The Mapleson Cylinders - Program Notes
|The Mapleson Transfers|
by Tom Owen
The first order of business was to design a playback apparatus capable of accommodating cylinders of all types, that imposed no mechanical vibration, offered light-weight tracking, accommodated speeds from 40 to 200 rpm, could be servo-driven, had digital mandrel readout in rpm, and could track vertically with no side angling at 100, 150 and 200 lpi (lines per inch). Obviously, this was no small order. Art Shifrin, a restoration engineer, developed a machine which accomplished most of these goals, and with further modifications at Rodgers and Hammerstein, it was able to fill our needs.
I made a search of the existing literature on the analysis and playback technology of acoustical recordings--an area that proved to be under-investigated. (This was before the Audio Engineering Society published their index of publications.) However, an article by Hans Meulengracht-Madsen, "The Transcription of Old Phonograph Wax Records" ( AES Journal, XXIV/1 ) describes an extremely thorough acoustical analysis methodology, and should be required reading for anyone working with this technology. It is based on cylinders recorded from 1919 to 1923, 23 years after Mapleson did his recordings. My associate John Fesler and I knew that in order to do viable acoustical analysis, the cylinders would have to be recorded methodically, under calibrated conditions, on equipment similar to that used by Mapleson. This was not difficult, since many Edison Home Model A's are still to be found in good working condition. We will discuss the recordings and their setup in a moment, but first let me discuss the acoustical concerns.
One of the most important articles to date concerning acoustical analysis is that of Maxfield and Harrison (in the Journal of the American Institute of Electrical Engineering, XLV [February 1926], 334-48), in which they relate acoustical energy to electrical energy, and describe what was later known as the "theory of matched impedance." In fact, they were describing the pre- and post- (record and playback) equalization, or pre-emphasis/de-emphasis that is still the manner in which records are cut and played today.
To ascertain some of the general characteristics of cylinders, I collected samples of every type of cylinder I could find from our archives, from collectors, and from other sources: a representative batch of brown waxes, Amberols, Blue Amberols, Concerts, Columbias and others. These were sent to the late Ed Catalano, chief chemist for the Gamma Omega Association. He reported that the surface deterioration of the cylinders was caused more by internal stress then by mold and mildew. Photomicrography was done on cylinder surfaces, and several samples were broken to enable photography of the various layers. A test washing of a moldy cylinder surface produced the results shown in Figure 1. Despite the frequent statements in the literature about Maplesons that the extreme surface noise is due to fungus attack, they do not in fact suffer from excessive fungus attack. The fact is that by 1937 Lionel Mapleson and his colleagues had played some of the cylinders to a physical point of no return.
I then sent our sample cylinders to George Alexandrovich of Stanton Magnetics, and asked him to evaluate the various cuts with the scanning electron microscope. He took wax impressions of the grooves and measured the different cutting parameters such as depth, width, and sidewall deformation characteristics. This was extremely important for determining tracking parameters, stylus size and shape, vertical/lateral phasing problems, and the like. In general, we found that the processed molded cylinders tracked at approximately 2-7 grams, but Blue Amberols were sturdier and could be tracked more heavily, if necessary. The brown waxes (the type used by Mapleson), on a cleanly shaved, freshly cut surface that had been warmed and prepared, were cut at an 8 mil diameter to a very shallow depth at 100 lpi. The commercially produced cylinders we evaluated fell within the same depth parameters for the most part, but were played back with a stylus in the 7.4 mil tip radius for the two-minute (100 lpi) type, and 4.2 mil for the four-minute (200 lpi) type (see Figure 2). The surface content varied
according to the material composition. On some of the unmodulated-groove brown wax cylinders that John Fesler cut for me, surface content was measured (with a BaDap and HP spectrum analyzer) at levels as low as 3 to 8 dB. Black waxes measure at least double that; unmodulated grooves cut on a shellac disc are triple.
As far as the level in recording originals, the horn and diaphragm like to see an spl (sound pressure level) of between 90 and 105 dB. When you consider that Mapleson was 40 feet away from his sound sources, it's amazing that he was able to get any modulation at all, even with his large horn. It would seem clear that the singers and orchestras of his day must have performed with a great deal of gusto.
With an spl of over 105 dB, the radius of curvature of the wound track modulation becomes equal to, or smaller than, the radius of replay stylus, and a curvature overload occurs as tracing distortion where high levels are coupled with signal frequencies. In the case of the Maplesons, the track configuration was seriously degraded by constant and improper playing, thereby nullifying any standardization applicable to the lot. We decided (mainly because of comparative performance considerations) to transfer all the cylinders at one speed: 184 rpm. A few more were transferred at 120 and 160, but most ran from 184 and up (the old machines had variable speed for pitching, but all tracked at 100 lpi via a feed screw mechanism). Tracing distortion, surface cracks, internal stress cracks, out-of-roundness, groove deformities, different signal-to-noise ratios, and different speeds rendered each cylinder a project in itself.
Probably the easiest way to describe the Mapleson transfers is to describe a typical work day. David Hall and I would take one of the seven original boxes the Maplesons were packed in, and work on it until the box was completed. The cylinders were packed randomly in the seven boxes; there was no program or performance order. To minimize handling, we took them as they came out of the box, often not knowing what was on the cylinder until we played it. Prior to playback, while still in their original containers, the cylinders were allowed to warm up under a 40-watt soft light. Each cylinder was examined under the microscope, and groove size and description noted. If cotton fibers remained on the cylinder from the box linings, they were blown off with an air gun. A treatment of LAST (Liquid Archival Sound Treatment), prepared especially for the cylinders by Ed Catalano, was applied if needed. The cylinders would then be placed snugly on the mandrel, but not forced.
The reproducer comprises a modified Rabco SL8E tangential tonearm, capable of tracking 100 to 200+ lpi, equipped with a Stanton 500-AL cartridge. The tonearm is driven by the groove, except when a locked or deformed groove is encountered, at which times a battery powered servo motor takes over. The mandrel is solid brass, approximately the same size as an original Edison mandrel, driven by either an Ampex two-speed hysteresis motor or a servo motor, and is mounted off the body of the reproducer, thereby eliminating any mechanical vibration. The motor is belted via a pulley (gated for the standard cylinder speeds) to a very heavy brass wheel, with the mandrel attached to it (see Figure 3).
After the correct stylus was fitted (often, many would be tried), the vertical and lateral components of the signal would be monitored. The cartridge was wired to a Mallory switch, then to a transformer-based Stanton 210 preamplifier in the linear mode wired to a patch bay. This setup allowed monitoring of the left, right, lateral, or vertically modulated groove. (One soon learns that those who insist there is no lateral modulation in a cylinder groove simply aren't listening.) The signal arrived at the patch bay at a level of--20 VU. It was then elevated by U.R.E.I. LA-3A amplifiers. Level was adjusted to peak at a maximum cylinder modulation of 0 to +3 dB, referred to NAB standards.
The next step in the chain was the Packburn Transient Noise Suppressor. Space does not permit an in-depth discussion of this device (for a detailed account, see the review in Stereophile, V/8 [October 1982], 8ff.), the chief contribution of which is to eliminate tricks, pops, cracks and other transient noises. When the signal left the Packburn, it arrived at the Owl 1 Restoration Module, used here as a low-end (20-60 Hz) and notch filter. The thump and rumble associated with most cylinders is either mechanically induced by the original machine or occurs as a result of out-of-round cylinders. This is easy to see with a spectrum analyzer. After the thumping was eliminated, the ticks and pops removed, the signal then went into a pair of U.R.E.I. 565 "Little Dippers." These filters were usually set as low/high-pass filters with an 18 dB per octave slope.
Throughout this whole process each step of the way was monitored on the BaDap and the Leader oscilloscope, checking resonances, polarity, etc. Unusual problems, as well as the final stylus selection, were noted by Mr. Hall. In all these steps, no boost or curve equalization of any kind was added or imposed. At this point, the cylinder, usually reproduced at 184 rpm, would be recorded onto 1.5 mil magnetic tape at 15 ips, on an Otari MTR 10-2. All tapings of each cylinder were verbally slated. Each cylinder was transferred completely flat, then with filtering, and followed by all previous 78 rpm and LP transfers, giving us the recorded history all in one place. The tapes were then collated alphabetically, opera by opera, and pitched to the printed music. Every effort was made to achieve accurate identification of casts and performance dates. We have compiled and edited a history of the Mapleson cylinders, with complete technical information, which is now available at the reference desk of the Rodgers and Hammerstein Archives. Analysis of some cylinders has been recorded onto video cassette for future examination. The initial transfer process began in mid-May 1981 and lasted until late June.
After completion of the transfer process onto thirty-one reels of tape, cassette copies were made for exact identification and pitching, following which a pitch-corrected working master was prepared, edited into alphabetical sequence by opera title and generally in performance order within each opera. By the end of 1981, a research tape was available for on-premises audition at the Rodgers and Hammerstein Archives.
When David Hall retired as Curator of Rodgers and Hammerstein in June of 1983, he passed to me the responsibility of publishing the disc edition. (My background, prior to working at Rodgers and Hammerstein, had been in the record business.) Funds to cover the costs were on hand at the Archives from the Rhode Island Corporation. David Hamilton and I have worked together over the past several years collaborating on the Metropolitan Opera Historic Broadcast Recordings, so I asked him to be co-producer.
The processing for the disc edition of the master tapes from the original transfers done with David Hall in 1981 took approximately four weeks. In most cases the "flat transfer" from the original tape was used, with some additional filtering through the Owl 1. Pitch was constantly checked with a Casio VL-10 synthesizer, tuned to standard A = 440 Hz (based on a study of the literature on pitch around the turn of the century). The album master tapes were recorded on a half-track tape recorder at 15 ips, using Ampex 406 tape. The lacquers were cut at Tru-Tone, Inc. on a Neuman lathe under my supervision, with plating and pressing by Europadisk in New York City.
I am often asked about digital restoration of the Maplesons--whether
In many cases the result is not at all what you would expect. I attended a demonstration in Washington, D.C., of a digital restoration of the famous cylinder of the Emperor Franz Josef. Although all the noise of the recording was gone, so was the sound of the human voice as we are used to hearing it. The human voice is a complicated instrument, with harmonics, overtones, timbre, vibrato, and many other characteristics. If any or all of these elements are deformed, whether digitally or otherwise, the sound is seriously impaired. (See B. A. Blesser, "Digitization of Audio: A Comparative Evaluation of Theory, Implementation, and Current Practice," AES Journal, XXVI/10 , 739-771.) The "ambience" or hall sound of the old Metropolitan Opera House stage is present in the Mapleson recordings, and is preserved in these transfers, whereas in previous issues it was not to be heard. In the present state of the technology, digital filtering for this type of material does not offer any miraculous cures. However, I am seriously investigating such technologies as digital delta modulation recording and filtering, voice identification Spectrographic analysis, and other methodologies that might serve as a yardstick in judging the results.