Tuesday, January 31, 2006

Klassics: Playback 2

The translation of sound to digital information could be thought of as a second divorce from the acoustic. Sound (at least until it reaches the inner ear) is a mechanical phenomenon. Just as a stone dropped in a pool creates waves in the water, things that make noise creates waves in the air.

The first sound recordings were also mechanical. Edison’s tin cylinder recorder collected sound with a simple horn and channeled it to a diaphragm attached to a stylus which etched the vibrations onto the cylinder. This created a physical trace of the mechanical vibrations of sound - in a sense, mapping the physical vibrations of sound, both the volume (how strong the vibrations were) and the pitch (the frequency of the vibrations, that is, how close together the waves were) onto a physical surface. Playback reversed the process, turning that map back into sound by running the stylus over the surface. The cylinder’s action on the stylus caused the diaphragm to vibrate, creating sound which was “amplified” by the horn. While the techniques and materials of recording were refined over time and greatly changed by electronics, the fundamental principle of a modern vinyl LP remains the same. You can try this experiment at home, if you still have a turntable and a album you don’t care about: stick a common sewing pin through something like a paper cup so that the sharp end is sticking out the outside of the cup. Bring the needle gently onto a spinning record’s surface and you can hear a very low fidelity but recognizable translation of the recording, amplified by nothing more than the cup acting as a diaphragm. The recording is that direct of a physical impression of the vibrations of sound.

Many methods and formats were devised for recording sound in the years that followed Edison’s invention, but all were based on making a physical trace of sound vibrations onto a solid substrate. However, the rapidly developing technology of electronics was making profound impacts on audio technology throughout this time. Indeed, these two technologies developed alongside one another and have always been closely interconnected. Early microphones started to be developed in the late 1800s. Many different principles by which sound vibrations could influence an electric current were explored in the development of the microphone. The dominant technologies that came out of this development were the dynamic microphone, which operates by fluctuating a wire coil through a magnetic field (the same basic principle as a generator), and the condenser microphone, where a static electric charge is applied to either the diaphragm or a back-plate behind the diaphragm, and the fluctuating distance between the vibrating diaphragm and back-plate creates the audio signal. But regardless of their technological foundations, all microphones work by translating the vibrations of a diaphragm into a fluctuating electric current.

There was no way to record this electronic audio signal, however, until the development of magnetic tape in Germany in the late 1920s. With this the first divorce of sound recording from the acoustic was complete. Sound had been transformed into a purely electromagnetic phenomenon.

Jump to the Twenty-first century. I still own a turntable and a fair number of records. Since that all-inclusive on-demand library of digital audio recordings over the internet hasn’t panned out yet, when I got a yen to own a copy of the theme song of the 1978 television mini-series Flambards, what I ended up with was a 1979 Phillips “Phonogram” label 45 that I ordered from a record shop in the U.K. Ah, nostalgia. This record went on to a life of near undisturbed tranquility in the stereo cabinet because playing a record is sort of a pain, particularly if you want to do as much as possible to maintain the fidelity of the recording. Clean the record, clean the needle, play it, put it back in its sleeve, put it away. I rarely play any of my records anymore.

But recently I acquired some cheap and accessible hardware and software to better suit my home computer to the recording of external audio inputs. The next time I played that single, the long-remembered theme song got recorded, burned to a CD, and resided in MP3 format on my computer. I think I’ll listen to it right now. Ah, that was nice.

What just happened?

Superficially what I did when I made a digital recording of a song from a 16 year old record is no different than if I had made a copy on the tape recorder connected to my stereo (which function I haven’t used in years). An electronic signal of audio information created by the physical vibrations of the record against the player’s stylus goes through some electronics and ends up as information on a magnetic medium (in the case of the computer, the hard drive). But there is a fundamental difference between these two types of information that is as great if not more so than the difference between vibrations in air and an electric current or magnetic recording.

Divorced as it is from the physical vibrations of sound, the electrical current created by, say, a microphone shares something of the nature of those vibrations. The information contained in both shares the nature of being defined by frequency and volume. A stronger current represents a higher volume. A higher frequency fluctuation in the current represents a higher pitch of sound. The cardinal characteristics of sound are captured and impressed in the nature of the current. A magnetic recording translates these characteristics as magnetic fields of greater or lesser strength fluctuating at higher or lower frequencies.

And there’s a problem in this similar nature, which is distortion, or noise. What would a perfect transcription of sound be? I suppose something capable of producing the exact experience of perfect human ears hearing the original sound under ideal acoustic circumstances. Microphones, electronics and magnetic media do not make a perfect transcription. A component of very electrical current and magnetic recording is unrelated to the original sound - electrical or magnetic noise caused by interfering magnetic and electronic fields in the environment and limitations of the recording equipment. Distortion is an inevitable component of recording: there is no signal without noise. The greater problem comes when this recorded information is duplicated. There is no way to duplicate this information without turning it back into an electrical current and recording that by the same or a similar process as the original was recorded. All the same kinds of limitations and noise sources come into play, so that in the second generation of the recording the amount of noise in the signal increases.

My audio cassette collection is slowly dying: the oldest cassettes are succumbing to mechanical failure and magnetic drift, the signal daily more corrupted by noise, and there isn’t a single thing I can do about it. Inasmuch as entropy rules the cosmos, all media are inherently volatile. But see how this relates to the duplication problem: for a sound recording of the old type to be preserved indefinitely it is necessary to duplicate a particular copy before its medium expires. In each copy the ratio of information to noise decreases. Inevitably, the information will be lost altogether over the very long term.

The digital recording I made of an old 45 is utterly different from the audio cassette recording I could have made. Although recorded on a magnetized surface its information is not defined by the strength or fluctuation frequency of the magnetic impressions. What is it?

It is information about the electrical signal rather than an impression of the signal itself. A digital audio file is composed of information of individual samples of a particular electronic impulse representing sound - on a CD each sample being composed of two bytes of binary information, taken at a frequency of 44,100 samples per second.

There are two major initial outcomes of this new paradigm of recording audio signals. The first is an (arguable, but in my opinion genuine) increase in the potential fidelity as compared to existing recording technique. Of course, you can still make a rotten recording at 44,100 rotten samples per second: put another way, making a digital copy of the Flambards theme song didn’t eliminate the hiss and pop of a beat-up old record. But in practice properly mastered CDs have higher fidelity than comparable quality records or cassettes.

The other, and I think more important, outcome is the duplication problem is eliminated. Rather than duplication of successive generations of a recording being carried out by continuously translating and retranslating between electrical currents and magnetic fields, a copy of digital data consists of an exact transcription of the data about the sound - the specific value of those two bytes of binary data for each 44,100th of a second. This may not matter much in the short term and my idiosyncratic transcriptions of old analog information. Consumer electronics giant Philips is probably not convening an emergency board meeting to discuss my making an immortal recording of a noisy 45. What happens when I put a CD into my computer is much more interesting. The copy of the CD data the computer makes is utterly identical to the data on the CD - as is a CD copy I make from the data transcribed to the computer - as is the data a friend, say, copies onto their computer from the CD duplicate I made... ad infinitum.

There are other significant effects of translating sound into digital information. This kind of information lends itself to being attached to meta-information - information about the information, like its title, composer, artist, and date. It divorces the information from a specific medium. It doesn’t matter how that data is stored - whether on magnetic, optical, or purely silicon media. It lends itself to manipulation of the data, particularly in two important ways. One is remastering - programs which analyze digital information in order to preserve the signal and decrease the noise have become increasingly sophisticated. The second is compression - approximating the same audio information in a smaller amount of binary information.

Very interesting things are happening to the way we can interact with recorded audio information as a result of the confluence of all these effects. The next chapter will look at the social equation: how recording and electronics changed the nature of music in society and why we could be on the brink of the greatest changes yet.

this is what is up with this.

No comments: