Digital Recording (Part 2)
The Compact Disc
Early next year we will witness the dawn of an exciting new era in audio production - the Compact Disc.
The disc is 120mm in diameter, plays up to 60 minutes of uninterrupted music, has a frequency range of 20Hz-20kHz, a dynamic range, signal to noise ratio and channel separation of greater than 90dB, total harmonic distortion of less than 0.005% and an unmeasurable amount of wow and flutter!
So how are these impressive specifications achieved? The answer - Digitally.
Digital Sound Reproduction
Sound is a variation of air pressure detectable by the ear. Until now, audio systems have worked on the analogue principle of converting air pressure variations into electrical voltage or current variations, processing them as necessary in the analogue state and then reproducing the sound waves through loudspeakers. This system, at its present state of development, works pretty well. But it is practically at its final performance limits. The waveforms can hardly be made more accurate, the levels of noise and interference can hardly be further reduced at reasonable cost, in analogue audio equipment. Still, it is not possible to reproduce the original sound both exactly and consistently. This is very largely because inescapable factors such as non-linearities, noise, power supply and temperature variations affect the actual shape of the analogue waveforms in the conversion processes.
With digitalisation, however, the road to substantial performance improvement is wide open. The audio waveform is sampled at high speed, and the value of each sample is measured, see Figure 2. Each value is then converted to a digital number, in binary code (in digital terms this is known as a word or symbol). The string of successive binary numbers is the exact digital equivalent of the audio waveform. As long as the numbers maintain their true values, the waveform is expressed with an accuracy that depends only on the sampling speed and the resolution of the binary number. The advantage of the binary code in this respect is that it has only two conditions, 0 and 1. These can easily be represented by electrical circuits being switched on and off. As long as the digital circuits can detect the difference between these widely different conditions, the string of numbers will be perfectly preserved. Noise in the ordinary sense can be high, yet has no effect, see Figure 1. Each unit of binary code is known as a bit (binary digit).
Further advantages of a digital recording and reproduction system are extremely low distortion, both harmonic distortion and intermodulation, and elimination of wow and flutter.
The low distortion figures are achieved by the high accuracy of the applied analogue to digital and digital to analogue convertors, whereas wow and flutter is eliminated by a Quartz crystal controlled clocking operation.
Because the digital-coded signal consists of a sequence of finite numbers, rather than the continuous and infinitely varying analogue, it is easy to insert extra information or to manipulate the sequence to gain extra advantage, without affecting the main (sound) information at all.
In this way, as in many advanced digital systems, automatic error correction is introduced. This error correction compensates for quite substantial marks on the disc or even temporary lapses in the electronic circuits. So if there are flaws in the system, perfect results are still obtained, because lost information in the system will be automatically restored.
In the CD-system each measured sample of the audio waveform is represented in 16 bits. This offers an extremely high resolution. Each bit in fact, adds a theoretical 6dB to the signal-to-noise ratio to produce a dynamic range much bigger than any previous system has been able to offer.
The audio frequency range that can be covered depends upon the sampling rate. This must be at least double the highest frequency that is to be preserved. At 44.1 kHz the Compact Disc covers easily the full audible range up to 20 kHz. Separation between stereo channels is practically infinite, since the 16 bit audio samples are derived independently for the left and the right channel. The resulting bits are combined ('time multiplexed') into one audio bit steam of 2 x 16 x 44.1 = 1.4112 Mbit/s without any interference between the original audio samples.
The audio bit stream is not recorded directly on CD. First the extremely powerful CIRC (Cross Interleave Reed Solomon Code) is employed to enable correction of all signal drop-outs that may occur later on. Theoretically it will correct an error burst of up to about 4,000 bits, equivalent to a track length of up to about 2.5mm. Even beyond that, it will conceal, by interpolation, for an error burst up to about 12,300 bits, occupying a track length of up to about 7.7mm. This is a very important factor in the CD system, providing a wide margin of tolerance, not only for disc manufacturing, but also for scratches and marks that are practically inevitable with continued use.
Other information that is added to the audio bit stream is concerned with synchronisation and tracking search and find facilities (including random track selection) and can include special symbols for visual display of supporting information such as titles, composers, score or script.
Still, all this is only part of the CD encoding, which is also aimed at increasing the amount of information that can be stored while ensuring accurate timing and eliminating low frequency components, which could upset the operation of the player servo system.
To these ends a process known as EFM (Eight-to-Fourteen modulation) is applied and finally merging bits are added for a further increase of LF-suppression. After this a stream of so-called channel bits at 4.3218 Mbit/s is obtained that is actually recorded on CD.
To accommodate all the features and the processing mentioned, the channel bit stream consists of subsequent groups of bits called frames, each comprising 588 channel bits. Each frame is composed in the following way. Starting point is the stereo information of six audio sampling intervals, i.e. 2 x 6 x 16 audio bits.
These are split in 24 audio symbols of 8 bits, to which 8 parity symbols of 8 bits are added for error correction according to CIRC. Next, one 8 bit control and display symbol is added to yield a total of thirty-three 8 bit symbols.
Each of these is converted to 14 channel bits by the EFM and extended with 3 more channel bits for merging. Finally, a synchronization pattern of 27 channel bits (including 3 bits for merging) completes the frame.
In summary one frame contains: a synchronization pattern of 24 + 3 channel bits; a control and display symbol of 14 + 3 channel bits; two blocks of 12 audio symbols: 2 x 12 x (14 + 3) channel bits; two blocks of 4 parity symbols: 2 x 4 x (14 + 3) channel bits = 588 bits.
Apart from digitalization Compact Disc Digital Audio introduces another very significant innovation - optical readout - using a laser beam. The system is shown in Figure 3.
In essence, a laser is a special light source, producing highly concentrated light. The word 'Laser', in fact, stands for Light Amplification by Stimulated Emission of Radiation, and because of the precise nature of the beam, the laser principle has already found many applications in science and industry.
The Compact Disc laser is a small, low-power semiconductor (aluminium gallium arsenide) unit, emitting invisible infra-red light. Its essential feature is a capacity for ultra-sharp focus, capable of reading a track with pits, only 0.5 um wide, 0.1 um deep, approx. 1 to 3 um long at a track pitch of 1.6um. The beam focus point, in fact, is less than 1 um in diameter (measured between the half-intensity points).
The laser beam is directed through the transparent side of the disc and on to the disc track by the optical system. This is carried on a servo-controlled arm, which tracks radially from the inside to the outside of the disc. To keep the constant linear velocity of 1.3 metres per second the rotation of the disc changes progressively from 500 to 200 rpm.
A second servo controls the position of the objective lens, to maintain ultra-sharp focus regardless of disc warp or any other unevenness in disc rotation.
The recording in the disc is a pattern of pits in a brightly reflective surface. When the beam falls on the flat surface, it is reflected back along the same path as it came. In this path is a semi-reflective prism. On its journey to the disc, 50% of the beam passes straight through the prism. On the way back, 25% is reflected away from the original path to be detected by an array of photodiodes. This is to prevent any light from the laser falling in the photodiode array directly and causing interference.
If, however, the beam falls into a pit on the disc, it is scattered. Very little light returns to the semi-reflective prism and the photodetector. The sequence of flats and pits on the recorded track, therefore, produces a sequence of 'on' and 'off' impulses in the photodetector, thus generating the data stream of channel bits in the player electronics. If the light spot falling on the disc moves off the correct track, the reflected beam is tilted; this causes a balance detector to generate an error correction signal for the tracking servo system.
Because the read-out is optical, the 'pickup' causes no more wear to the recording than reading causes to the words printed on this page. In fact, the reflective track surface is covered with a transparent plastic seal, which affords permanent protection. Scratches, dust and dirt on the outer surface have little effect; the very high opening angle of the objective lens keeps these irregularities well out of focus, and only larger deviations are detected by the photodetector. Furthermore, the CIRC error correction system increases the capability to cope with severe optical distortion arising from dirt or damage.
The optical read-out system of Compact Disc Digital Audio ensures a read-out accuracy, from an extremely high-density digital recording, that no mechanical system, could ever hope to attain.
The signal picked up by the photodetector is a frame structured data stream containing a great deal of information in EFM (eight-to-fourteen modulation) format. The first stage of decoding is to establish clocking (timing), and to retrieve the 8-bit format from the EFM format.
For this purpose, the synchronisation pattern is separated from the control and display symbol and the (audio) data symbols.
Clocking will eliminate small timing errors (jitter) in the data stream, caused e.g. by disc speed variations.
The second stage of decoding is application of error correction and interpolation to the data symbols.
In the third stage, left and right channel data words are demultiplexed (split apart) and separately converted back to analogue to provide normal style stereo pre-amplifier input signals.
Meanwhile, the derived clocking signal is compared with a quartz crystal controlled oscillator reference frequency. Any discrepancy generates an immediate correction signal for the disc motor speed servo system. This servo system, together with the above-mentioned clocking operation of the data stream, makes wow and flutter completely inaudible.
It can be seen that before the signal is returned to its analogue state it is subjected to considerable processing in digital form. Digital processing is entirely a switching operation, with the necessary timing and synchronisation. For the Compact Disc player, with 588 channel bits per frame and a channel bit rate of well over four million bits per second, thousands of instantaneous switching circuits are needed. Only with the development in the last few years, of large scale integrated circuits (LSI's), has it become possible to produce such circuitry within the small dimensions and economics necessary to make Compact Disc Digital Audio a practical proposition.
Like the conventional record, the disc is compression or injection moulded. It goes through the same stages of pre-mastering, mastering and replication, but the production process is different in many respects, because the final product is of a much higher technological level. Several of these steps even have to be performed under clean room conditions. Similar to those required for Integrated Circuit production.
The audio recording and sub-coding, of course, must be encoded into the characteristic CD frame format, including synchronisation and error correction.
From an approved CD-Tape Master, the CD-Disc Master is produced by the disc mastering process. A glass disc, optically ground, polished and spotlessly clean is coated with about 0.1 um photoresist evenly distributed by a spin-coating technique. This forms the Resist-Master Disc (comparable with a photographic film) for the recording process.
The encoded digital information is recorded (cut) on a CD-Disc Master Recording System. The high power recording laser, modulated by the signal from the CD-Tape Master, writes a pattern of pits in the photoresist. The exposed parts are etched away (developed) to generate the final pit structure. After a silvering process, the CD-Disc Master results, carrying the actual pit structure required in the finished discs.
At this stage, the CD-Disc Master can be transferred to the replication facilities (disc production plant).