The idea of recording sound on optical film is so closely associated with the idea of recording moving pictures photographically that it is practically impossible to separate them. I do not propose to try. All I can do is to list some techniques which sound operators in other genres have up their sleeves, which could be applied to optical tracks (and should be, if we agree to “restore the original intended sound”).

Before I do that, I shall mention two or three optical recording formats which were intended for sound-only applications. These did not survive long because of the costs. Photographic systems, such as the Selenophone and the Tefi-Film, required silver-based chemicals, which were coated onto bases requiring precision engineering techniques, and suffered from the further disadvantage that the soundtrack could not be replayed (or even seen) until the film had been developed much later.

The Philips-Miller mechanically-cut optical film did not have the processing lag, but it was confined to areas where mass production could make no contribution to reducing costs, and it so happened that the films were on a cellulose nitrate base. The originals now seem to have been scrapped, either because they were degrading, or because of the fire hazard. I do not know of any original Philips-Miller recordings surviving anywhere; we only have tape and LP transfers. It is possible, of course, that some of the techniques for optical sound restoration might still be applicable to these. Details of the mechanical cutter may be found in Ref. 1.

But even when we have a sound-only optical format, it is the author’s experience that it is better to make use of the expertise and equipment of a film-based archive to be certain of reproducing the power-bandwidth product in full. As I said in my Introduction that I shall not write about esoteric media, I shall just describe the soundtracks accompanying moving pictures.

Oddly enough, the history of moving films has always meant that sound has taken second place. I do not mean to say that sound recording was a neglected art - although it sometimes was of course! - but that sound reproduction was generally made “idiot proof”. It makes the job of the film archivist easier for two reasons. Firstly, film sound engineers usually took reproduction difficulties into account when they made their soundtracks. Secondly, standards were maintained with few alterations, otherwise the “idiot-proof” advantages would be lost.

On the other hand, if we are working in an environment where we wish to restore the original sound (rather than the original intended sound), then it is a different matter, because the working practices of contemporary engineers were largely undocumented. This problem might occur if we are taking bits of optical film soundtrack for sound-only records, or if we are working for the original film company and they have changed their practices.

Thousands of improvements were proposed to the idiot proof methods over the years, but they were all essentially experimental in nature, and very few broke the mould. As I said I would only concentrate on mainstream media, I feel absolved from listing the experimental setups. As far as I know, all these experimental films can be tackled by an experienced operator with a thorough grounding in the basic principles listed here and elsewhere, in the same way as an operator experienced with grooved media can recover the full power-bandwidth product from alien cylinders and discs.

From now on, I shall be making frequent reference to different “gauges” of optical films. Gauge is always quoted in millimetres, and includes both the picture and the sound when they are combined onto the same print.

Archivists are hampered by the fact that there is no satisfactory non photographic method of storing the pictures. As I write this (early 2001), we are a very long way from being able to convert the best moving picture images into electronic form, much less digital form. Indeed, we can only just envisage methods of dealing with “average” quality film pictures - say standard 35mm prints - and even with this material, the apparatus isn’t actually available yet. I am talking now about “archive” and “objective” copies.
It remains perfectly feasible to use quite inexpensive video formats for “service copies.”

As a sound operator, I consider the prime difficulty is simply that no electronic method can give pictures to the speed accuracy normal in audio. Until this problem is solved, I cannot see that it is worth bothering with the pictures! But, as someone who has had to work professionally with film pictures, I can see that questions of definition, contrast range, field of view, etc. also desperately need attention.

Even if these problems were to be solved, we would probably not be able to reproduce the images in the manner intended by the original makers. An adequate “screen” for electronic display of film pictures is only just beginning to appear (the flat screen plasma or LCD display); but this will remain peripheral for the following reasons.

Films were mosty exhibited in cinemas and "peep-shows" before the arrival of sound, and a standard running speed wasn’t needed because the human eye is generally more tolerant of speed errors than the human ear. As projected pictures grew brighter, faster frame rates became desirable because the human eye’s persistence of vision,” which is shorter in brighter light. Average frame rates gradually climbed from 12-16 frames per second in the early 1900s to between twenty and twenty-four by the 1920s. When the “talkies” started, it was a natural opportunity to make a clean break with former practice, and upgrade and standardise at the same time. The result was 24 frames per second. Later developments have shown the choice to have been very wise - it is practically the only speed at which you can shoot under both 50Hz and 60Hz mains lighting). So when 24 frames per second was adopted with sound, that was it - further changes were not possible without upsetting the sound.

The boot was now on the other foot. National standards authorities recommended a peak screen brightness for indoor cinemas in the neighbourhood of 10 foot-lamberts. (Ref. 2). This was done to minimise the motion-artefacts for the cinemagoer. It was also assumed that the screen would occupy most of the field of view of the observer with no surrounding visual distractions, and there would be no stray light to fog the image and reduce the contrast range, which might otherwise be as much as 50: 1.

However, electronic displays were developed for domestic viewing (television). This meant pictures many times brighter than 10 foot-lamberts. Under these conditions the persistence of vision is shorter, and motion becomes appreciably jerky (especially on a big screen). To ameliorate this, television’s brightness range rarely exceeds 10: 1, otherwise details may be lost because of stray light from the surroundings. Domestic viewers will never put up with having their rooms stripped of visual distractions, nor will they put up with a screen which occupies most of one wall. So material shot for television has a faster pace of editing, more camera movements to provide instant recognition of depth using parallax, larger screen-credits, and frequent closeups of characters’ faces. With no consumer base, the finance for developing suitable processes for cinema-style images in the home seems unlikely.

For this reason, most film archives have kept photographic film as the destination medium for their restoration work. It is a well known medium. It can be projected easily to paying audiences in an idiot-proof way, and converted (“downgraded”) to whatever electronic medium users might demand. Its costs are high but well known, and it is a much more stable platform to work from than rapidly shifting video technology. It is not a format with great longevity; but the problems are well understood and the timescale is predictable. Unhappily, the technique of copying photographic films onto photographic films gives difficulties with the soundtracks, because they are essentially being copied from an analogue medium to the same analogue medium, hence potentially halving the power-bandwidth product.

Controversy even rages on this point. Modern optical film soundtracks are capable of slightly more power-bandwidth than many early ones (especially those made before 1934). Should these be raised in volume to fill the power-bandwidth product of the destination medium to reduce the losses, and make them ready for screening without rehearsal? Or should the original tracks be photographically copied as they are (warts and all) to preserve the style of the soundtrack unadulterated? Unfortunately, the cost of picture copying and storage is so high that it is usually impossible to afford two versions, “service” and “objective” copies to use our language. This, in turn, means that if we get the ideal high-definition variable-speed digital video format and its screen, it will also have to hold many alternative synchronous PCM digital soundtracks.

Because of all this, I make no apology for the ivory tower nature of the following sections. I shall simply make recommendations for recovering the power-bandwidth product of optical film soundtracks, without any reference to how the results might be stored or displayed.

There are many types of optical soundtrack2. The simplest is the variable density type and its equivalent is the variable area or variable width type. When either of these passes through a narrow uniform slit of light shone across the track, its brightness is modulated by the film, and the remaining light can be converted directly into an electronic audio output. The device which performs this task has had many names. Until the 1960s, it was always a “photo-electric cell” or “photocell”; but nowadays we use the catch-all term ‘photoelectric devices’.

Somewhat more complex types of soundtrack are a dual variable-width track, in which the net effect is the same as the variable area type but the modulation is split into two halves. In the pushpull track the two halves are out of phase. This cannot be reproduced with a single photoelectric device. The scanning beam has to be split into two and fed to two separate photoelectric devices, and the output of one of them reversed in phase electrically. Many sources of even harmonic distortion are cancelled this way.

In some types of variable width track, in the absence of sound, the clear area closes down to a thin narrow line. Thus the effects of dirt and halation are reduced; but it is necessary for the timing of the circuitry to be carefully optimised to eliminate thumping noises and the breathing effects following transient sounds. Again, a variable density equivalent exists, which also has the advantage that a limited amount of volume changing can be done during the film printing process if required. With a push pull equivalent of this format the thumps are automatically cancelled, and the noise reduction can therefore be faster.

In a Class B push-pull track the positive going and the negative going halves of the wanted sound are separated. This gives optimum immunity to dirt whilst using a single photoelectric device, but has the risk of crossover distortion at low signal volumes, and increased distortion at all signal levels if the illuminating beam is not uniform and on-centre.

The advantages and disadvantages of most of these formats were worked out during the first twenty years of sound on film, and are described at great length in Reference 3 and elsewhere. The sound restoration operator is obviously helpless to do anything about these types as they come to him, except understand the disadvantages and how they may be minimised (e.g. by ensuring the beam is uniform, the print clean, the dual photoelectric devices are in place, etc).

Because cinema pictures undergo intermittent motion in the projection gate whilst the soundtrack must run at constant speed, the sound is displaced with respect to the picture. On all optical sound formats it is nearer the head of the roll. As the film usually travels in a downwards direction when it is being screened optically, the sound head is placed beneath the picture gate. The intervening film forms a flexible loop while a flywheel stabilises the speed at the sound head. Compliant and resistive mechanical parts (springs and dampers), and power assisted mechanisms, may be provided to allow the flywheel to get up to speed in a reasonable time without the film being scratched. All this needs critical levels of care and maintenance to avoid wow and flutter, particularly noticeable on narrow-gauge formats.

Students of the history of technical standards for 35mm film will find that the amount of this displacement has apparently varied, with 19 frames (14.25 inches) and 20 frames (15 inches) being quoted. It wasn’t until the ISO Recommendation of December 1958 that this was made clear. There, it was set at 21 frames, and an additional paragraph explained that when the distance was 20 frames, the picture and sound were in synchronism for an observer at a distance of 15 meters (50 feet) from the loudspeaker.

In theatre projection circumstances, these figures might be altered to allow for the time taken for the sound to travel from the loudspeakers behind the screen to the middle of the auditorium, or perhaps to the most expensive seats (which tended to be even further away). The size of the lower loop was the usual way of regulating this; so if you are taking the sound from a conventional projector and moving it to another medium, this displacement should be checked and, where necessary, allowed for.

On 16mm films the nominal difference is 26 frames, and on super-8mm films it is 22 frames.

It is worth noting that all three gauges may be found with magnetic stripe instead of, or in some cases in addition to, the optical soundtrack. In an attempt to keep to the standards, early workers with magnetic stripe kept their displacements the same as optical film. Towards the middle of the 1950s magnetic sound-on-film became dominant for handheld shooting, such as news film. Because television news programmes frequently needed to mix optical and magnetic stories on the same reel and transmit them without relacing, the magnetic heads were thereafter displaced by different amounts: 28 frames in the case of 16mm and 18 frames on super-8mm.

Magnetic stripes for 35mm and larger gauge formats were generally used with wide-screen spectaculars having stereophonic sound. Because the conventional projector layout could not accommodate extra heads in the usual place, they were mounted above the picture gate. On fourtrack 35mm Cinemascope prints, for example, the distance is minus 28 frames.

All the standards authorities agreed that the tolerance should be plus or minus half a frame, and that the measurement should be from the optical slit (or magnetic head gap) to the middle of the appropriate picture frame.

This writer has many times been faced with the question, “What if it looks wrong?” A skilled film editor can easily judge less than half a frame on critical scenes. (This isn’t difficult to comprehend when you realise that orchestral musicians can consistently follow a conductor with an accuracy approaching a hundredth of a second). This may mean separate archive and service copies.

As with any analogue recording, the basic principle is that the nearer you are to “an original,” the better the quality is likely to be. There are, however, a couple of special caveats for optical film.

The first is that the original may often be an optical negative. If this is the case, the principle of the narrowed track - preventing the reproduction of dirt - is stood on its head, and it no longer works. However, the high frequency response and the harmonic distortions will be better on the negative, and printing to positive stock will lose some of this information. I have no practical experience of my next suggestion, but as a sound operator I consider it would be worthwhile reproducing both the negative and the positive print, and combine them using the digital equivalent of the process outlined in Section 0.

I must also record that Chace Productions Inc., of Burbank, California, have a process for scanning an optical negative digitally. This allows the process of ground-noise reduction to take place in the digital domain; but I have no experience of the results.

Where there is no narrowed track (or its variable density equivalent), the maximum power-bandwidth product is bound to exist on the negative. (Or, to be pedantic, the original positive in the extremely unlikely case of a comopt reversal original). But even-harmonic distortion will occur because the “gamma” (the linearity of the relationship between the brightness and the desired output voltage) will normally have been controlled in the printing and development stages. However, provided the negative is transferred as it is without any phase shift, even harmonic distortion can (in principle) be corrected in the digital domain.

An optical soundtrack may be copied to another in three ways. Contactprinting holds the master and the unexposed films in close contact and shines light through one onto the other. This gives perfect transient response, of course; but it is not often done, because the films have to be held in close registration by their sprockets, and this can increase flutter. Optical printing shines light from one film running on one continuousmotion transport through an optical system directly to the unexposed film running on another continuousmotion transport. Various aperture effects can cause high-frequency losses, and careful sensitometric testing is essential to avoid harmonic and intermodulation distortions, but the system is sometimes used for copying sound from one film gauge to another.

When used for this purpose, the frequency characteristics are not necessarily optimised for the destination medium, so the third method is preferred (although much more expensive). This is called electrical printing, and simply means that a reproducer is connected to a recorder using an electrical connection (the same means as copying magnetic tape). Thus aperture effects may be compensated at the expense of putting the sound through twice as many optoelectronic and electrooptical transducers. Frequency re-equalisation and possibly dynamic compression may be employed to give better results, especially if the destination medium is a narrower gauge; but the full disadvantages of analogue copying (doubling the distortion, noise, and speederrors, and introducing subjectivism) can occur.

The soundtrack is usually scanned by a narrow rectangular beam of light, corresponding to the gap in a magnetic tape head. Because the film is usually moving downwards rather than sideways, the dimensions are called height (the narrow one) and width (the wide one). It is the height of the beam which plays the most important part in the playback frequency response. Accurate focussing and azimuth also play a part, but for many years there was a constant battle to get enough light to operate the photoelectric cell without backgroundnoise coming from that component. All other things being equal, the better the highfrequency response, the worse the photoelectric noise. Nowadays, solid-state photoelectric devices have noiselevels near the thermal limit and there is less of a problem. We can therefore concentrate on extending the frequencyresponse, and the penalty of noise will be due to the film itself rather than the playback mechanism. Which is the otcome expected by archivists.

The light should be focussed on the emulsion side of the film, and if the optical system is welldesigned, scratches and dirt on the clear side of the film will be defocussed and will give less output.

The slit height is typically 0.5 to 1 mm. At this sort of size, significant amounts of optical diffraction can take place, especially towards the red end of the spectrum; and blue light (or even ultraviolet light) is to be preferred if there are no disadvantages.

Azimuth is much more critical than with analogue tape. Fortunately, severe azimuth errors at the recording stage are uncommon, and conventional idiotproof projectors do not even have an azimuth adjustment. It isn’t just high frequencies which can be affected, although this might be very significant with narrower gauges. Very considerable amounts of intermodulation distortion are also generated. This usually comes as a surprise to operators used to tweaking azimuths on analogue magnetic tape; but because films are recorded to constantamplitude characteristics (as we shall see in the next section), extremely steep waveforms may be encountered. Some highpitched waveforms on variablewidth tracks may slope at an angle of 89 degrees. A mathematical description of how the distortion will occur on three types of soundtrack may be found in Ref. 4; but I have found it helpful to describe the problem to audio engineers by asking them to consider what would happen if you played a lateralcut disc with 89 degrees of tracking error!

This seems to be the place to remind operators that the equipment must have low levels of distortion itself, besides the extended frequency-response and low noise. Subjective experience backs this, particularly when the soundtrack is poor. It seems that under these conditions the ear needs all the clues it can get to perceive the recorded content, and relatively small amounts of distortion (either at high or low signal volumes) can have similar perceived effects as restricting the frequency response. Unfortunately, there is no known way of testing the overall system distortion (including that of the photoelectric device), because it is difficult to make good test films.

This section will extend concepts first introduced during the chapters on disc cutting, notably sections 6.4, 6.13 and 6.14, which I suggest you read first.

Both variabledensity and variablewidth soundtracks could be recorded by shining a beam of light through an electromechanical “light valve”, comprising lightweight but stiff metal ribbons in a magnetic field. These ribbons were arranged to have a high frequency of resonance, usually between 7 and 12 kHz. Electrically, they functioned as resistors rather than inductors, so they gave constant amplitude characteristics on the film below these frequencies. That is to say, the resulting variations in width or density were substantially the same at all frequencies for constant inputs.

This also applies to variabledensity recordings made by exposing the film to a gaseous discharge lamp driven from the audio, or a modern photoemissive diode driven in a similar manner. In the vocabulary of Chapter 5, we always end up with a constant amplitude characteristic on the film. A great deal of high-frequency noise is masked by this technique, but the disadvantages are greater risk of highfrequency distortion, and worse lowfrequency hums and thuds.

Much of the work of the sound camera engineers consisted in minimising the high-frequency distortions, including volume limiters with side-chain pre-emphasis (Chapter 10) and innumerable intermodulationdistortion tests around each developing and printing stage. We must especially respect the wishes of those engineers today, especially since we are likely to add to the difficulties if we copy to another optical medium.

Linear noise reduction techniques like preemphasis were never used on releaseprints because of the desire to make things idiot-proof. But, within the studios, intermediate stages of the final soundtrack might carry a standard preemphasis curve remarkably similar to a modern digital preemphasis curve. Its time constants were 42 and 168 microseconds, giving a 12dB step in the recorded high frequency response between about 1000 and 4000Hz. However, it seems it was always used with the pushpull system to give greater fidelity in the stages prior to the final mix. You are unlikely to be in the business of extracting sound from such a component track; but if you are, it is worth knowing that (say) original music can be recovered with enhanced fidelity this way.

Another canofworms occurs when we are using cinema sound for another purpose (e.g. domestic video). In about 1940, the SMPE (Society of Motion Picture Engineers ) introduced something called dialog equalization, to be applied to cinema film soundtracks to compensate for a number of psychoacoustic effects resulting from cinemas reproducing speech much louder than natural. (The thinking is explained very clearly in Ref. 5).

Therefore, it might well be advisable to reverse this standard (which I believe became universal) when converting cinema-film sound to another medium.

In 1944 the SMPTE Research Council proposed another standard equalisation characteristic to be used when 35mm films were printed to 16mm. This was designed to overcome the subjective woolliness of the severe highfrequency losses then inherent on 16mm optical soundtracks. A 1949 Revised Proposal flattened this equalisation somewhat, because there were less boomy loudspeakers and better H.F high frequency recording by then (Ref. 6), but at the same time dynamic compression was adopted. This latter was intended to overcome the fact that 16mm films were commonly shown with the noisy projector in the auditorium. None of these proposals became standards, but the principle remained - and still remains today - that a 16mm version of a 35mm film is very unlikely to have the same frequency and dynamic characteristics as the original.

During the days of the photo-electric cell, the reproducing equipment often contributed a significant amount of hiss of its own, even on 35mm film. It was necessary to make a trade off between having a broad optical gap to let a lot of light through, which meant concomitant equalisation, and a narrow gap which caused problems of hiss, focussing, and diffraction. The engineering journals of the time are filled with different people’s compromises about this; but fortunately everyone seems to have recognised the idiot-proof advantages of inviolate constant-amplitude characteristics. All we have to do is follow in their footsteps with our improved technology.

Today, the background hiss of photoelectric devices is less of a problem, even on narrowgauge formats, but it still isn’t good enough to allow much electronic extension of the high frequency response. All we can do is hurl as much light through the film as we can (provided we don’t overload the electronics). We must focus it on the emulsion as sharply as we can (unfortunately this is often made difficult by idiot-proof designs), and keep the hum and noise of the exciterlamp as low as possible (hum is endemic on narrowgauge formats which tend to have A.C filament lamps). We could also explore the use of shortwavelength light in conjunction with new photoelectric devices to reduce the diffraction problems (there is room for new research here).

It seems scarcely worthwhile for me to say this, but the optical alignment of the sound system should start with a high-frequency loop film, and the maximum output should be sought. Next a frequency testfilm should be run, and the equalisation adjusted to get a flat result. Intermodulationdistortion testfilms exist for checking the uniformity of illumination and the azimuth of the slit, and buzztest films for centering the beam on the track; but both these parameters may require empirical adjustment when running actual films which have been printed imperfectly.

All the electronic techniques listed in Chapter 3 could be used to clean up the results, especially with Class A pushpull tracks or other systems which in effect give two versions of the same sound. We should probably not try to expand the dynamic range (the subject of Chapter 10), except when a narrowgauge version of a 35mm original is the only surviving copy; but the techniques of thump removal mentioned in chapter 10 are valuable for dealing with some tracks with imperfect noisereduction. As an outsider, I am also vaguely surprised that the equivalent of a liquid gate is not used to cut down some of the noise of scratches.

From 1974, release-prints on 35mm or larger may be encoded Dolby A or Dolby SR for added immunity from background noise when the track is split into two for stereo. Please see sections 9.4, 9.12 and 10.5 for further details.

REFERENCES

• 1: Van Urk, A. Th: “Sound Recorder of the Philips-Miller System,” Philips Technical Review, 1936, no. 1, p. 135.
• 2: British Standard B.S.1404 “Screen luminance for the projection of 35mm film on matt screens” specified 8 to 16 foot-lamberts; and American Standards Z22.39-1944 and PH22.39-1953 “Screen Brightness for 35mm Motion Pictures” recommended 10
• foot-lamberts plus or minus 1 foot-lambert. These recommendations are relaxed (i.e. the picture may be dimmer) off the axis of directional screens, or in outdoor theatres.
• 3: John G. Frayne and Halley Wolfe, “Elements of Sound Recording” (book), New York: John Wiley & Sons and London: Chapman & Hall, 1949. Chapters 15 to 20 deal with most of the listed types.
• 4: ibid., pp. 350-357.
• 5: D. P. Loye and K. F. Morgan (Electrical Research Products Inc.), “Sound Picture Recording and Reproducing Characteristics” (paper), Journal of the Society of Motion Picture Engineers, June 1939, page 631.
• 6: John G. Frayne and Halley Wolfe, “Elements of Sound Recording” (book), New York: John Wiley & Sons and London: Chapman & Hall, 1949, pp. 560-1.

-