Go back to article: The Art and Science of Acoustic Recording: Re-enacting Arthur Nikisch and the Berlin Philharmonic Orchestra’s landmark 1913 recording of Beethoven’s Fifth Symphony
Findings from the perspective of sound engineering: preparation and processes of acoustic recording on disc
Surviving documents, photographs and first hand accounts relating to acoustic studio practices tell us little about the precise methods of recording, which were shrouded in secrecy and had been largely developed through tacit knowledge. Formal portraits were taken of artists and experts alongside the recording horns, but the recording apparatus, the ‘business end’ of the horn, remained hidden in a closed booth or out of shot. In order to fully understand how an orchestral recording could have been made acoustically, it is therefore necessary to reverse engineer the process of acoustic recording, using the same apparatus or constructing replicas of it, and to recreate the recording media itself – the recordable wax discs.
The following section examines the processes, technology and media of acoustic recording from the perspective of the project’s sound engineer who, in addition to making the recordings, was also responsible for the manufacture of the blank wax discs and the construction of the replica recording horns, disc recording lathe and recorders.
The primary component parts of the recording apparatus and media are discussed individually and their performance assessed, beginning with the manufacture of the all-important wax discs for recording. This is followed by a description of the recording lathe, which consists of a powerful turntable on which the blank discs are rotated and an advancing mechanism (or feed screw system) which in this particular design propels the turntable underneath the stationary recorder and also regulates the ‘pitch’ (the number of grooves per inch on the record). The sizes and shapes of recording horns that were used to collect the sounds to be recorded are described, followed by an examination of the recorder that is connected to the horn and receives the sounds. The recorder consists of a sealed chamber that houses a circular diaphragm, typically made of mica or glass. The sound energy received from the horn vibrates the diaphragm and these vibrations are transmitted to the stylus that cuts the spiral groove on the disc via a linkage from the centre of the diaphragm.
The section ends with an account of how some of the recorded wax discs were processed and duplicates made from them that could be played back on a mechanical gramophone.
Wax disc preparation
The recording medium was of fundamental importance to the project, since the re-enactment would be assessed by the sound reproduced from the discs. Therefore, the first major technical task was to reproduce the wax disc blanks that were used for recording in 1913. Wax recording is a lost art and careful research and practical experimentation were required to replicate as closely as possible the discs of the time. Since most of the formulae for the waxes were commercial secrets and the actual wax blanks were an industrial intermediate, access to original recording wax is very limited. The material would have been recycled when in use and then scrapped, either when the recording company ceased to trade or when it went over to lacquer/nitrocellulose masters. A chemical analysis of any original wax will indicate some of the final components but not the method of manufacture or the exact raw materials used to produce them. Very often the quoted recipe in published material is wrong or misleading, or the materials have changed and current versions of them are produced from different sources or by 'better' methods that remove essential components now considered as impurities. In fact these ‘impurities’ often imparted vital qualities to the finished recording blank.
Blended waxes versus soap waxes
Contemporary descriptions of wax masters hint at the material being a blend of waxes, and certainly such a blend would be a quick and desirable way to make a blank. Blended wax would also allow for easy adjustments in the cutting characteristics of the disc. However, a search of the literature for this type of wax produces few credible references (Barnes, 1936). In The Reproduction of Sound (1916, p 37) Henry Seymour alludes to a mixed non-soap wax made in Germany. However, the current author’s many experiments with both original and modern waxes have failed to produce a blended material that comes close to a usable wax in this way. Blends tried included various proportions of montan, carnauba, paraffin wax, and petroleum jelly. In most case the waxes are not totally miscible and separate on cooling, giving crystalline patches, and they produce a material which is either too soft or too brittle and always inherently noisy when cut – if they can be cast into blanks without splitting on cooling. H Courtney Bryson does not mention this material in his 1935 book Gramophone Record. He does suggest a mixed wax, but this uses pre-made soap components as well as waxes. Preparation of this blank recording material becomes particularly difficult when trying to define the actual composition of the aluminium oleate, a key ingredient which, according to Bryson, is a brown jelly-like material. However, when recently prepared from pure ingredients, it precipitated as a white putty-like substance. This material was at one time an item of commerce and used in some grease formulations and it is difficult to reproduce a modern analogue.
A small sample of later Gramophone Company wax, analysed by chemists at Kings College London, shows that this was a soap wax, containing a wide range of other hydrocarbon compounds. This suggests at least one of the components was a complex natural or fossil wax, but no conclusion could be reached and without access to the original ingredients it would be difficult to replicate the wax from the chemical analysis alone. These waxes often contain montan wax as a saponified or non-saponified ingredient (and sometimes as both), the major saponified component being stearic acid. The chemical specification of montan, stearic acid and some other ingredients for these waxes has changed in important but subtle ways since the 1920s. Many of these extraction and production changes have happened since the 1960s and some of the materials have become commercially unavailable.
Internal documentation by the Columbia Graphophone Company in 1931 gives a specific process and recipe. However, duplication of the 1931 Columbia recipe with modern stearic acid (several grades were tried) and montan waxes (several types were tried) produces a material that is far too hard compared with the requirements of the disc cutting process and also compared with the original wax sample. Disc blanks require 2.5 to 3 kg of wax per blank and around 25 blanks were needed for tests and takes. A saponified wax with as close to the characteristics required as possible was developed and tested. The material could be compounded in one of two forms – a basic wax which would need to be used at a higher temperature, or a modified version tempered with a softening wax that could be used at room temperature. At the start of wax production it was not certain that a warming cabinet for the wax would be available for the re-enactment, so the tempered wax had to be used. The tempering wax can, at certain percentages, increase the cutter noise.
The objective in producing a suitable recording wax was to match the cutter noise and thus the possible signal-to-noise ratio of the original waxes. When analysed, the noise of the historical wax was 30dB lower compared with a 0dB signal level. After numerous test formulae, a recording wax was produced whose noise measured approximately 33dB lower than the signal. Analysis showed the newly made wax to be entirely consistent in quality to those that would have been in daily use in 1913.
The wax was made in 20kg batches, cast into moulds, cooled, drilled, trimmed on the outside on one machine and then rough shaved on another. Once this was done, a second shaving machine could be fitted with a sapphire blade and the final surface produced. From raw material to final blank, each recording surface required from three to four hours of work, although after use the discs could be shaved for reuse.
© Aleks Kolkowski
Wax disc on shaving apparatus
Session results regarding wax
The wax behaved consistently and it was possible to warm the blanks and shave them on site during the sessions. Fully warming the blanks took several hours. The warming cabinet had a capacity for six blanks so the number that could be made ready at any time was limited. Maintaining the surface temperature could have been achieved by adding an overhead heat lamp to the lathe but this was not available so it was essential that recording commenced as a soon as the blank was on the turntable.
The playback of the waxes on the modern turntable still affected the soft groove to some extent and the early tests especially had a poor signal-to-noise ratio. This turns out in hindsight to have been due to the low level of recording rather than an especially high level of cutter noise. Noise and imperfections in blank wax records were regular problems in acoustic recording, with contemporary reports of entire sessions being abandoned and recording masters and the negative metal copies produced from them frequently rejected (Batten, 1956, p 35).
The specially constructed acoustic recording lathe was designed to use the traversing turntable system. This was especially important where a large recording horn or a group of horns might be needed, as the movement of the horn(s) would put strain on the lead screw’s driving unit and would also cause the horns to point at a different angle at the start and end of the recording, moving from right to left. This can change the balance of the recording.
The turntable was belt-driven and the belt stretched over the entire circumference of the platter. Similar belt-driven turntables on recording lathes were used by the Columbia Company in the 1920s. These have the advantage of transmitting no noise to the turntable shaft, which can happen with a bevel or similar gear drive. An electric motor with a high quality speed control was used in preference to the weight-driven motor that would have been originally used as the weight motor would have been too expensive to build and would have added unnecessary bulk to the equipment. The lead screw was driven by a suitable motor giving the possibility of a very wide range of feed pitches as needed.
© Aleks Kolkowski
The disc recording lathe in operation
The machine was designed to be mounted on a stand that would bring the horn up to a working height – normally head level for a standing vocalist. The transport system of the lathe performed consistently throughout the sessions.
While not revealing the interior of the recording booths, historical photographs do show the recording horn, sometimes with one or two additional horns fixed to it, the narrow end of which is attached to a recording soundbox via rubber tubing. ‘Couplings’ or tubing that connected auxiliary horns to the main recording horn would have a narrower bore that attenuated the incoming sound of instruments, enabling a soloist, for example, recording through the main horn to remain in the foreground of the recording. Multiple horns, their couplings and the positioning of musicians in relation to the horns were the means of sound balancing or ‘mixing’ during the acoustic era.
Only a single horn is visible in the 1913 photograph of the BPO in the recording studio, but additional horns used in the sessions may have been removed for the photograph, to allow room for the camera equipment or to avoid blocking its sightline. Using multiple horns for recording has the advantage of isolating and balancing groups of musicians and is ideal for recording soloists with accompaniment, but a disadvantage in that it reduces pressure at the recorder diaphragm, resulting in a lower level of recording. For the recording experiment at the RCM, it was decided that a single recording horn should be used throughout and no experiments with multiple horns were attempted.
It should be noted that most of the sessions that exist in the photographic record, where two or more horns were used, normally consisted of small orchestra arrangements where the principal players were positioned close to the mouths of the horns. There may be two main reasons for this approach; one would have been to reduce the expense in musicians’ fees and the other would have been the typical size of the studios at the time. As a result of this project, it has become apparent that increasing the number of players may well increase the orchestral quality of the sound but moves many of the quieter instruments much further from the recording horn, giving a reduction rather than an increase in overall recorded level.
© EMI Archive Trust
A studio layout for recording a vocalist with ‘orchestral’ accompaniment. Note that the violins are represented as horns (to signify Stroh instruments). Fred Gaisberg: Report from the Victor Talking Machine Studios, 1907
The taper of gramophone recording horns were generally 1:4 so that a forty-inch long horn finishes at ten-inch diameter. Four such horns were specially fabricated for testing during the RCM recording sessions: two 40-inch (length) x 10-inch (diameter) horns made from sheet zinc (0.7mm thickness); a slightly larger straight horn made from galvanised sheet steel (0.7mm thickness); and a flared horn made from the same material.
Session results regarding horns
Both the straight and flared horns were tried without changing the initial set-up of the orchestra, the flared horn providing significantly better results. Between sessions it was possible to do a voice test to establish the areas of the room that were effectively covered by the horn for the purposes of recording. With the flared horn the sensitive workable area corresponded directly with the extension of a cone of the same angle as the main part of the horn. A secondary, less sensitive working area was represented, roughly, by the extension of the cone flare angle. Areas outside these two cones were effectively dead, meaning instruments outside this area hardly registered at all. The effect was far more evident than expected and the floor was marked with tape to show the workable areas, thus allowing for rearrangement of the orchestra.
At the start of this project, two duplicate Gramophone Company-style recorders were available. These had been made in 1997 to facilitate a historic recording celebrating the centenary of HMV, the oldest company within the EMI group, and had been used to successfully cut both lacquer and wax masters of tenor vocals with piano accompaniment. The recorders were replicas of the so-called type ‘R’ recorders, and once the rubber gaskets were replaced to ensure an air tight seal with the correct amount of resilience around the edge of the diaphragm, the performance of the recorders’ voice tests was found to be consistent with the original set-up as used in 1997. According to documents at the EMI Archive, this style of recorder was also in use in the 1907–08 period for military bands and small orchestras and they were still in use up until 1924. There is no direct evidence that these standard-type recorders were the ones used in Berlin in 1913, as no surviving documentation exists from the Deutsche Grammophon Berlin studio, but it is a reasonable assumption. The diaphragms used were mica, with a diameter of 42mm, and a thickness of 6.5 1/000 inch for one, and 7 1/000 for the other. Flat rubber gaskets were used for both. The author’s previous experience showed that diaphragms thinner than this will produce a ‘tubby’ sound with reduced top frequency response and that thicker diaphragms will considerably reduce the sensitivity and low frequency response.
There was a possibility that for orchestral use a larger diaphragm would be more suitable, and another type of recorder was made specifically for the RCM re-enactment with the capacity to hold a 50mm diameter mica diaphragm. Both types of recorder were tested during the first day of the re-enactment to determine which would most closely approach the quality of the original recording.
© Aleks Kolkowski
Recorder used during the RCM sessions
Session results regarding recorders
Two things that affect the recording process are the freedom of movement of the recorder up and down, and the air-tightness of the seal on the rear of the recorder box. When tested, the two type ‘R’ recorders were insensitive compared with the new recorder and because of this the main recordings were carried out with the 50mm diaphragm unit.
One of the 48mm units was fitted with an aluminium diaphragm but in a test this increased cutter noise and was very resonant in response. In fact, a flat response is the desired result, as a resonant response causes certain frequencies to record much louder than others, giving an unnatural sound to the finished recording.
The new 50mm recorder did present some problems with the slight rise-and-fall that can occur in each revolution of a recording blank. This causes what is termed 'sweep noise' in the recording, a cyclic rather than constant background hiss. It was possible to improve this with some on-site adjustment, although a certain amount of re-working, with access to workshop machinery, would have been desirable to eliminate the effect. As this is a type of recorder that floats on the wax, its inertia and that of its trunnion need to be as low as possible. Simply offsetting the mass with a counter-balance does not automatically solve the problem. Nevertheless it was the better option for the recording.
Time was the limiting factor as the substitution of diaphragms into the recorder boxes would take around 15 minutes each time plus the recording test. Glass diaphragms were not at this stage an option for this re-enactment as they have to be specially made to size and thickness. Mica is practical because it can be made to a specific thickness and cut to size with ordinary workshop tools.
Processing the recordings as 78rpm discs
During the recording sessions it was decided to make a limited number of recordings that would not be played back in any way on the day but would be turned into discs. This would give the opportunity to play copies on a gramophone, thus replicating the full cycle of processing of the original 1913 recording. To this end the wax was tested in advance so it was known to be proof against the plating-bath solution allowing for the possibility of subsequent processing.
To create a more permanent disc after the recording session a few of the wax masters were treated with graphite to render them conductive and then processed in a copper plating bath to deposit a layer of copper around 0.75mm thick. This process takes around thirty hours. Once the wax was separated from the copper shell and the surface cleaned, it was possible to use the shell to produce accurate resin mouldings. The mouldings produced are similar in hardness to a shellac pressing and with the right type of needle it is possible to play the recording back on an acoustic gramophone and get a good idea of how it compares with an original 1913 recording played back on the same equipment. This is important as the mechanical playback has a particular natural filtering characteristic. In addition, a copper shell is an archival item in comparison with a wax blank, which tends to deteriorate in storage.
Resin mouldings of the RCM re-enactment were produced and replayed successfully. In this way the entire acoustic recording and production processes were replicated.
© Duncan Miller
Left: Preparation for electroforming; Right: Part-processed copper shell on wax master
© Aleks Kolkowski
Finished 78rpm duplicate disc
© Aleks Kolkowski
Beethoven’s Fifth Symphony – Allegro (side one), Robin O’Neill and the RCM Chamber Orchestra on a moulded resin 78rpm 10” disc
Assessment of the results
All the primary components of the acoustic recording system that were constructed for the re-enactment performed well throughout the testing and recording sessions and were analogous to those used in the acoustic recording studios from the early 1900s.
From a technical point of view, the sessions at the RCM have clearly demonstrated that the original recordings by the BPO from 1913 under both Hertz and Nikisch were pushing at the very limits of the acoustic recording system of the time, in terms of what it was possible to capture sonically using recording horns. The larger the group of musicians in the studio, the more difficult it becomes to balance the instrumental sections and maintain a satisfactory overall recording level as many of the musicians are placed further away from the recording horn. Even the later orchestral acoustic recordings from the 1920s, when compared with earlier military band versions of the same piece and even from the same studio, are generally recorded at a significantly lower level. To get the recording level up and keep the noise level down requires everything to be just right; every single element has to be optimised.
By the end of the sessions at the RCM, we had managed to increase the level of recording and improved the instrumental balance. While the recording levels were at best 10dB lower than that of the Deutsche Grammophon 1913 recording, the technology and media were refined as closely as possible to match the historic antecedents. Importantly, in the process of experimentation, the factors that would in all likelihood have allowed the recordings to reach the historic levels were identified. The experiment has provided valuable information for future research into the practices of acoustic recording.
Component DOI: http://dx.doi.org/10.15180/150302/004