Interesting papers and lively discussion at the IEE
Almost every conceivable aspect of stereophony was discussed at the excellent two-day convention held recently at the Institution of Electrical Engineers in London, but only the more recent developments or lesser-known aspects of the subject which were discussed are reported here. The full papers and discussion will be published in the Proceedings of the Institution.
Theory of Hearing
F H Brittain described some of the results on stereophonic listening which have been obtained by Dr B Sayers and Dr D M Leakey working under Professor C Cherry at the Imperial College of Science and Technology. These results indicate that in hearing the position of a sound source is obtained from the difference between the times of arrival at the two ears of the sound, as distinct from the intensity difference at the two ears of the sound as has sometimes been thought. The variations of this time difference produced by head movements as small as one tenth of an inch are used to remove the ambiguity as to whether the sound source is in front or behind (both positions would produce the same sounds at the ears). When listening to two loudspeakers in a stereophonic system each ear hears both loudspeakers, but the sounds at each ear from the two loudspeakers apparently combine to form a single sound provided that the time difference between the sounds from the two loudspeakers is less than about 0.5msec. For longer time differences up to about 3msec, such as might be produced by off-centre listening or a spaced-microphone recording technique, partial apparent combination at each ear of the sounds from the two loudspeakers can still occur.
The apparent arrival time of each combined sound relative to the arrival times of each of its two constituent sounds from the two loudspeakers depends on the relative intensities of these two constituent sounds, and also, for delays between the two constituent sounds of more than about 0.5msec, increasingly on the arrival time of the earlier of the two sounds. This latter phenomenon is known as the Haas or precedence effect.
The difference between the apparent arrival times of the combined sounds formed at each ear gives the apparent single source position. Thus intensity differences between the two loudspeakers are converted in hearing into apparent time differences between the two ears. This somewhat surprising asymmetry between sounds from the loudspeakers and the corresponding sounds at the ears can (in case any readers are sceptical about it) be derived trigonometrically (see for example the paper by H A M Clark, Dr G F Dutton and P B Vanderlyn in Proc.IEE Part B, Vol. 104, Sept. 1957, p. 422). This asymmetry arises because each ear hears both loudspeakers - an elementary fact of loudspeaker as distinct from headphone stereophonic reproduction that is sometimes forgotten.
Reproducing Conditions - Rooms
In stereophonic reproduction interfering reflections from the walls of the listening room should be avoided by absorbing material in suitable positions according to T Somerville of the BBC In the excellent BBC demonstrations, given at the IEE during the convention, reflections were avoided by hanging curtains so as to form a sort of room within a room.
The deliberate use of reflections from the room walls using directional loudspeakers has been suggested as a means of producing apparent loudspeaker sources farther apart than the actual loudspeakers and thus as a means of increasing the apparent size of the overall sound field beyond the region between the two loudspeakers. However, this increase will not occur for the reverberant sound which consists mainly of low frequencies at which loudspeakers are not directional.
Reproducing Conditions - Loudspeakers
A discrepancy between the apparent size of the reverberant and direct sound fields also arises in the sometimes used system in which signals at frequencies below about 300Hz in the two channels are added together and reproduced by a single loudspeaker. This economical system is generally advocated on the grounds that stereo information is mainly confined to the upper frequencies and that sound source positions can be located less and less accurately as the frequency decreases. This latter point was challenged by H A M Clark and by F H Brittain who stated that he had been able to locate mortars fired on Salisbury Plain at great distances to an accuracy of one degree although their sound on arrival did not contain any energy above 50Hz and had a maximum energy at 20Hz.
A combination of the two stereophonic signals to form a single signal which can be reproduced by ordinary unmodified single-channel equipment so as to sound as good as or only slightly inferior to the corresponding non-stereophonic signal is said to be compatible.
The sum of the left and right stereophonic signals is often stated to be compatible, but a number of criticisms of this view were put forward in two papers by T Somerville and D E L Shorter of the BBC. This sum signal is least compatible when a multi-microphone recording technique has been used. Such a technique may be necessary to correct for poor acoustics and is often used in 'pop' music or drama to correct the relative intensities of the various parts of the sound source. When the alternative of a single pair of coincident (or nearly so) microphones is used its optimum position may not be the same as that for a single microphone for a single-channel recording. Also when a coincident microphone recording technique is used, this sum signal may be equivalent to that from a microphone of unsuitable polar response, though according to Dr W S Percival this response can be improved by altering the phase of one of the stereophonic signals by ninety degrees before addition. Incidentally, use is often made of the fact that signals obtained with microphones with one kind of characteristic can be combined to produce signals which would have been produced by microphones with different polar responses. When a pair of spaced microphones is used the difference between the times of arrival of the same sound at the two microphones may produce some cancellation at various frequencies in the sum signal. A revealing ear opener was provided in the BBC demonstration when, with a soprano, choir and orchestra, the emphasis on the soloist was very much greater when the sum of the left and right signals rather than the stereophonic recording was played.
Two new systems were mentioned by T Somerville and D E L Shorter. These use primarily amplitude modulation, and this can be at frequencies in the ordinary medium Waveband as distinct from the more usual frequency-modulation multiplex systems at VHF. In the RCA system the left- and right-hand signals are carried one on the upper and the other on the lower sideband, so that this might be described as a double single-sideband system. An ordinary detector will give an output proportional to the sum of the left- and right-hand channels. In the Philco system the sum of the left- and right-hand signals is used to amplitude-modulate the carrier in the usual way and provide a compatible output from an ordinary unmodified receiver, while the difference between the left- and right-hand signals is used to phase-modulate the carrier. The difference modulation is confined to frequencies above 300Hz to avoid certain complications in the receiver circuit. In both these systems distortion is possible in the detector when one modulation is much deeper than the other.
In an interesting paper by I I Geluk of the Netherlands Broadcasting Union (some of the results from which have also been given by the same author jointly with H I van der Heide in EBU Review, Part A (Technical) for Feb. 1959, p. 15) a comparison was made between a number of versions of the VHF FM multiplex system in which the sum of the left- and right-hand signals is used to frequency-modulate the main carrier and provide a compatible signal for an unmodified receiver, while the difference between the left- and right-hand channels is used to frequency-modulate a sub-carrier of the main frequency. In all the cases considered the total maximum deviation was kept at 75kHz while being sub-divided in various ratios between the two modulations. Three factors of importance in comparing various systems are the loss in the power transmitted from the main compatible carrier caused by the power which must be diverted to maintaining and modulating the sub-carrier; the differing amounts of filtering between the carrier and sub-carrier necessary to secure a sufficiently low cross-talk and finally, the various amounts of spurious radiation which could cause adjacent-channel interference. A possible detector for the difference signal was described using four double-triodes in which the difference signal is used to synchronize a multivibrator. By limiting the output from the multivibrator rather than limiting the difference signal itself very efficient limiting is obtained. With efficient receiver limiting, the sub-carrier deviation can be decreased so that the power loss on the main channel is only 2.7dB and also cross-talk between the two signals and spurious radiation are reduced.
Systems in which the sub-carrier is amplitude rather than frequency modulated were also considered by Geluk. The detector can be simpler than with an FM sub-carrier, but the RF stages are more difficult to design, since a flat amplitude response over the channel bandwidth is required to avoid distortion rather than the more easily obtained flat phase response required for an FM sub-carrier. The cross-talk tends to be higher using an AM sub-carrier so that, unlike the case with an FM sub-carrier, it is unlikely that two independent signals could be satisfactorily transmitted. As was pointed out by D E L Shorter, a time-division multiplex system would be equivalent to using a sub-carrier spaced by the switching frequency and AM by the difference signal.
In considering the optimum division of available power between the sum and difference signals, it is often assumed that the peak level in the sum is considerably greater than that in the difference. In fact, the usual difference between these levels is only about 4dB, according to Shorter, and can be less even than this value.
Considerable interest was aroused by the discussion of the Percival system, of which some details have already been given in Wireless World (Vol. 64, Nov. 1958, p. 521) and in which the directional information is separated from the signals themselves and processed so as to occupy a bandwidth of only about 100Hz. The processing is primarily based on the Haas or precedence effect which can be stated in the form that in hearing the apparent direction of a sound is determined primarily from the first few milliseconds of it. Thus the directions of a number of nearly simultaneous sounds can be apparently given by stereophonically reproducing only their beginnings provided that these beginnings are not simultaneous to within a few milliseconds. Even when the whole orchestra is supposed to start together the attacks of the individual instruments seldom coincide exactly and the system usually remains unbeaten. The processing of the directional information thus mainly consists in heavily weighting the initial transients of signals.
One advantage of separating the directional from the main signal is that the latter signal can be obtained from one or more microphones independently of the directional signal, and so can be made more compatible. Some attention was given to the possibilities of stereophonic sound in television. This is surprising since in this case the widest possible sound source corresponds to the widest possible picture so that little benefit would seem to be obtainable from stereo in home viewing even considering large projection models. The difficulty of changing the apparent sound field to correspond with every change in the field of view is also obvious.
Two lesser-known but still very important sources of distortion which apply equally well to monophonic or stereophonic pickups were discussed in a paper by D G Iaquess relating to the Decca stereo pickup. These sources of distortion are, first, impacts between the stylus and groove when vertical groove motions are not satisfactorily traced (which cause most of the wear on the record), and secondly, longitudinal stylus movement in the direction of the unmodulated groove.
The reasons for using two balanced pairs of magnetic gaps in the design of variable-reluctance stereophonic pickups were discussed in a paper by S Kelly. If each magnetic gap between the moving armature and a fixed pole piece is not balanced by a corresponding gap on the other side of the armature, then, as the side thrust varies across the record due to changing tracking error and frictional force in the modulated groove, the lengths of the un-paired gaps will alter so as to alter the relative sensitivities of the two channels. The sensitivity of at least one of the channels will for a similar reason vary with the playing weight, though for gap lengths parallel to the two modulation directions this will not affect the relative sensitivities of the two channels. Another disadvantage of an unbalanced gap system is that induced hum cannot be cancelled by suitably winding two separate coils for each channel.
An interesting sidelight on the hazards of pickup manufacture mentioned by P Wilson in the discussion is that the rejection rate is about 75% for the coils of two American pickups using 54 and 56 gauge wire.
Four-track Tapes and Cartridges
Two important disadvantages of tape relative to disc - its higher price per reproducing time (particularly for stereophonic tapes) and the inconvenience of threading it on to the tape deck - have both been tackled recently in America. Some of the results were described in papers by A D Burt and D R Andrews and also by Dr G F Dutton. The price per time of tape has been reduced both by using four instead of two tracks across the ¼in width and also by halving the speed to 3¾in/ sec. To secure the same response at high frequencies at 3¾in/sec as at 7½in/sec the effective gap length of the reproducing head must be halved down to about 10-4in, but this can be achieved.
To keep losses caused by spacing of the tape from the replay head due to the tape roughness to an acceptable value the tape must be made smoother, Smoother tape (and also greater freedom from dropouts) are also necessary to allow the halved wavelengths at the slower speed to be recorded satisfactorily, The allowable angular azimuth error of the gap length direction is proportional to the recorded wavelength divided by the linear error along the tape between the two sides of the gap, ie, this error is proportional to the tape speed divided by the track width, and thus remains unchanged. In stereophonic reproduction crosstalk problems are reduced by staggering the four tracks so that no two stereo tracks are adjacent to each other, The signal- to-noise ratio tends to deteriorate when four rather than two tracks are used since the signal is proportional to the number of magnetic particles, ie, the tape width, and the noise is more nearly proportional to the square root of the number of magnetic particles. An increased recording emphasis at frequencies near the region of maximum hearing sensitivity has been suggested as a method of reducing the apparent signal-to-noise ratio. According to Dr G F Dutton however, results on orchestral output spectra show that peak signal levels in this region can only be increased by about 4dB or less before overloading occurs.
The slower speed also increases the difficulty of making the transport mechanism sufficiently free of wow and also (particularly) of flutter. For example, a slipping clutch providing take-up tension tends to be jerky and produce longitudinal oscillations of the tape. The flutter produced by such oscillations will be more serious at the shorter wavelengths recorded at the slower speed. The inertia of a given capstan which tends to smooth out irregularities in the tape motion will be quartered at the slower speed. The wow and flutter for the RCA 3¾in/ sec tape cartridge on its deck was quoted as about 0.3%.
Easier handling of the tape has been achieved in a number of ways. For example, half loops at the ends of the tape are automatically caught and pulled through the transport mechanism in a development of the Armour Research Foundation of America, The paper by A D Burt and D R Andrews described the RCA cartridge. In this the tape is enclosed in a plastic container which can be readily slotted into its correct playing position on the tape deck. Thus the cartridge can be removed even when the tape is only partially played through, and a particular tape cannot be 'lost' by being transferred on to another reel by mistake. The two hubs carrying the tape before and after playing do not have any flanges on them at the side of the tape so that the whole space between the hubs can be filled with tape throughout the playing time. Windows in the cartridge allow the amount of tape which has been played to be seen. The tape is secured by a loop at each end which slips over a peg on each hub. At the end of each reel the change in the position of the end of the tape as the peg completes its last revolution can be used to actuate a trip mechanism to stop the tape or reverse it and replay on another track. Interlocks (which can be removed) between the transport mechanism and cartridges carrying pre-recorded tapes prevent such tapes from being accidentally erased.
Is Stereo Worth While
Dr W S Percival was brave enough to give an opinion on this question. With both stereo and mono reproducers their value increases proportionately to their price at low price levels but tends to flatten out to a constant value at high price levels. At first the value for money increases more rapidly with single-channel than with stereo reproducers. The maximum attainable value is, however, greater with stereo so that at some point the value-for-money curves cross for the two types of reproducer. Dr Percival would not, however, commit himself as to what price this occurred at!