The United Kingdom patent 394325 'Improvements in and relating to Sound-transmission, Sound-recording and Sound-reproducing Systems' is a fundamental work on stereophonic sound, written by Alan Blumlein in 1931 and published in 1933. The work exists only in the form of a patent and two accompanying memos addressed to Isaac Shoenberg. The text is exceptionally long for a patent of the period, having 70 numbered claims. It contains a brief summary of sound localization theory, a roadmap for introduction of surround sound in sound film and recording industry, and a description of Blumlein's inventions related to stereophony, notably the matrix processing of stereo signals, the Blumlein stereo microphone and the 45/45 mechanical recording system.
In 1933–1935 Blumlein built experimental stereo recording equipment and recorded two sets of stereo recordings using mechanical and optical media. Commercial implementation of his invention became a reality in the late 1950s, when the patent had expired. Blumlein's 45/45 system became a worldwide standard for stereo LP records, and Blumlein himself was proclaimed "the inventor of stereo".
Background
In 1881 Clément Ader presented the Théâtrophone – a working system for live delivery of opera performances over telephone lines. Also reprinted in The théâtrophone was a one-to-one network, employing one carbon microphone to energize one remote telephone receiver, and required rows of microphones placed along the stage. While experimenting with parallel lines, Ader accidentally discovered stereo effect. By placing pairs of microphones at either side of the stage, near the footlights, Ader achieved strong binaural sound localization, simulating the effect of sitting at the edge of the stage, hearing actors and instruments as if they were spread in front of the listener. Ader himself explained the effect as the result of the differences in apparent loudness registered by the ears of the listener. Binaural théâtrophone, advertized as "auro-stereroscopic" or "binaural audition" failed to attract customers due to the need to have two telephone lines per subscriber, and overall low fidelity. Conventional, monaural théâtrophone operated successfully in France, Hungary, Italy and the United Kingdom until the end of the 1920s.
In the 1900s Lord Rayleigh formulated the scientific sound localization theory. In Rayleigh's model, human hearing localizes low-frequency sounds based on phase difference between the signals registered by left and right ears (interaural time difference, ITD); high-frequency sounds are localized based on relative loudness of two signals (interaural level difference, ILD). Rayleigh's duplex (two-factor) model remains valid in the 21st century, with the addition of a third mechanism, the analysis of spectral cues provided by mechanical filtering of incoming soundwaves by human torso, head and pinna.
During World War I acoustic location was actively researched for military applications of air defense, artillery sound ranging and naval hydroacoustics. Valve amplifiers that emerged at about the same time allowed reproduction of sound via loudspeakers. Early experiments with amplified reproduction of binaural signals ended in failure: binaural effect that was easily reproduced via stereo headphones was either weakened or completely absent.
In the end of the 1920s researchers of American and British corporations approached the amplified stereo problem; basic strategy for solving it had taken shape in the early 1930s. Stereophony could not bring immediate financial gains: the main potential customer, cinema, was content with crude monaural sound film equipment; the Great Depression ruled out investments in new sound systems. However, the corporations were eagerly accumulating patent portfolios in anticipation of economic recovery, and continued financing research. Arthur C. Keller and Harvey Fletcher of Bell Labs and Alan Blumlein of Columbia Graphophone Company and EMI were the first to obtain practical results. Each of the three inventors had a different objective, and followed a different course of research.
Fletcher followed Bell Labs strategy for the improvement of traditional telephony. He concentrated on transmission of sound field images of the original for binaural reproduction via headphones. His experimental equipment successfully recreated high-quality spatial sonic imagery, but like all binaural systems did not work well with loudspeakers.
Keller was primarily interested in amplified public address sound; he placed arrays of microphones on the stage and transmitted parallel audio signals to arrays of amplified loudspeakers in a remote listening hall, trying to capture and recreate the original "sound front". Best results were obtained with expensive two-dimensional arrays, capable of recreating both width and depth of the original; linear (one-dimensional) arrays could produce almost perfect sense of width, but not depth. The minimal working configuration required three channels (left, centre and right). It was adequate for recreating width, and even a limited sense of depth, but only for the listeners sitting close to the axis of the centre channel, and too expensive for the consumer market. A cheaper two-channel stereo setup could not reproduce the sound field; the sound inevitably broke up into left and right point sources with a "hole in the middle".
Blumlein envisaged introduction of surround sound in the film industry. He rejected the binaural model from the start. Instead of recreating spatial sound fields or "sound front" radiated by the orchestra, he settled on recreating the sound that is heard by a listener sitting in a concert hall, or by a camera operator on a film set. He reasoned that the microphone array should mimic human hearing apparatus, thus the two microphones must be placed close together (and close to the film camera). The resulting binaural signals cannot be used directly in an amplified stereo setup. However, wrote Blumlein, a two-channel recording with carefully altered phase and level differences can deceive the listener with a lifelike spatial illusion.
Publication
Very little is known about Blumlein's work on stereo prior to filing the patent application. According to biographer Robert Alexander, theoretical studies probably commenced not earlier than March 1931. Blumlein did not keep work journals, and did not publish journal articles; the first written evidence of his studies, a work paper explaining the shuffling technique, is dated 25 September 1931. Practical experiments in 1931 were impossible due to the merger of Columbia and the Gramophone Company into EMI, subsequent restructuring and relocation of Blumlein's laboratory to the new building in Hayes.
On 14 December 1931 Blumlein filed patent application at The Patent Office. The final revision of the application was filed 10 November 1932 and was granted patent status 14 June 1933 with priority right since 10 November 1932. The first, official publication was 24 pages long (22 text pages and 11 illustrations on two pages). The text contained an extraordinary (The author is a staff expert at the British Library) 70 (or "more than 70") claims (a typical patent of the period contained six).
On 4 July 1932 Blumlein compiled an eighteen-page long handwritten summary of the patent, probably intended for Isaac Shoenberg. The second, much shorter memo contains eight typewritten pages. It was signed by Blumlein on 21 July 1932 and duly received and read by Shoenberg. Both memos are now preserved at the British Library.
Neither the patent, nor the memos ever mention the word stereo or its derivatives: Blumlein used the term binaural. There are no references to preceding works, apart from the unnamed military hydroacoustics researchers. In the 1950s and the 1980s American critics hypothesized that Blumlein, who worked in the London branch of Western Electric in the 1920s, could have been familiar with concurrent work by Keller and Fletcher, however, no substantive evidence was ever found. The only certain connection is the fact that Blumlein used Western Electric microphones and disk recorders, which were already retired from EMI studios. According to Barry Fox, the issue of priority has no answer. The origins of the ideas and the paths of thought remain unknown; the technical implementations of these ideas were too different to suspect any exchange. There was no commercial incentive to beat the competition in developing a marketable product; the inventors were working, literally, for the next generation.
Ideas and inventions
Psychoacoustics of sound localization
Blumlein was a modest man who never sought publicity; according to Alexander, "quite often [he was] not fully aware of his genius" or the value of many of his inventions. This wasn't the case with patent 394325: Blumlein "certainly had some feeling for the enormity of this work", and thus prefaced the patent formula with a summary on psychoacoustics of sound localization. Blumlein's theory follows Rayleigh's duplex model, with minor amendments: At low frequencies or long wavelengths localization is determined by the difference in phase. The head is too small to register any differences in sound pressure levels, leaving phase as the only directional clue; At high frequencies or short wavelengths phase ceases to be a reliable clue. However, now the head forms a relatively large baffle, physically separate left and right signals. Thus the brain can rely on relative differences in sound intensity; Fast percussive high-frequency sounds are localized based on the time differences between arriving attack transients, in a manner similar to low-frequency phase differences.
Rayleigh drew the line between low and high frequencies at 1.5–3kHz. Blumlein noted that the "line" is actually several octaves wide, starting at around 700 Hz. Within this band, human hearing can register and evaluate phase and level differences simultaneously. Phase mechanism alone works only below 700 Hz. It is the low-frequency phase information that is lost when binaural signals are reproduced via loudspeakers.
Blumlein tracked the cause of this loss to omnidirectional pressure microphones which were the principal type used in studio recording. Blumlein proved mathematically that phase differences registered by pressure microphones and clearly heard via headphones will be inevitably lost when reproduced via loudspeakers. He suggested compensating losses of phase with pre-emphasizing low-frequency level differences between two stereo channels. When spatial clues present in source signals indicate that the virtual sound source must be positioned to the left of the listener, playback equipment must increase the gain in left channel and attenuate the right channel, and vice versa. These manipulations had to be limited lo low-frequency content; Blumlein specifically warned against tampering with treble signals. The known anomalies of high-frequency hearing were not properly researched and understood yet, so the inventor confined his research to already well-understood low frequencies.
Matrix processing of stereo signals
Popular perception of stereo that emerged decades after the death of Blumlein usually treats stereo signal as the aggregate of two independent channels, left (L) and right (R). Blumlein proposed an alternative approach: stereo consists of a monaural signal M, which is common to both left and right louspeaker channels, and a differential side signal S that defines spatial distribution of sound. The M and S signals are easily derived from L and R by addition and subtraction: M = 0.7017 (L+R) S = 0.7017 (LR), and can be converted back to L and R just as easily: L = 0.7017 (M+S) R = 0.7017 (MS).