sixty degrees). So LR may be four times the size of AB.

A central loudspeaker at C is often advantageous, and cinemas always use one behind the centre of the screen; but any reasonable-sized domestic loudspeaker will be obstructed by the video screen, so it will have to be fitted beneath. The purpose of such a centre loudspeaker is to “anchor” dialogue for off-axis viewers, so the voices appear to come from the picture. This centre loudspeaker only needs to be capable of handling speech frequencies, so good bass response isn’t vital. Therefore another option is to place two (paralleled) good-quality “bookshelf-sized” loudspeakers immediately adjacent to the video screen at A and B (provided their magnets don’t affect the picture). However not every domestic viewer likes this degree of complexity, and most Dolby Surround and Pro-Logic decoders have an option for no centre-loudspeaker at all.

The rear loudspeaker is at R, and again this might actually be more than just one. (Cinemas commonly have three or five, plus more if there’s an upper circle). In practice, film soundtrack personnel do not route sound continuously to the rear, but use it for special dramatic effects. Such dramatic effects usually involve loud and powerful low frequencies (e.g. gunfire, aircraft, bombs, etc), so the rear speaker(s) must be able to handle lots of low-frequency power. It is also helpful if they do not sound like point sources of sound, and sometimes placing them at varying heights helps.

The rear-channel sound is actually encoded as out-of-phase information in the two main channels (so it will disappear if the two main channels are reproduced in mono). I therefore repeat my message that no matrix system can conserve all the directional information; you cannot have a surround-sound system and retain downwards compatibility at the same time, and this is why “Dolby Stereo” licencees are generally accompanied by a Dolby Laboratories representative during the mixing sessions. Various tricks can be invoked at the post-production stage to ensure the listeners believe they’re hearing the original sound when they aren’t (Ref. 19)

The surround-sound decoder uses the out-of-phase property to tell the difference between front-centre and rear-centre sounds, and it is also helped because the rear sounds are supplied with a frequency limit of only 7kHz. Thus the presence of high frequencies also tells the decoder they are meant to be at the front, and it routes them accordingly, even though they may consist of stereo music with many out-of-phase components. This technique also ensures that the effects of the outer ear are not brought into play, so listeners are not tempted to turn their heads. It is therefore an implicit assumption that the audience is seated facing forwards in a way which discourages turning around.

Setting-up the rear speaker(s) is quite a complicated business. To prevent tape hiss coming out continuously during the long periods of inactivity, the rear channel on “Dolby Stereo” recordings is further encoded Dolby B. (This is in addition to any noise reduction used for the distribution medium). Thus the decoder must be aligned so that the Dolby B threshold is correct (section 9.5). Furthermore, rear sounds should be delayed, so the front sounds arrive at the listener’s seat first, and the Haas Precedence Effect (Ref. 20) does not destroy the front stereo image. Ideally, the delay should be about ten or twenty milliseconds at the listener’s ears. In a large cinema this can be quite difficult to achieve, because sound takes about three milliseconds to travel one meter, and setting a delay which is appropriate for the entire audience may need several directional loudspeakers. But this is less difficult in a small room; the Dolby Surround circuit incorporates a delay control which, once set for a particular room, needs no further attention.

In the film world, it can also be assumed that the two channels (the two in the middle of the 4-2-4 system) are encoded Dolby A or Dolby SR on analogue release-prints (sections 9.5 and 9.13); I have ignored this additional complication here. “Dolby Stereo” is quite separate from Dolby Noise-Reduction.

Dolby Laboratories have since developed a digital audio process which can be added between the perforation-holes of analogue SR-coded prints (called “SR-D”), which may also include “Dolby Stereo” matrixed information.

I have mentioned all these details for a very good reason. I consider that it is not appropriate to make four-channel objective or service copies of a “Dolby Stereo” soundtrack, because the delay (and, for that matter, the gain) of the rear channel needs to be attuned to the listening-room in question. And because Dolby engineers were on hand during the mix to give advice about the inevitable consequences of a 4-2-4 matrix, it is better to store the sounds they way they intended them. All three copies - archive, objective, and service - should be two-channel versions. In my view, the “archive copy” should not incorporate decoded Dolby A or Dolby SR noise reduction where it exists, but the others should.

A modified surround-sound system for domestic viewers has been introduced by Dolby Labs (Ref. 21); the audio is in a digital format. The system (called “AC-3”) is a combination of several Dolby inventions, including their digital data-compression methods and the psychoacoustics of the “Dolby Stereo” system. With data-compression, Dolby originally claimed to be able to store their surround-sound information in a datastream of 320 kilobits per second. As with “Dolby Stereo” five full-range audio channels are encoded, plus another for low frequencies only (gunfire, aircraft, bombs, etc.).

Dolby’s data-compression technique employs masking psychoacoustic tricks, but these are augmented by another technique to reduce the bit-rate. When similar sounds are found in more than one channel, the coder economises on data by coding the information only once, registering the similarity. However, it became clear that 320 kilobits per second were not sufficient; Hollywood have been using 384 kilobits per second for their DVD movies, and Dolby themselves now recommend 448 kilobits per second. Fortunately, AC-3 decoders recognise the bitrate!

The geometrical layout seems to be similar to, and is claimed to be compatible with, cinema “Dolby Stereo.” But instead of one channel of sound behind the listener, there are now two, to prevent the “point-source” effect if the listener should turn. The low-frequency sounds may be routed to one or more “sub-woofers”, which (as they are often large) may be situated behind the viewer where he can’t see them.

As described, this writer considers that “Dolby Stereo” and Dolby AC-3 are incompatible, because the cinema and domestic versions will need different source-mixes at the dubbing stage. In addition, the domestic version is in fact a "5½-1-5½" system (the trade press has changed this clumsy expression and calls it “5.1 channels.”) Finally, it seems that the data-compression which results from steering single sounds would result in large “lumps” of data, whose sizes vary with the complexity of the scenes being simulated. It is not clear how this is implemented while maintaining synchronism with the pictures; but with digital video this is a relatively trivial problem. However, broadcasters have never liked it, because the licensing process means ongoing costs for television when broadcasters use it.

However, international politics have entered the arena, and DVD Video discs for the European market were originally mandated to use another bitrate compression system called MPEG-2. This was an “open” system (so anyone could use it; royalties are only payable on hardware); but delivery of the necessary processing software and hardware was about a year behind AC-3, which therefore dominated the first year of DVD. MPEG-2 is used for some European DVDs, and many DVD players are capable of decoding it; but AC-3 is now “the compulsory option” for DVD Video. Yet another system (called “DTS” – Digital Theater System) appeared in 1993.

In the case of DVD, all this was feasible because the physical disc carries the necessary software to decode the bitstream it bears. The implications for archivists, I leave for the reader! As I go to press, the MPEG Forum has now developed a non-backwards compatible version of surround sound called “AAC” (Advanced Audio Coding). It is not the purpose of this manual to predict the future; so I shall stop at this point.

This system, launched in the spring of 2001, is a way of modifying conventional analogue Dolby Surround so the effect may be heard on headphones. There are no transducers picking up movements of a listener’s head, so the circuit (which is an analogue one requiring special processing chips) can feed many headphone listeners at once. It emulates one of three sizes of room-space around the listener; where an entire audience is watching the same film (on an aircraft, for example) only one processor is necessary (Ref. 22). From the point of view of a sound archive, you would certainly need to provide “bypass circuitry” for ordinary stereo or mono listening.

This, also introduced in the spring of 2001, is an upgrade to the normal analogue Pro-Logic decoder for stereo sounds (section 10.12). It must have a “centre front” loudspeaker, and routes dialogue psychoacoustically into that loudspeaker (also Ref. 22). The mono “rear signal” now becomes one with a wider frequency response, and is “spread” to avoid the point effect. Apparently it also offers a “music mode” to generate surround-sound for listening in a car.

The following four sections deal with “4-4-4” systems, beginning with two which were distributed on vinyl LP discs and meet the definition of “quadraphonic” given at the start of section 10.7.

“CD-4” was a discrete quadraphonic system invented by the Japanese Victor Company in 1970 (Ref. 23). More than 500 LPs were issued in the next five years, the chief protagonists being JVC themselves, and the RCA and WEA companies of the USA. A few highly specialist issues appeared (for example, some steam-train noises, which conventional stereo couldn’t handle adequately).

The channels comprising the left-front and left-back sounds were matrixed using conventional “sum-and-difference” techniques. The sum-signal was cut onto the left-hand groove wall with the usual equalisation in the usual manner, so the special CD-4 pickup could also be used for playing conventional stereo discs. The difference-signal was encoded using JVC’s “ANRS” noise reduction system (section 9.7), and modulated an ultrasonic carrier at 30kHz. (This was actually frequency-modulated below 800Hz and phase-modulated above 800Hz, the total passband being 20-45kHz). The right-front and right-back channels were cut on the right-hand groove wall in a similar manner. Reproduction used a pickup with a response to ultrasonic frequencies, fed directly to a CD-4 decoding box. (Some quadraphonic amplifiers were also available with a decoder built-in). After the decoder had been matched to the pickup and stylus (an alignment disc being necessary), this simply reversed the foregoing processes, and provided four channels of audio.

I haven’t written the above paragraph “just for the record”. The process was so complex that I cannot believe any archivist would build a CD-4 decoder from scratch. But I need to make the point that no standard digital recording system has a frequency-range wide enough to permit a “warts-and-all archive copy” as defined in section 1.5. For this and other reasons, it seems that present technology may force us to break our rule, and convert the sounds back into four channels for storage in digital form.

Here are some of the “other reasons.” The ultrasonic-carrier principle strains analogue disc technology to the limit; it is much less likely that vinyl bearing such grooves will have a long shelf-life. Pickup cartridges with a response to 45kHz and good crosstalk figures are essential, and it is much less likely that these will be available in future decades. Even if they were, supplies of “Shibata” diamonds (section 4.11) will also be scarce. Finally, it is known that one playing with a normal stereo pickup is enough to wipe the information from the ultrasonic carrier, and the longer we leave it the more likely this is to happen.

I mentioned that CD-4 alignment discs were necessary for setting up the gear; the only one I actually know is Panasonic SPR111 (which is a 7-inch “45”), but there must have been others.

I haven’t mentioned a certain important principle for some time; but as always, I recommend that the archivist should start by locating the four-track master-tapes. If this isn’t possible, then I recommend that CD-4 discs be decoded and transferred to 4-channel digital recordings as a matter of some priority. These would then combine the functions of “archive copy,” “objective copy,” and “service copy.”

A quadraphonic disc system. The basic idea was a hierarchy of matrixes invented by Duane H. Cooper in 1971, and was developed by Nippon Columbia for LP discs from 1973 onwards. The proponents called it a “4-4-4” system, but this is flattering; there was inherently a certain amount of crosstalk, although always uniform round the square. Despite some optimistic announcements, very few LPs were actually made. The only example I know is the Hi-Fi News “Quadrafile” album of May 1976, comparing the four different systems QS, SQ, CD-4 and UD-4 on four sides.

When announced in mid-1974, UD-4 was described as combining features of several quadraphonic disc systems in an attempt to make a “universal” one, but later revelations showed that it was “just as unique” as its predecessors. The groove contained a true omnidirectional sound channel (rather like Ambisonics, in the next section), and a second difference-channel. This was also omnidirectional except for a frequency-independent phase-lag that, in comparison the first channel, was made equal to the panorama bearing angle for each of the images to be portrayed.

It is not clear how this was supposed to represent two or more sources at once; but on single-point sources the crosstalk was something like that of the regular matrix (-3dB for neighbouring loudspeakers and -infinite for diagonal loudspeakers). It would not have resulted in a stereo-compatible audio signal, but no-one seems to have noticed this point. After de-matrixing, the theoretical “emission polar diagram” for a point source of sound reproduced at the centre of a square array of identical loudspeakers was shaped like a cardioid. The ultrasonic channels improved this in much the same way that two cardioid microphones can be combined to give a “second-order cardioid” response. It was possible to reduce the bandwidth of this second-order signal without much effect on the subjective result, and the UD-4 carrier did not swing beyond 40kHz. (Ref. 24)

Archivists may like to know, however, that Denon did make a “universal decoder/demodulator” (their Model UDA-100). This would decode QS, SQ, CD-4 and UD-4 without any of the “enhancing” circuits, so this is a machine to be sought by sound archivists. (Ref. 25)

“Ambisonics” is a hierarchical system for capturing and reproducing sounds in various directions with respect to the listener. Although this isn’t how it’s actually achieved in a practical session, the “core” of the system may be imagined as an omnidirectional microphone, picking up sounds from all around. (Conventional terminology calls this the “W” channel). Associated with this are up to three additional signals representing the directionality in the three dimensions. (The same convention calls these X (front-back), Y (left-right), and Z (up-down), and if the complete sphere is recorded this way on a four-channel medium, it is called “B-format.”) All these signals are assumed to have been picked up at the same point in space. Time differences do not enter into Ambisonics, so the system is inappropriate for headphone listening. Given perfectly-engineered microphones, the system picks up sounds with the correct polar diagrams, with no crosstalk or dead zones, unlike previous systems.

Ambisonics also comprises means for manipulating audio captured using this basic principle. Such signals may be stored or transmitted on two, three, or four audio channels. Two channels might permit the effects of “stereo,” three channels might permit the effects of “quadraphony,” and four channels might permit the reproduction of height information as well. All these uses of the word “might” are because the results would depend on how the channels were manipulated. Furthermore, loudspeakers are not be confined to any idealised geometrical layouts; the reproducing system could reconfigure the channels to any loudspeaker layout without compromising their information-content. So ambisonics allows infinite flexibility of directional reproduction. (Ref. 26)

So far, all the pre-recorded software has been two-channel for stereo listeners, usually encoded with a matrix called UHJ. I have not been able to find any account of how the UHJ process works, but I gather this too is hierarchical, enabling two dimensions to be matrixed onto two channels and three onto three. A review of the matrixing process considered alone on a B-format signal (encode and decode) is given in Ref. 27. This is the only example of a matrix being reviewed “before and after,” and an implication of this review was that the UHJ matrix was not supposed to compromise the directionality of any of the original four-channel images (unlike other matrixes). It did, though! Once again, the first step must be to locate four-track Ambisonic master-tapes.

I write this section with some diffidence. It seems to me that the existence of a four-channel version of any recording should not go unremarked in a programme of conservation-copying. The extra spatial information should be preserved as well as the best power-bandwidth product. The media in this section are almost certain to have inferior power-bandwidths, but will carry spatial information uncorrupted by matrixing processes. Considerable discographical research will be needed to discover what was going on.

In the USA there were a number of specialist manufacturers of pre-recorded four-track quarter-inch tapes, giving a reasonable-quality 4-4-4 quadraphonic distribution medium for material recorded by various commercial companies. Unfortunately, these were often on acetate-based tape, needing priority copying to preserve the sound.

There was also a “superior” version of the eight-track endless tape cartridge for cars. Instead of having four sets of stereo tracks, this had two sets of quadraphonic tracks. A few were made and sold in Britain, the most important being some of EMI’s best-selling pop albums from the years 1975-1980. Also there was some quadraphonic material which was never available in any other form, notably from the British Decca companies. Unfortunately, the power-bandwidth product of an eighth-width quarter-inch tape running at 9.5cm/sec was very poor.

Whilst these media might represent the intended position of musicians more accurately than a matrixed disc of the same session, we must also remember that the sound-balancers may have been monitoring through a matrix encoder/decoder system. The sales of matrixed discs were usually insufficient to pay for themselves, so it is very unlikely that another mix would be done specially for discrete tapes. I frankly do not know whether this avenue would be worth researching, nor can I say which medium would best match the producers’ intentions.

The official Philips “Red Book” standard for compact digital discs also has an option for 4-channel sound. Nobody has yet taken this up as it implies CD players with four digital-to-analogue converters, although both hardware and software was promised for the second half of 1982. (Ref. 28). But I mention this because it offers the only hope of resolving the difficulty. If commercial discrete 4-channel CD players become available, an archive should pursue them as a matter of principle.

Greater numbers of channels seem to have been confined to film media, although I suppose it’s always possible they could crop up anywhere. For example, “Cinerama” had seven discrete channels. It was claimed that it used seven separate microphones at the shooting stage, and the seven tracks were reproduced on nine loudspeakers (five behind the screen, one “on each side of the orchestra”, one at the back of the auditorium, and one at the back of the balcony). (Ref. 29)

So far I have been unable to find whether similar spatial resolution was actually captured by other systems. “Todd-AO” used six tracks, and the Russian “Kinepanorama” system used nine; but it is at least possible that these were made up from one- or three-channel master recordings with suitable panning, simply as the easiest way of ensuring the sounds came from the right places when reproduced.

When I was working with wildlife filming, the “height” dimension proved essential. I greatly regret that only one system (Ambisonics) has ever offered this, although I developed something using six bi-directional microphones around a cube of absorbent material; this also preserved some “time information”.

I shall describe the effects of multitrack audio recording upon musicians in section 13.5; but here I shall concentrate on the issues affecting an archivist. In all the previous sections, I have concentrated upon reproducing the sound from analogue media “faithfully” (that is, reproducing the sound the way the original engineers intended); but in nearly every case, a multitrack tape implies subjective treatment at a mixing console. Although the analogue tape standards I described in Chapter 6 imply only a problem with “hardware”, there are no international standards in the digital domain, so both “software” and “hardware” is needed for playing most digital multitrack media - let alone the “artistic” elements of the mix-down. Even digital multitrack equipment implied an analogue mixing console until roughly the year 2000.

Unless a satisfactory hardware-based multitrack digital medium is used, there is no way in which a digitised version of an analogue multitrack tape can be preserved in a satisfactory manner (section 3.7). This has a serious side-effect for students of musical history. What happens to performances which were never “mixed-down” at the time?

The only two solutions I can see are as follows. One is to play the unreleased performances into a sound-mixer set with the controls at “flat”, so students will at least get some idea of what is on the multitrack master. The other is to employ a mixing operator familiar with the subject matter, and who can “imitate the style of” the original mixes. This would enable the character of the music to be emulated, rather than a “warts-and-all” reduction. And - preserve the originals!

REFERENCES

• 1: Michael Gerzon, “Surround-sound psychoacoustics,” (article): Wireless World Vol. 80 No. 1468 (December 1974), pp. 483-486.
• 2: Christina Morgan, “Thorn-EMI audio move,” London: Broadcast, 22nd October 1993, p. 16.
• 3: Barry Fox, “Sound waves in sync for better stereo,” London: New Scientist, 23rd October 1993, p. 20.
• 4: IBS News (The Journal of the Institute of Broadcast Sound), No. 22 (November 1985), pp. 16-19.
• 5: Barry Fox, “Business”, Studio Sound Vol. 32 No. 11 (November 1990), p. 32.
• 6: European Patent Application 357 402.
• 7: Barry Fox, “TV Audiences in 3-D,” Studio Sound Vol. 33 No. 7 (July 1991), pp. 58-59.
• 8: Michael H. Gray, “The Arturo Toscanini Society” (article), Journal of the Association of Recorded Sound Collections, Vol. V No. 1 (1973), page 29.
• 9: Barry Fox, “Pairing up for stereo,” Studio Sound Vol. 28 No. 3 (March 1986), p. 110.
• 10: L. F. Rider, “Magnetic Reproduction in the Cinema” (article), Sound Recording and Reproduction (the Journal of the British Sound Recording Association), Vol. 5 No. 4 (February 1957), pp. 102-5.
• 11: Wireless World, December 1972 p. 597.
• 12: Geoffrey Shorter: “Four-channel Stereo - An introduction to matrixing,” Wireless World January 1972 pp. 2-5 and February 1972 pp. 54-57.
• 13: P. Scheiber, “Four channels and compatibility,” J. Audio Eng. Soc. Vol. 19 (1971) pp. 267-279. (Presented at the AES Convention 12th October 1970; also reprinted in “Quadrophony” Anthology, AES 1975, pp. 79-91.
• 14: “Quadraphony and home video steal the Berlin Show,” Wireless World, vol. 77 (1971) pp. 486-8.
• 15: R. Itoh, “Proposed universal encoding standard for compatible four-channel matrixing,” JAES April 1972. (First presented at the 41st A.E.S Convention, 7th October 1971). Also reprinted in “Quadraphony” Anthology, AES 1975 pp. 125-131.
• 16: B. B. Bauer, D. W. Gravereaux and A. J. Gust, “A Compatible Stereo-Quadraphonic (SQ) Record System,” J. Audio Eng. Soc., vol. 19 (1971) pp. 638-646. Also reprinted in "Quadraphonics" Anthology, AES 1975, pp. 145-153.
• 17: Geoffrey Shorter, “Surround-sound Circuits - Build your own matrix circuits using i.cs,” Wireless World, March 1973, pp. 114-5.
• 18: (Haas precedence effect)
• 19: Simon Croft, “Europe Is Surrounded - Using Dolby Surround: how does it affect production?” (article). TVBEurope, October 1994, pp. 38-9.
• 20: P. S. Gaskell, B.A., and P. A. Ratcliff, B.Sc., Ph.D., “Quadraphony: developments in Matrix H decoding” (monograph). BBC Research Department Report RD 1977/2 (February 1977).
• 21: Barry Fox, “Video viewers to surround themselves with sound,” New Scientist No. 1819 (2nd May 1992), page 20.
• 22: Barry Fox, “Dolby’s ’phones and PL2”, London: One to One (magazine), January 2001 page ?
• 23: T. Inoue, N. Takahashi, and I. Owaki: “A Discrete Four-Channel Disc and Its Reproducing System,” J.A.E.S. vol. 19 (July/August 1971). Originally presented at 39th AES Convention, 13th October 1970; and reprinted in “Quadraphonics” Anthology, AES 1975, pp. 162-169.
• 24: Duane H. Cooper and Toshihiko Takagi, “The UD-4 System,” Hi-Fi News & Record Review, Vol. 20 No. 3 (March 1975), pp. 79-81.
• 25: Gordon J. King, “Equipment Review: Denon UDA-100 Decoder/Demodulator,” Hi-Fi News and Record Review, Vol. 20 No. 8 (August 1975), pp. 117-8.
• 26: Richard Elen, “Ambisonics - Questions and answers,” Studio Sound, Vol. 24 No. 10 (October 1982), pp. 62-64.
• 27: Peter Carbines, “Review of Calrec UHJ Encoder,” Studio Sound, Vol. 24 No. 9 (September 1982), p. 88.
• 28: Mike Bennett (Sony Broadcast Ltd.), “Letter to the Editor,” Studio Sound Vol. 23 No. 9 (September 1981), p. 51.
• 29: D.W.A. (= almost certainly initials of Donald W. Aldous), “Cinerama” (article), Sound Recording and Reproduction (the Journal of the British Sound Recording Association), Vol. 4 No. 1 (December 1952) page 24.

-