Best Practices
The International Association of Sound and Audiovisual Archives (IASA) lays out the following guidelines for signal extraction from original sources:
- Selection of Best Copy
The selection for use in digitisation usually follows a previous stage where all available apparently identical copies of a particular commercial record are collected in one place, from which one or several copies are selected. An archive may choose to retain a number of identical copies. In selecting the best copies for digitisation, co-operation with other collections should be considered. Selection is primarily made visually, for speed and to prevent wear. The staff should be well versed in the codes and identifiers used by the various record companies and usually placed just outside the label. This will help to discover alternative takes, i.e. non-identical recordings.
- Cleaning and Carrier Restoration
Recordings not stored ideally since their manufacture 40-100 years ago may have suffered damage or accelerated decay. Unfortunate combinations of temperature and humidity as well as the storage container surface material (sleeves, cylinder boxes) touching the recording may cause anything from mass craquelure of the surface to mouldy spots (fungus). In many cases instantaneous discs suffer from shrinkage of the recording layer, causing huge cracks or partial loss of surface.
Cleaning depends on the time and apparatus available. The intention is to remove dust and deposits from the grooves while not affecting the surface itself. The effect of cleaning should always be checked. Some carriers such as the gelatine instantaneous disc may be soluble in water, cleaning others in water, such as pressings with absorbent fillers, may cause an increase in replay noise. In such cases dry wiping with a clean micro-fibre cloth in a circular movement is preferable to no cleaning.
A standard piece of equipment is a Keith-Monks record cleaner, which cleans one side unaided. A manual process may comprise application of demineralised water (blood temperature) with an addition of a non-ionic wetting agent (known from photographic processing). The application should be by a stiff, thin-fibred brush (max. 30 mm wide and fibre length 10 mm), wiping in a circular movement (as opposed to a radial movement) while protecting the label against any drops of water. Intermediate rinsing may use tap water, if the process needs repeating, but the last rinse must use the solution. Wiping with a squeegee in a circular movement coupled with evaporation, or wiping with a non-shedding cloth may achieve manual drying.
Cleaning will remove dust particles, some tar residue from tobacco smoking, and organic matter related to fungus attacks, however the digested and damaged parts of the surface will remain rough.
- Replay Equipment
The least invasive equipment is the best for the job; Optical replay is increasingly becoming feasible, however at present it is perceived as more elaborate than the use of pickups. When using pickups, the stylus dimensions of the pickup have to be suited to the task - a range of widths will enable riding high or low in a lateral groove thereby avoiding the worst surface roughness of the grooves.
- Speed
The recording speed of historical mechanical and other obsolete formats may vary considerably from agreed standards or normal practice for many reasons, not least being that standards were not widely accepted, or that early mechanical drive motors could not be locked to an external standard such as the mains supply. Determining the correct reproduction speed for a carrier may be time-consuming. For commercial records and cylinders one may assume that a given speed would be valid for a whole recording session, however cases are known where the speed varies across the side of a record. Speed is determined by a mixture of experience, knowledge, and analysis of the recorded signal characteristics. Consequently any speed changes are, to some extent, subjective and must be carefully and accurately documented and retained in the metadata.
- Replay equalization
Either by nature (all acoustic recordings) or by design, sound pressure levels at different frequencies do not result in directly proportional variations in the amplitude of the recorded signal. Rather it is the instant velocity that is proportional, which indicates that the amplitude is reduced at 6 dB/octave increase in frequency. Early research in this field served as a basis for playback characteristics developed in subsequent years (Maxfield and Harrison 1926)
When transferring a disc it is necessary to choose between transferring audio, equalised to the standard that may be achieved today, or transferring total waveform from the groove unequalised. In both cases it is necessary that the pickup be able to give a non-distorted signal representative of the groove.Where the curve is accurately known equalisation may be applied either at the preamplifier prior to making the copy or applied digitally after making a flat copy. In this case it may generally be said that equalisation before digitisation is a dynamic range issue, in that a flat recording can require more than 20 dB of extra headroom when compared to that of an equalised copy of the same sound recording. The noise floor of the encoding system and format is therefore 20 dB poorer if recorded without equalisation, however, with a high bit depth (24 bit as recommended in section 2.3 Bit Depth) there is sufficient dynamic 'range to convert the raw, unequalised signal from any groove into digital form without compromising the overall Signal to noise.
Some modern noise reduction techniques depend on a simple model for the noise, and heavy filtering may impair their functioning and in this case the equalisation should be performed after noise reduction. Furthermore, some digital noise removal techniques experience greater problems when dealing with clicks and crackles obtained from an equalised replay. In other cases the predictive model works better on equalised audio. In acoustic records, where equalisation of the frequency response at recording will generate noise bands at discrete frequencies, noise reduction models may work better on unequalised audio because narrow noise bands are far more difficult for the system to model.
- Corrections for Errors Caused by Misaligned Recording Equipment
The worst misalignment in mechanical recording is rotation of the cutting stylus about its longitudinal axis. This was deliberately performed in instantaneous recording in which it was desired to collect the swarf (chips) by means of a co-rotating central brush. The result was a time delay between the two groove flanks when intersected by a radius.
- Calibration Discs
Calibration means knowing the output when a defined input is provided, for instance over a range of frequencies.A pre-amplifier/equalizer may be calibrated by supplying the input with a constant signal of variable frequency while loaded with the correct impedance, and the measurement consists in plotting (or data-logging) the output against frequency. Automatic apparatus exists for this. In practice, the input comes from a pickup cartridge, at transducer that converts a mechanical input to electrical output, and for this we need a mechanical calibrating signal. When mechanical recordings were commercially available Test Discs were produced for this purpose. It is envisaged that in the future, pickups may have to be calibrated by means of controlled vibrators, however none dedicated to this purpose exist commercially at the present. The Audio Engineering Society (AES),via its Standardisation Committee, runs an ongoing and active project of developing and publishing a series of simple Test Discs, both for coarse groove work and for microgroove.
If the calibration by means of a Test Disc has been performed with sufficient resolution, the plotted curve may be regarded as a plot of the transfer function of the pickup or the pickup- preamplifier- equalizer combination. Apart from the fact that visual inspection of the curve will tell the operator of gross deficiencies, it may actually form the basis of a digital filter that may filter the digitised signal from the mechanical record, so that it becomes independent of the actual pickup (and pre-amplifier and equalizer) used. All it takes is to be certain that no adjustment has been changed between using the Test Disc and the mechanical record to be transferred (and ideally that the record materials for those two inputs behave the same way). (For further discussion see Brock-Nannestad 2000).
- Time Factor
A complex transfer may easily take 20 hours for 3 minutes of sound (a ratio of 400: I). An average transfer may take 45 minutes for 3 minutes of sound (a ratio of 15: I), which represents time spent on finding the correct settings for the equipment and choice of stylus, based on an analysis of the recording as it relates to others of its time and storage history. Some experienced archives suggest that, for the transfer of unbroken cylinders in average condition, two technical staff, (one expert and one assistant) can transfer 100 cylinders per week (a ratio of about 16: I). Obviously experience will improve both the ratio and the ability to estimate time required.
