Web Edition

The Influence of Recording Media on Composition

A production thesis by Tom Ellard 10542304

University of Technology Sydney 2008

Acknowledgements.

With thanks to Peter King Record Production for the speedy manufacture of the discs, Michel Faber for putting me in touch with C. S. J. Bofop, Brian Eno for admitting the Bofop joke, advice on process music and the tips on dub platters, The Powerhouse Museum, UWS, 8 Track Heaven, the community at pianoworld.com, the Ellard family piano business, Associate Professor Norie Neumark and especially to Chris Caines for the occasional hurry up.

Yanone Kaffeesatz by Jan Gerner

http://www.planet-typography.com/download/typo-kaffeesatz.html

Introduction.

It seems reasonable to claim that sound media are engineering solutions to the needs of sound producers. Also, that development within and across recording formats is consistent with a desire to improve this function. This is stated by the designers, shown via their specifications, and can be deduced from the media itself.

The reciprocal claim that music production responds to the media requires argument. Music existed before media and exists outside of it in performance. Nevertheless there is evidence; in the documentation of recording artists consciously addressing the media, in methods that overcome limitations of the media – for example the unique grouping of musicians found in acoustic recording, in stylistic devices used in recorded music – e.g. flourishes marking an excerpt from a longer classical piece, and in the development of styles that draw from the media itself – ‘DJ’ culture being a common example.

Supportive research is fragmented across a wide methodology from textual analysis of sleeve notes, to interviews, to the empirical measurement of recordings. As Chanan laments, ‘there is no comprehensive survey on the shelves,’ and his own (I think invaluable) overview is ‘modest’ (1995, p. ix). In his (again important) book Katz admits being ‘selective in my choice of case studies’ and that ‘there is a whole world of phonograph effect waiting to be studied’ (2004, p. 7). Method tends to align with genres, with ‘serious’ music more likely to receive empirical study. For example the Centre for the History and Analysis of Recorded Music processes only ‘serious’ C20th music on 78RPM discs (CHARM 2008). One simple reason is that popular musicians are still alive to interview.

I don’t believe it possible to provide a useful written overview of recorded music in the space of this document. A production thesis provides the greatest utility as a new source for study. Here I’ve intended to research and then produce a representative range of recording formats, documenting my production processes. This collects in the one artefact a range of musical solutions by a single composer. Where it is feasible, the physical medium has been supplied so that the listener can inspect the mechanics, otherwise the most reasonable facsimile is provided.

To further clarify the process I have based each recording on the sound of a piano, which applies over the entire history of recorded media – including the pianola. A piano sound should also be familiar to most listeners and so transformations should be evident. In each case I have allowed manipulation of the recording that is appropriate to the period of medium – for example, digital manipulation in digital formats.

One way to view this process is an engineering collaboration between the designer of the media and the recording engineer. But I would go further and argue that rules are necessary to composition, and as some of the rules are imposed by the media, the designer is effectively a collaborator in the music. My process should make some of this collaboration apparent.

Although this cannot be a historical survey (as my work can only be anachronistic) it should be possible to discern some contributing factors to musical styles that came about in the last century. I have tried in my reading to gauge the potential of each media as it was first discovered, then decide upon a ‘best use’ of each and use this as a starting point for my own work. This should make it at least responsive to to recording history and at best part of a social process.

Exclusions and special cases.

Some media has been excluded by obscurity. DAT, DCC, Elcaset and Open Reel are examples of valid formats that have had little impact on the music consumer. (MiniDisc however was successful in Asia, if not elsewhere). Excluded are CD-ROM and game music which are a parallel study. The RCA 45rpm 7 inch format has been excluded as having common ground with other discs, but also an important and complex symbiosis with auto stacker turntables and jukeboxes that far exceeds the resources here (e.g. Hoffman 2008, p. 141).

Compact Cassette has an important social history; the ‘mixtape’ as a mating ritual, the independent ‘K7’ (K-sept) music scene of the 1980’s, mobile music via the Walkman, home recording on the Portastudio, and so on. All are notable but outside this study. Considered only as a recording medium the cassette offers few unique features that aren’t more effectively implemented elsewhere. To acknowledge the disproportionate social influence of the cassette I have decided to not include a composition here – it would misrepresent the format.

Score.

The musical score has an interesting multiple nature.

It remains the legal documentation of a composition. For example in the case of Bright Tunes Music v. Harrisongs Music in 1976, George Harrison’s My Sweet Lord was found to plagiarise He’s So Fine by Roland Mack solely on the basis of a transcription. As Cronin notes:

Had the court acknowledged that since the advent of rock music ... the commercial value of a song like “My Sweet Lord” or “He’s So Fine” depends far more on a specific performance of it than on rudimentary musical elements reduced to notation, it might have reached a less speculative conclusion. (Cronin 2002)

The utility of the score has been in doubt for some time. When transcribing field recordings of Eastern European music, Bartok discovered nuances that could not be placed in standard notation (Chanan 1995, p10). Many composers have since expanded upon or abandoned notation. Cage used diagrams to describe his process music. Stockhausen coloured a print of the Pro Tools recording of his Helicopter Quartet and declared this the score (Stockhausen 2007).

Locally, Percy Grainger described notation as, “a tyrannical... rhythmic pulse that holds the whole tonal fabric in a vice-like grasp and a set of harmonic procedures... that are merely habits, and certainly do not deserve to be called laws” (Grainger 1938).

Some notable 20th century composers embraced recorded music to overcome interpretation of their score. Stravinsky was not alone in deploring the ‘notorious liberty’ of conductors that ‘prevents the public from obtaining a correct idea of the author’s intentions.’ (Katz 2004, p. 109)

Notation remains perhaps to satisfy the lawyers and pedagogues. As recently as 2005, the Australian Guild of Screen Composers required a written score for film soundtrack – a task I found vexing at the time, but useful practice for achieving the current work.

My intention here was to provide a small, reasonably bland melody with some obvious percussive elements that would easily show the transformations brought about by the various recordings that followed. I started by performing into a ‘piano roll’, one editor within the computer based music sequencer FL Studio. Some editing and rearranging took place on the computer, most was done in my head, e.g. anticipating that certain groups of notes harmonise. I necessarily avoided the abilities of contemporary MIDI composition – channels of pressure, controllers, pitch bend and so on – that cannot be included in a written score. It also had to be cleared of any ‘swing’, or timing variation as the score becomes a neutral starting point for a performer’s interpretation.

I then exported a MIDI file for use in scoring software (and eventually player piano). Here I used the scoring module in Logic Studio - not the most able software but certainly more able than me. From there I could layout and print the enclosed document.

As a recording media it fails to fully document my intentions: concepts such as using sections of the piano as percussion are not revealed and performance is entirely open to Stravinsky’s ‘notorious liberty’. More usefully the score becomes a control against which we can compare the recorded works that follow.

Player Piano.

“The Pianoforte”, says The Oxford Companion To Music, “is really nothing but a keyed dulcimer” (Scholes 1991 p788). While we assign its invention to Cristofori around 1720, many hands reshaped it; Johnson (1991, p. 129) pays particular attention to Beethoven’s ever more violent demands – the Hammerklavier of 1818 required a piano of 6½ octaves, 4 strings per note and a metal frame to perform – even then Beethoven managed to break it. In this progress automation seems a consistent refinement.

Two strategies developed – Fourneaux’s Pianista of 1863 was a player positioned in front of an existing piano (Dolge 1972, p. 134), which opposed Pain’s pneumatic self playing piano of 1880 (p. 137). The latter was ultimately triumphant, but the ‘Pianola’ brand piano player became the generic term. It was popular up to the first war, after which it succumbed to radio and a Zeitgeist for change. I use ‘pianola’ here for economy.

I hold Stravinsky as a pioneer pianola composer. At London in 1914 he encountered a night of Russolo’s futurist noise music, followed closely by a demonstration of the Orchestrelle pianola, and a few days later was struck by the generative process in a church carillon. Two years later he contacted the Orchestrelle Co. with plans to use pianolas for Les Noces (Walsh 1999, p. 236). The simple mechanics of the bells had inspired the precision of the machine in creating the rhythmic structure he had in mind.

I must also mention Nancarrow as taking the pianola to its greatest refinement. In general his studies involve difficult ratios; for example 4 against 7 and 5 against 3, canons, tempo accelerations mixed with whimsical jokes – one piano roll had holes punched out to spell HELLO. He avoided academic analysis & never dated or numbered his work, nevertheless Gann has set out a definitive account in The Music of Conlon Nancarrow (Gann 1995).

Piano rolls are no longer commercially manufactured in Australia; the Mastertouch company closed doors in 2005. Peter Phillips developed the first computerised pianola in Australia around 1980, based on an Apple 2 computer. He now sells ‘e-roll’ variants for his ‘Phillips-Ampico’ and other commercial machines. (Phillips 2008)

Several pianola descendants are available – Yamaha’s Disklavier is perhaps better known than the PianoDisc. Both use motorised hammers under the control of a computer, which reads the score either from a standard MIDI file or ‘eMIDI’ or ‘Bar/Ann’ versions ‘tweaked’ to best reproduce on one system. The newer Disklavier Pro uses a proprietary version of MIDI with an extended resolution. Not able to make paper rolls, I’ve targeted a standard Disklavier with my score.

Scores intended for pneumatic pianola must be readjusted to emulate the build up of air behind the holes of the piano roll. A MIDI file intended for a Disklavier needs to overcome the inertia of the hammers, which draws part of the velocity. The ‘ground level’ is raised, while the highest level is lowered. The opposite is true for MIDI recorded on Disklavier that is reused for computer. This kind of information is best found in the many online discussions that take place between pianola enthusiasts, for example:

When I do midi recording on my Pianosdisc equipped piano, I calibrate before each session for minimum velocities of between 3 and 15... before I put any Pianodisc-recorded midi files on the web...What I do is first remap any velocities in the range of 3-15 to a minimum of 15 and then remap the range of 15-65 into the range of 40-65.

I sometimes go the other way with midi recordings... originally recorded with Fatar-manufactured weighted keyboards of the era connected to Akai samplers, and then in the late 90’s with Gigastudio. For these I try to reverse the process, remapping the range from 65-40 down to 65-20 assuming they had minimums of 40 to start with. I also eyeball the top end and make sure nothing exceeds 121 or so. (‘hv’ 2007)

My piece is a ‘4 handed’ work of the sort popular in the pre war era where phantom hands work the extreme upper and lower registers as percussion or counterpoint. The MIDI velocity has been raised and compressed to suit the physics of the Disklavier and the file saved to floppy diskette in a form ready to perform. While it was tempting to create a pseudo Nancarrow piece, without a constantly available machine there was not sufficient feedback to test the effect.

Edison Cylinder & Berliner Gramophone Disc.

Although Scott’s 1857 Phonoautograph has recently been decoded, the first sound reproduction device remains Edison’s Phonograph of 1877, intended for dictation. That intention may explain a design flaw – it’s not possible to stamp copies of cylinders, which had to be pneumatically dubbed in sets from a master, leading to poor quality ‘second generation’ sound.

Berliner’s Gramophone of 1888 used a disc and altered the ‘hill and dale’ recording action to a horizontal movement. Copies could thus be stamped and the discs played around four minutes at a time where an Edison ‘Standard’ cylinder could store two. Double sided discs introduced by Odeon in 1905 immediately doubled the duration.

The difficulty of fitting existing work on disc is illustrated by Verdi’s opera Ernani released 1903 on 40 single sided discs. (Scholes 1991, p.421) Most early performances were thus truncated to fit a side, often with ‘a modest ritardando’ to acknowledge the edit (Beardsley R. & Leech-Wilkinson 2008). Popular music soon adapted to the time limit and jazz recordings flourished over the life of the ‘78’ (Rust 1982, p.i).

Early recordings were acoustic, the performance directed at the horn of a recording machine with the expectation that the volume of the sound would drive the mechanism. The mix required that the instruments be arranged around the horn in relationship to their power. Instruments such as brass and loud voice were more effective and so became prominent in music recordings. Pianos were rarely heard alone. Amplified recording only became standardised in the 1930s with the Western Electric patent. (Beardsley & Leech-Wilkinson 2008)

In 1928 Kurt Atterburg was awarded £2000 by Colombia Gramophone for a recorded symphony dedicated to Schubert. This is perhaps the first documented ‘serious’ work for gramophone. (Scholes 1991, p. 58). Two years later Paul Hindemith performed several pieces for voice, viola, xylophone and gramophone at Neue Musik 1930, with some evidence that he used two machines at once – arguably the first DJ performance (Katz 2004, p.100). John Cage was in the audience, and would go on to make works such as Cartridge Music in 1960 (p. 113).

The last 78rpm discs were deleted from EMI’s catalogue in March 1962 (Scholes 1991, p. 420)

I desired to create an acoustic recording for Edison Standard cylinder based on my pianola score, truncated to two minutes and played at the frantic pace of many early recordings. I enquired about cylinder recorders held at the Powerhouse Museum and was told that the collection is for preservation, not use. An American recording service seems to have closed down. In the remaining time frame I decided to construct a sonic model.

Commercial modelling software (e.g. Audio Ease’s Speakerphone) use an ‘impulse response’ recorded through a gramophone horn. By using this as a convolution source, the acoustic space of the horn can be imposed on a new signal. The failing of this technique is that an acoustic recording passes through such a horn twice, to and from the recording mechanics – which aren’t included in the model at all. Speakerphone also uses sampled surface noise to augment the illusion, whereas I sought the sound of a new, idealised cylinder recording.

A useful impulse would need to be acoustically recorded and reproduced on a cylinder. However my hunch was that the frequency spectra of a piano has remained roughly static over the last hundred years and so a cylinder recording of a piano could be used to convolve a new acoustically flat piano recording, the difference being the changes imposed by the recording process.

The library at University of California, Santa Barbara holds over 6000 cylinders which have been digitised at 24 bits (Seubert 2008). From there I found a recording of The Laughing Song from 1904 and with a high pass filter removed the rumble of the mechanism, which lay below the actual recording. I then edited together sections where only a piano is heard.

Using the Match EQ module of Apple’s Sound Edit Pro I captured the frequency response of the cylinder recording, then imposed that spectra on my own piano recording. The result is quite plausible when compared to the actual historical recording used and can be heard with intermediate stages on the documentation CD.

It was then simple to adjust the relative velocities used in the score to best reproduce through the modelled playback – in effect, directly composing for the playback device.

To create the spectra we need the sound of a piano that has passed through the entire acoustic recording process. This 1909 cylinder has a mixture of singing voice and piano but the voice could be edited out. There is then the rumble of the player’s motor, which was easy to remove as visibly distinct from the sound recording. The high pitched scratches have little or no effect on the frequency analysis which is driven by overall sound pressure. This screenshot shows the spectrum captured in Soundtrack Pro using the EQ match filter. It shows the cylinder to have a complex frequency response between about 200 - 3000Hz, keeping in mind that the impulse here is only a piano and not a white noise sweep. The EQ shape is now stored as a preset to be imposed on the new signal.

 

I then took a piano recording free of acoustic information and forced it to match the EQ shape from the cylinder. The result was a new sound recording that carries the equalisation effects of the acoustic recording process. The underlying assumption is that the original sound of the piano is the same in both cases - not entirely likely but within the proximity which allows a reasonable model to be created. I could then rework the original recording slightly while hearing it in real time ‘through’ the 1909 transformation.

Columbia 33rpm Long Playing Disc.

The ‘LP’ was an engineering response to the collapse of the record industry during the great depression. In 1939 Columbia Records started work on a new 12-inch format. The label had been recording on 16-inch ‘talkie’ motion picture discs at 33RPM and so used that speed. In 1946 Columbia president Wallerstein rejected a 16 minute prototype, and by measuring the durations of CBS classical holdings determined a specification of 34 minutes total. The LP was finally ready in 1948 (Hoffman 2008, p. 113). To avoid the license fee, rival RCA developed the 7 inch ‘45’ in 1949, targeting the pop market. The companies eventually cross-licensed, bringing about ‘albums’ and ‘singles’ at both speeds (p. 115).

In all disc formats sound quality decreases as the needle approaches the centre of the disc. This ‘tracing distortion’ comes as sound waves are recorded over less linear space, and the curvature of the groove distorts the waveforms. Discs are cut linearly but the pick-up is usually mounted on a pivot, causing azimuth errors. This creates a ‘sweet spot’ at the exterior of the disc. The engineering response is mixed – in my experience Australian cutting engineers placed grooves in the sweet spot (Fig. 1a), while Europeans spread them over the entire available surface (Fig. 1b). It seems known anecdotally by recording artists, but a compositional influence is difficult to prove. While the last track is often less energetic than the first, that could be due to the importance of the first track. Here I’ve simply responded to the known physical process.

Morton Subotnick claims his 1967 Silver Apples of The Moon to be the first ‘original large scale’ composition for disc, written ‘in two parts to correspond to the two sides of an LP’ (Hertzog 1992, p. 894). Usually the second side was used initially to extend overall duration, but over time the ‘flip side’ of a record became a counterpoint. The introduction of the Billboard charts in July 1940 caused studios to concentrate on one hit recording (Hoffman 2008, p. 91). 45rpm discs were marketed as a ‘single’ track ‘cut with’ another – the A side clearly marked for disc jockeys. From about this time came the definition of ‘a flip side’ as an alternative to any process (Room 2000, p. 247)

One distinct B side form is the ‘dub version’. ‘Ruddy’ Redwood has been proposed as the inventor of the ‘version’ around the early 70’s, by accidentally neglecting the vocal line in a mix for a ‘dub’ or acetate (Barrow & Dalton 2004, p. 216). ‘King Tubby’ Ruddock is often cited as the inventor of the ‘dub mix’ (e.g. Larkin 2006, p. 840) but this is disputed by Barrow & Dalton. The first commercial 12 inch extended single was Ten Percent mixed by Walter Gibbons in 1976 by DJing of acetates – creating a new record from records (Gilbert, Pearson 1999, p. 135).

All these innovations eventually became formalised as a 12 inch ‘dance’ or ‘extended’ 45RPM disc with a B side holding a ‘dub version’ or elements that could be used by a DJ to further compose the recording.

My design here was based around three aspects of an LP disc. Firstly to contour the frequency spectrum of the recording to illustrate the gradual effect of ‘tracing distortion’. Secondly to create a ‘B side’ where parts of the main recording are analysed in a ‘dub version’. Thirdly to carve a groove pattern on the disc such that the music is presented visually for DJ use (cf. Fig 2).

I needed a wide frequency range to reveal the process and to have loud impacts to make noticeable grooves. So I extended my piano percussion by kicking, hitting and scraping an old upright piano and recording samples to MiniDisc. I then built a prepared piano sample set from this. (See the photographs on the following page spread.)

For the A side I resequenced my score at 100BPM expecting that at 33RPM a triangular pattern would form on the disc face. Then I slowly reduced bandwidth over 16 repeated plays of the score, but the result was uninteresting. Instead I used GRM Tools software to design a set of 16 equalisations of the sound that reduced the bandwidth of each repeat in a more musical way – this became a new ‘meta score’ and is shown over the page. On the ‘flip side’ is the original bandwidth idea in reverse, along with the separate elements that are combined on the ‘A side’. A DJ can thus create new workings of the piece.

Preliminary results have been surprising. The disc included here was hand cut on transparent polycarbon by Peter King on a 1960’s lathe. There is little evidence of a physical pattern, possibly as the lathe is not microcomputer controlled, so all grooves are the same distance. All the lathes I have previously used changed their groove distance automatically on amplitude. It also seems that the extreme frequencies in the recording are causing the polycarbon to resonate, making it hard to determine tracing distortion. I intend to try again with pressed vinyl to verify the design.

Piano percussion: the three images spread across these pages show some of the techniques used to create percussive sounds on a piano. From left to right - strumming the strings creates a harp like glissando. To create a midrange thud sound, dropping the lid from a small height. A large bang is made by slapping the face of the sounding box with an open hand. None of the actions damages the piano.

These sounds were used to create the 12 inch disc, the cartridge and MD recordings that needed a wider frequency range than the usual piano sound.

Lear Jet 8 Track Cartridge.

8 track cartridge is perhaps the most unusual format covered here. The design draws from the ‘Fidelipac’ cartridges used in radio broadcast, which were adapted into a 4 track system in the early 60’s by Earl Muntz. Bill Lear reverse engineered the Muntz design and announced a cheaper 8 track version in 1964. The prototype was adopted rather quickly by Ford as Motor-ola, another Lear partnership, was under their contract. In 1966 some 65,000 players were sold by Ford and manufactured by Motorola. (Schmidt 1999)

The cartridge holds a continuous loop of ¼” recording tape, which is pulled out from the central spool up to the playback head. A pad in the cartridge presses the tape up against the head. The capstan is spun by the player’s motor rolling the tape against the pinch roller in the cartridge pulling the tape at 3¾ inches per second. The tape loops back to the outside of the spool. When the entire length of tape has looped, a metal foil splice in the tape passes by a solenoid sensing coil, kicking the playback head along the width of the tape to play a new program. The topology is thus 4 equal spaces joined in an endless loop. The gap between sides is quite short. It’s impossible to rewind the tape and so very difficult to cue recording. Most cartridges were adapted from LP records with songs rearranged into roughly equal quarter sides – in some cases (for example Sgt. Peppers) songs were re-edited to add or subtract bars.

In 1975 Lou Reed released Metal Machine Music as a possible homage to, or parody of La Monte Young, as ‘noise music’ or to void a recording contract. He has provided conflicting evidence over the years. The album was divided into 4 equal sides of 16:01 minutes with an endless groove at the end of the last side on vinyl. This indicates a desire for an endless playback and a 4 sided structure – perfect for the 8 track version released by RCA.

Unfortunately in The Band That Would Be King, a documentary film of the underground musician Jad Fair, Lou Reed claims not have seen the cartridge (Feuerzeig 1993). I was not able to obtain a reply from Reed for this study. My best conclusion is that this perfect 8 track was devised by RCA staff.

My own design used four parallel variations of one piece that can be switched at will or allowed to run as one endless programme (see the diagram across the page). I designed each soundscape with smaller tape looped recordings which are also unable to be rewound or cued – this forms a coherent epicyclic construction where ‘like begats like’. The loops were dubbed from minidisc recordings of my upright piano.

I was able to buy an 8 track recorder and six 84 minute Ampex cartridges on eBay. The duration of these antique carts set the length of the piece. Unable to amass many analogue tape recorders I played 2 hired Revox B77 machines into digital multi-track as a temporary buffer as shown in the diagram on the facing page. Each set of tape loops was performed live into a 8 part orchestration. Four performances are dubbed onto each blank cartridge. It was obviously impossible to make the three examination copies exactly the same. The documentation CD has two 21 minute passes over a cartridge with some manual switching.

Compact Disc.

Compact disc was announced in prototype by Philips in 1979, smaller than the final version at 11.5cm and holding one hour of music. The specification was drawn from the diagonal length of a compact cassette, and aimed at European automotive installations. The final diameter was allowed by Philips only after Sony engineers proved that a 12cm disc would fit into all suit jackets available for measurement. From the diameter and the error correction follows the 16 bit and 44.1KHz sample rate.

Sony’s official reason for the 74 minute duration is that founder Norio Ogha decreed, “Just as a curtain is never lowered halfway through an opera, a disc should be large enough to hold all of Beethoven’s Ninth Symphony” (Sony Corp. 2008). The story possibly reflects a more calculated assessment that, as with the LP before it, the majority of Sony/CBS Records’ classical inventory would fit.

The CD is a single side of 74m 42s duration, with a consistent frequency range of 22,050 Hz. Compared to previous formats it has a noticeably low noise floor. It precludes the kind of A/B comparison of vinyl records – it invites one recording or flow of ideas.

My chosen expression of the compact disc format is Brian Eno’s 1984 recording Thursday Afternoon. Originally a video soundtrack, the compact disc is a redesign 61 minutes in length. In the sleeve notes for the CD we read that it:

... is perhaps the first recording specifically for the compact disc and it utilized two new freedoms of that format: it is 61 minutes long... and it is occasionally very quiet.... It seems likely that, just as the 78-rpm record set the scene for the 3-minute song, so the compact disc will foster an interest among composers in long-duration pieces like this one.

The liner notes are credited to C. S. J. Bofop, which is B. R. I. Aneno with the letters moved along one position – which I take as the real author. I contacted Brian Eno for further clarification and received this reply:

I think Thursday Afternoon was the first thing ever conceived and recorded specifically for CD in the sense that I made it deliberately to occupy the additional running time that a CD offered. I’d been wanting to make ‘endless’ music since I released ‘Discreet Music’ in 1975 - in the case of that piece I made it as long as possible for one side of a vinyl album (I think it was 30’34”) (Eno 2008)

Endless music requires a generative process. My mechanism draws upon cross section of Eno’s work at the time (particularly Discreet Music) as well as digital tools that have come about since. The system is diagrammed over the page. Three independent look up tables of MIDI notes are accessed at random. These control two sound generators – a physically modelled piano and a simpler sampled piano with a hard low pass filter to provide accompaniment. The sound from these are passed through a bank of four hard band pass filters that allow only small set of the overtones from the piano sound to be heard, such that one note can have a wide variety of tonal ‘windows’. These sparse soundings enter a two channel delay system that has been tuned to soften the repeated notes over time in a natural decay.

The system is essentially endless: hearing all possible combinations of filter and chord would require about 716800 hours with no repetition. However I have addressed the ‘arc’ of CD duration by programming a side effect chain which is eventually mixed in over the whole period. This closes all potential outcomes into one performance, the CD document. Due to a rare problem with some older DVD players and ‘burned’ discs I have limited the performance to 72 minutes.

Composing for this system is done by selecting the notes that are held in the lookup table including rests, and the frequencies that are available to the filters. The score is both this ‘seed’ state and the layout of the process – which provided as an Ableton Live document.

Digital Versatile Disc.

At the start of the 90’s two competing prototypes existed – Multimedia Compact Disc by Sony and Philips, and Super Density Disc by Toshiba and 7 partners. Intense pressure from the computing industry brought compromise on the Toshiba design administered by a DVD Forum. Given later battles, Sony’s biography is ominously silent on DVD. (Sony Corp. 2008)

DVD has multiple uses. A single sided DVD ROM disc holds about 4.7Gb of any data, but not in a standardised form. DVD Audio is an audiophile format that failed to gather enough popular interest. Although DVD Video is not a music container by design, it’s the most popular and convenient format for the end user and the version I used here.

Media on DVD Video is compressed – video as MPEG2 and sound as one of AC3, DTS, MP2 or PCM. These streams and others are combined in an MPEG2 container. Depending on the compression sound can be 48 or 96KHz in 16 or 24 bits with as many as 8 channels. The nominal bandwidth of the whole stream is 9.8Mbit/s. The average duration of a DVD Video is 2 hours.

This specification is a recipe that has been implemented in hardware with various success. For that reason there is a lowest common denominator of the recipe. I have learned that video at 7Mbit/s combined with Dolby AC3 surround at 384Kbit/s will reproduce on most players.

The DVD programme is represented as a virtual layout including menus, titles and chapters and these are the elements by which a musical work can be organised. Unlike CD where a set of tracks are played in order, chapters may branch to any other, which creates a 2D flow rather than a linear path. Each object in the flow employs a ‘state machine’ in which 16 ‘GPRM’ registers in the DVD player can be set and interrogated. This allows quite complex programmed branching based on user input, for example scoring for a multiple choice test. Amid vast potential I have this time chosen a relatively simple process.

My design addressed these potentials; 5.1 surround sound at 48KHz and a flowchart of chapters which autonomously link according to a generative process. It also highlights the variety of the DVD recipe. Perhaps due to the simplicity of the design I haven’t found a previous realisation of this kind of album.

I recorded 85 individual piano chords in 5.1. surround, simultaneously generating a small piece of live video to illustrate the notes played. (There must be some video in the stream and so I’ve included this signal which is part of another design project). I then cut each chord apart, creating MPEG2 clips about 3 seconds long. GPRM 14 provides a pseudo-random value based upon the system clock which I have used to jump at the end of every clip.

The clips are organised into two groups based on their key. By adjusting the opportunity for a jump to reach the alternate group I arranged for the DVD to switch keys on the odd occasion. Once started the disc will play indefinitely at a speed based upon the hardware that hosts it. The score in this case is; the chords chosen to be played, the way they are grouped and the actual hardware playing the disc. A screen shot of the DVD Lab authoring tool over the page illustrates the flow.

The project hit some issues. In a previous sketch I’d used a tempo of 120BPM, which fit nicely into the 25FPS grid of PAL video playback. This time I wanted the tempo to be 70BPM to suit the music style – which doesn’t fit well into 25 frames. This gives what could charitably be called a ‘human feel’ to the piece. In future it might be best to use 30FPS NTSC. Another factor is the seek time of a hardware DVD player which struggles to keep pace – copying the disc to hard drive is extremely helpful.

Layout of the DVD in DVD Lab software.

Mini-disc.

The MiniDisc format was introduced by Sony Corporation in 1992, intended to replace the compact cassette (Sony 2008). The physical container is a 6.4cm optical disc housed in a permanent caddy. MiniDisc (henceforth MD) records a digital signal compressed by Sony’s ATRAC codec. While it anticipated and bettered MP3, ATRAC failed to gain market share. It and the MD have been discontinued. The original MD was available in 74 & 80 minute lengths, while a newer Hi-MD format offered 94 minutes of uncompressed PCM audio, or approximately 8 hours with compression. For this project, the original MD format was used as the most popular version.

MD audio is random access data with an index at the head of the disc. The design allows seamless joins and extremely fast seeking in the sound recording. Data is read at 4 times the required rate, filling a sample buffer. Eric Woudenberg found that in the worst case of recordings made either end of a disc, his test machine was able to provide seamless playback of segments at least 3 seconds in duration (Woudenberg 1996). This sets the smallest sound element that can be used in a composition.

In 1998 the UK group Gescom released the album Minidisc, with 88 short segments of abstract sound accessed at random by an MD player to form an endless montage. The disc was hailed by Sony in the UK as a creative use of their technology (Phillips 1998) and went on to gain a distinction in the Prix Ars Electronica of 2000 (Ars Electronica 2000). It has recently been reissued on CD by the Or label, which rather misses the whole point, or contributes to the point I am making: the album makes sense only on MD. I was misled that the Gescom MD was designed to be performed in multiple copies, and so I have made my MD do this by designing my segments to roughly complement each other. It uses the same process of on-the-fly assembly as the Gescom MD but all sounds are created from a large palette of heavily processed (mainly via granular resynthesis) piano. The exception are three samples of Internet CAPTCHA (password systems for the visually impaired). After some experimentation I wanted to have some recognisable feature that would occasionally alter the feel of the piece.

The enclosed documentary CD has examples of one and two MiniDiscs performing. Three at once is not pretty.

Web and Thumb Drive.

A decline in sales of physical recording media has accompanied an increase in music available for download, indicating that music may eventually be sourced entirely online as a kind of advanced ‘radio’. The trend is mostly an issue in publishing, outside of this study. Here we need to define the effect on composition.

The medium involves compression technologies such as MP3 and FLAC. It also involves a virtual container of the compressed files – the HTML of a web site or simply a torrent directory that holds raw files. The container could also be the playback hardware, such as an iPod.

The quality of a downloaded recording will usually be less than the uncompressed original. I have compared my own masters to versions sold via the iTunes Music Store and found significant loss of fidelity. As music will be recopied and encoded by persons unknown in formats unknown, it seems impossible to anticipate this signal loss in the original composition.

Individual tracks are freed from the artist’s control. An ‘album’ becomes whichever tracks the consumer happens to collect. I’d argue that the main business of the composer becomes one of encouraging the listener to preserve the collection as designed. Even an album sold as one unit may soon be redistributed in pieces. There needs to be an electronic container for the work that equals the traditional sleeve.

In July 2003 I wrote to Apple CEO Steve Jobs suggesting a PDF container for iTMS music sales, as it would preserve artists’ designs. He replied that, “We disagree. We think that it’s the music, not the package that the music comes in, that’s important here” (Jobs 2003). In October 2004 Apple then announced ‘the online music industry’s first-of-its-kind Digital Box Set’, with PDF files now suddenly recognised by the iTunes software (Apple 2004). While perhaps a victim of the legendary Jobs ‘reality distortion field’, I see this as a win for composers.

A cheaper yet tangible means to contain compressed files was developed by the German punk group WIZO that year with their Stick-EP, a USB thumb drive with 5 music tracks, photographs and texts (WIZO 2004). A customised stick presents the convenience of MP3 files while discouraging their dispersion. Currently companies such as All Access Today offer sticks as instant live merchandise, holding a performance just attended by the consumer (All Access Today 2008)

I first set up an electronic version of this production thesis online, at http://tomellard.com/album. The site holds revisions of this text divided into web pages, as well as videos and sound samples from all the various recordings. I had considered using iTMS or http://archive.org as more popular sites, but felt that their visual presentation was inadequate to enforcing the idea of an album.

That HTML site design then became the basis for the USB thumb drive included here, which replaces the lower quality web audio with variable bit rate MP3. The drive thus represents the original ‘album’ even though the individual files may end up copied and dispersed. It also reflects the entire work in an economic format – as a USB drive now costs around the same price as a cassette.

Reflection.

Encased, the individual album formats become an even larger ‘album’, which is the work. To my mind another case is made – even if the examination of each format has been limited by my time, knowledge and ability, considering the work as a whole it is obvious that significant influence has been exerted by the appropriate rules of each container. I argue that any composer met with both a 2 minute cylinder and an endless DVD will develop a wide gamut of approaches. This alone calls for the preservation of all antique formats.

I’m also suprised how the collected music evokes so much history, when made by a person untrained in earlier music forms. If we allow that the developers from Edison to Toshiba are co-composers of some sort, the collaboration here seems as strong as working with the members of a living band. It leads me to doubt there is or will ever be the one perfect medium for recorded music. Importantly, it calls for musicians to take greater influence in future recording formats, acknowledging that these are part of the process of music making.

I hope that this work, for all its limitations, will be of use and inspiration to any further exploration of recorded music.

References.

All Access Today 2008, Client Services, Houston, viewed 1st April 2008, <http://allaccesstoday.com/clients.html>

Apple 2004, Apple Introduces the U2 iPod, Cupertino, viewed 1st April 2008, <http://www.apple.com/pr/library/2004/oct/26u2ipod.html>

Ars Electronica Linz 2000, Digital Musics Distinction 2000, Linz, viewed 1st April 2008,
<http://www.aec.at/en/archives/prix_archive/prix_year_cat.asp?iProjectID=14&iCategoryID=9555>

Barrow, S. and Dalton, P. 2004, The Rough Guide To Reggae, 3rd ed., Rough Guides, London

Beardsley R. & Leech-Wilkinson, D. 2008, A Brief History of Recording, Research Centre for the History and Analysis of Recorded Music, London, retrieved 4th May 2008, <http://www.charm.rhul.ac.uk/content/KCL_resources/beardsley_brief_history.html>

CHARM 2008, Research Projects, Research Centre for the History and Analysis of Recorded Music, London, retrieved 4th May 2008,
<http://www.charm.rhul.ac.uk/content/projects/projects.html>

Chanan, M. 1995, Repeated Takes: A Short History of Recording and Its Effects on Music, Verso, London

Cronin, C. 2008, Bright Tunes Music v. Harrisongs Music, UCLA Law School, Los Angeles, retrieved 3rd April 2008,
<http://cip.law.ucla.edu/cases/case_brightharrisongs.html>

Dolge, A. 1972, Pianos and Their Makers, Dover, NY

Eno, B. 2008, Email to the author, personal communication, 17th April 2008, London.

Frith S., Straw W., Street J. (eds.) 2001, Cambridge Companion to Pop & Rock, CUP, Cambridge

Gann K. 1995, The Music of Conlon Nancarrow, Cambridge University Press, Cambridge.

Gilbert J.& Pearson E. 1999, Discographies, Routledge, London

Grainger, P., 1938 in Crab, S. 2005, The Electric Eye Tone Tool Cross-Grainger for Playing Graingers Free Music, Crab S., London, viewed 3rd May 2008, <http://www.obsolete.com/120_years/machines/free_music_machine/>

Herzog, ‘Morton Subotnik’ 1992, in Morton B. & Collins, P. (eds), Contemporary Composers, Saint James Press, Chicago

Hoffman, F. 2008, Chronology of American Popular Music, Routledge, New York

‘hv’ 2007, Playing Yamaha Disklavier MIDI files on non-Yamaha DPs, pianoworld.com, Nahant, viewed 12th Dec 2007, <http://www.pianoworld.com/ubb/cgi-bin/ultimatebb.cgi?ubb=get_topic;f=6;t=003438;p=0>

Jobs, S. 2003, Email to the author, personal communication, 4th July 2003, Cupertino

Johnson P. 1991, The Birth of the Modern, Harper Collins, New York.

Katz, M. 2004, Capturing Sound: How Technology Has Changed Music, Uni. of California Press, London

Larkin. C. 2006, Encyclopaedia of Popular Music, Vol. 4., Oxford University Press, Oxford

Phillips, A. 1998 in Or 2003, GESCOM minidisc, Or, London, viewed 1st April 2008, <http://www.itsaor.net/only3.html>

Phillips, P. 2003, About the PA eroll files, Peter Phillips, Padstow, viewed 18th May 2008, <http://www.petersmidi.com/About_the_erolls.pdf>

Room, A. (ed.) 2000, Brewer’s Dictionary of Modern Phrase & Fable, Cassell & Co., London

Rust, B. 1982, Jazz Records 1897 – 1942, Vol.1, Storyville Publications, Essex

Scholes, P.A. 1991, The Oxford Companion To Music, 10th ed., Oxford University Press, Oxford

Schmidt, F. 1999 in Forster R. 2007, The Man Who Invented ‘Kerchunk’, 8 Track Heaven , Chicago, retrieved 20th May 2008, <http://www.8trackheaven.com/frankschmidt.html>

Seubert, D. 2008, Cylinder Preservation and Digitization Project, University of California Santa Barbara, viewed 1st April 2008,
<http://cylinders.library.ucsb.edu/overview.php>

Sony Corporation 2008, Sony History, Sony Corporation, Tokyo, viewed 18th May 2008, <http://www.sony.net/Fun/SH/index.htm>

Stockhausen, K. 2007, HELICOPTER String Quartet, Stockhausen, Kurten, viewed 12th December 2007,<http://www.stockhausen.org/heli_graphic.html>

The Band That Would Be King 1993, motion picture, Vanguard Cinema, Buena Park

Walsh, S. 1999, Stravinsky: A Creative Spring: Russia & France 1882-1934, Alfred Knopf, NY

WIZO 2004, “WIZO STICK-EP” ist der erste USB-Speicher-Stick einer Band, Wizo, Stuttgart, viewed 1st May 2008 <http://www.wizo.de/stick/index.htm>

Woudenberg, P. 1996, MiniDisc FAQ: Disc Topics, minidisc.org, New York, retrieved 18th May 2008, <http://www.minidisc.org/faq_sec_3.html>