PART 5: Following on from last month's look at digital tape recording formats, Hugh Robjohns turns the spotlight on the techniques and technology of disk‑based recording. This is the fifth article in a six‑part series.
If you read last month's article, you will recall that there are two fundamental approaches to the design of tape‑based systems: rotary head and stationary head. There is a similar dichotomy within disk‑based systems, this time between optical and magnetic formats. There is also an increasingly significant third option which represents a blend of the two.
The most obvious digital disc format relying on optical technology is the CD — something which we all tend to take very much for granted fifteen years after its launch. However, it is worth revisiting the system as many of its concepts enable understanding of the newer formats.
CDs are 'pressed' in much the same way as vinyl records although the dimensions are obviously smaller and the tolerances much finer. The audio data is encoded along with timing and various other pieces of information as either pits or flats along a continuous spiral groove which starts close to the centre of the disc and works outwards at a fixed rate. The idea of starting on the inside edge of the disc was to allow production of discs with different diameters: the laser system would always start at the same point, but continue outwards until it found an 'end‑of‑disc' instruction. In the early days, for example, we had three‑inch CD singles. However, these never really caught on and it has subsequently proved cheaper to release singles on standard‑sized discs.
Once the blank polycarbonate disc has been stamped, the surface carrying the audio data is coated with a microscopically thin layer of metal. Usually, this is aluminium, although gold and silver are also used. This is then protected from oxidisation by 30µm‑thick lacquer, which also carries the screen‑printed labelling. Although CDs are extremely tough, damage to the lacquer layer usually means the disc no longer plays properly, so never place them label‑side down. Ideally they should be put back in the caddy, or, failing that, it is safer to place them playing‑side down. This might sound odd, but because the laser is focused onto the aluminium surface through 1.2mm of plastic, small scratches and marks on the 'playing' surface become insignificant (the laser beam is more than 1mm wide on the disc surface, but focused to 1.7µm on the reflecting layer). It is the same concept as not noticing the squashed bugs on the windscreen of your car while your eyes are focused on the cars in front. Small, troublesome scratches on the disc can usually be polished out quite successfully with toothpaste or the finest version of T‑Cut car body paint restorer. Be warned that it takes time and patience. If it doesn't work, don't blame me or SOS.
Reading data from a CD is a marvel of modern technology and I am still impressed every time I think about it. The pressed pits in the disc are seen as raised bumps from the playing side and they are arranged to be 0.125µm high — a quarter of the wavelength of the 780nm infra‑red laser light. To give you some idea of what that means, if the disc was scaled up so that the bumps were 1cm high, the disc would be around 10km across (just over six miles)!
A monochromatic and coherent light source is essential to 'read' the disc — the light must be at one frequency and in the same phase. The laser is focused to a spot about twice as wide as the bumps and therefore when one is encountered, light is reflected from both the top of the bump and the area around it. Since the bump is a quarter wavelength high, the light reflected from its top travels half a wavelength less than the light reflected from the surrounding surface, and is therefore out‑of‑phase with it, resulting in cancellation. The pressing tolerances mean that the bumps are unlikely to be exactly a quarter‑wavelength high, but as long as they are close to that value, partial cancellation will result. The photo‑detector in the optical pick‑up simply has to be able to recognise the difference between lots of reflected light from the flat areas and the dimmer light reflected when a bump is present (see Figure 1).
Unlike vinyl records which rotate at a constant speed (33 or 45rpm), the rotational speed of a CD varies so that the laser passes over the data at a constant rate of 1.4 metres/second (1.2m/S in the case of 80‑minute‑plus discs). This means a rotational speed of 500rpm at the start of a disc, falling to around 200rpm at the end. It has to work this way because the size of the bumps is related to the wavelength of the laser light and cannot be changed. The mechanism for controlling speed is very simple. As the data is extracted from the disc, it is stored in a buffer memory before being passed on at the correct sampling rate to the decoding circuitry. If the disc is spinning too fast the memory fills up: if the disc is spinning too slowly the memory empties. A simple feedback system based on the memory capacity is used to control the spin‑motor servo. The precise rotational speed of the disc is not important provided the data comes off at an average of 44,100 samples per second — hence the acceptability of linear speeds between 1.2 and 1.4m/S.
Embodied within the data on the disc are eight sub‑code (auxiliary data) channels: P, Q, R, S, T, U, V and W. Of these, the most important is the Q sub‑code as this carries the Table of Contents (TOC) at the start of the disc which says how many tracks there are, their timing information, running times, track and index identification, and copy‑prohibit and pre‑emphasis flags throughout the disc. The P sub‑code, which provides a very simple means of locating the start of each track, was originally intended for 'dumb' portable players, but it is not particularly important for most machines these days. The rest of the sub‑codes were unspecified in the Red Book (see the CD Colours box) and hence they are rarely used, although they have found applications in remote projector control for AV presentations, karaoke lyrics and other still graphics (the CD+G format), and computer data.
The CD‑R is a WORM disc (write once, read many). Once an area has been recorded, it cannot be erased and re‑recorded. There are no editing facilities, although discs can be recorded in stages with one or more tracks at a time up to the Red Book CD limit of 99 tracks. The prices of CD‑R recorders and their blank media are similar to those of high‑end cassette machines (£500 for CD recorders and less than £2 for a 74‑minute blank), so they are becoming increasingly common as demo and master recorders in studios.
The construction of a CD‑R disc is slightly different to that of a standard CD, featuring an additional layer in the disc construction. On top of the polycarbonate substrate (which is stamped with a 'pre‑groove') a yellow/green dye layer precedes the reflecting layer (usually gold) before the standard lacquer and label printing (Figure 2).
The principle behind the CD‑R is that if a high‑powered laser (about 10 times stronger than a normal CD player) is focused on the dye layer, the green dye will absorb the red light and get hot. At around 250?C the dye deforms and shrinks to become much more dense and the substrate swells to fill the space, neither of which can be reversed. When played in a conventional CD player, the laser light is reflected from the gold layer in the usual way, but where the dye has deformed, less light is returned, thus resembling the appearance of a bump on a conventional CD. The only snag is that because the light has to pass through the dye layer twice (ie, both on the way to the gold reflector and back again), the overall reflectivity of CD‑Rs is lower than a normal CD and thus some older players are not sufficiently sensitive to read CD‑Rs reliably.
Since the dye inevitably varies from batch to batch, CD recorders first perform what's called an optimum power check on each new disc to find out just how much power is required to deform the dye in the optimal way. When a new blank disc is inserted, the machine's display shows 'OPC' (or something similar) while the machine makes a series of test recordings with different laser powers in a dedicated area near the centre of the disc. The disc is then replayed to ascertain the best setting and the result is stored for reference in the Program Calibration Area (PCA) — a 'Regulo 6 for 20 minutes' kind of thing!
Once the OPC is complete, the display shows the available recording time, obtained from the data encoded in the pre‑groove of the disc. The pre‑groove is a wondrous thing, fulfilling several functions in a very elegant manner. It is cut at a precise pitch to define the spacing between adjacent turns of the spiral (specified as 1.6µm) which is critical to the correct tracking of the laser beam for both recording and replay. The pre‑groove also has a small 'wobble' from side to side as it runs around the disc. When the disc is spinning with a linear speed of 1.4m/S, the wobble causes the tracking servos to oscillate at 22.05kHz (half the sampling rate), thus enabling the machine to maintain the correct linear disc speed throughout the recording. The wobble is itself wobbled in such a way that timing information can be encoded (known as ATIP or Absolute Time in Pre‑groove) and this in turn allows the machine to know how much recording time is left and where the laser is within the disc — much the same as the Q sub‑code information on a standard CD (Figure 3).
When a recording is made on a CD‑R the audio data is recorded directly to the main data area of the disc, but the TOC information (track number, duration and so forth) is stored in a temporary area near the centre of the disc. This means that a partially recorded CD‑R cannot be played on a conventional player as there is no recognisable TOC from which it can ascertain the number of tracks and where they start. However, a CD‑R can be made playable by going through a process known as 'Fixing Up'. This translates the temporary table of contents into a Red Book‑compatible version which is recorded in the lead‑in area before the audio data section of the disc. The process also writes an end‑of‑disc scroll after the final track. The whole process can take several minutes and once done, no further recordings can be made to the disc. To a CD player, the CD‑R now appears to be perfectly normal, albeit with low reflectivity.
The specifications for the CD‑R, laid down in the Orange Book, permit the use of Skip IDs which allow unwanted tracks on the disc to be passed over during replay. However, conventional Red Book players do not recognise skip instructions, so this facility is of very limited use. If you make a mistake when recording a CD‑R, I would advise ejecting the disc, adding it to your beer mat collection, and starting again!
CD‑RW or CD‑Rewriteable discs use a different technology called Phase Change. This is a reversible recording process taking advantage of a material which has two stable but very different states. Life expectancy of the disc is between 1000 and 10,000 recording cycles, a factor of a thousand worse than professional MO disks and MiniDiscs (described later). However, as an essentially domestic product, the CD‑RW disc is perfectly adequate for typical audio and home computer applications, even though the blank media are ten mores more expensive than write‑once CD‑Rs.
In its original state, the recording layer of the CD‑RW disc is polycrystalline, During recording, a high powered laser is used to change areas of the disc into an amorphous phase of the material. The amorphous areas have much lower reflectivity than the crystalline areas, so audio data is recovered as bright or dark areas, just as with CD and CD‑R. When the disc is over‑written, the amorphous areas can be returned to the crystalline phase by using lower intensity from the laser. CD‑RW discs exhibit even lower reflectivity than CD‑Rs and cannot be replayed in conventional CD players at all.
Long in gestation, the DVD has finally arrived and is now available as a format for video movies and for some computer applications. The audio‑only version is still being held back because various hardware and software companies have failed to agree on a specification. Technology is available for both recordable and re‑recordable versions of the DVD, but this is also being delayed to allow for greater penetration of the format into the market place.
DVD started life as two similar, but incompatible formats: the Sony/Philips Multimedia CD (MMCD) and the Toshiba/Matsushita/Time‑Warner alliance format of the Super‑Density Digital Video Disk (SD‑DVD). Fortunately, common sense prevailed and the two formats were combined into DVD, which is now appearing on the shelves of the bigger video retailers across Europe.
DVD is nothing more than an increased density version of CD and takes advantage of improvements in CD manufacture and replay technology which have been introduced over the past 15 years. The disc structure is slightly different, comprising a pair of stamped, 0.6mm thick substrates, glued together with the data surfaces in the centre of the composite disc.
In order to accommodate smaller data bumps on the disc surface, the DVD laser operates in the visible‑red at 650nm instead of the 780nm infrared of conventional CD players. This reduces the wavelength and allows the bumps to be smaller, but it also requires different optical arrangements which force the data layer to be brought closer to the pick up. A disc only 0.6mm thick is not sufficiently robust and hence the idea of gluing two together to make a composite as strong as a normal CD.
As the composite disc has two pressed surfaces, it can be made single‑ or double‑sided (although double‑sided versions don't leave anywhere for the label). There are techniques available to build up two data layers on each surface using a semi‑transmissive reflector and a second data layer. The laser can be focused on either layer as required.
In terms of data storage, not only are the bumps smaller (in height, width and length), but the spacing between adjacent turns of the spiral is also halved. Consequently, a single‑layered disc provides 4.7Gb of storage capacity as opposed to the measly 650Mb on a standard CDs. A dual‑layer disc offers 8.5Gb which is enough to store the equivalent of about 1.5 million A4 pages of text (a pile of paper about 700 feet high) — rather more impressive than a standard CD‑ROM which can hold 'only' 95,000 pages. A dual‑layer, double‑sided disc could potentially store 18Gb of audio, video or data.
The linear speed of the DVD has also been increased from the 1.4m/S of a CD to 4m/S in order to achieve workable data transfer rates for real‑time video (albeit with MPEG data reduction) of 1.1Mb/S. This compares with 153Kb/S from a CD at standard speeds.
There are other enhancements built into the DVD format such as a different channel coding structure called EFM Plus and revised error protection which is around 10 times more robust than that of conventional CDs.
Pre‑recorded MiniDiscs are made in exactly the same way as a normal CD, even down to the channel coding and error protection systems. The only difference is that the disc is just 64mm (2.5 inches) in diameter, although the audio information is data‑reduced by the ATRAC process to allow a full 74 minutes of replay time.
One of the recognised failings of the CD is its lack of resistance to damage on the playing surface which can result in tracking problems and replay glitches. Rather than redesign the error protection systems, the MiniDisc designers simply encased the disc in a plastic caddy to reduce the likelihood of scratches and marmalade reaching the disc surface! Re‑recordable MiniDiscs will be covered in the section on magneto‑optical discs.
The alternatives to optical formats are formats that use magnetic technology similar to conventional magnetic recording on tape. However, erase heads and recording bias are not required as the digital media is fully saturated N‑S or S‑N. Linearity is also not required and the signal‑to‑noise ratio is minimal (about 10dB). Unlike tape recorders, a hard disk unit involves no contact between the record/replay head and the media surface — the head 'flies' just above the surface of the rapidly spinning disk on a cushion of air, so media wear is insignificant.
Depending on the design and capacity of a hard drive, there may be a number of aluminium disks within the unit, each coated with a magnetic layer and mounted on a common drive spindle. Each surface has its own record/replay head which is mounted on some kind of shared, movable arm assembly allowing the heads to be positioned accurately for record and replay of data (Figure 4). The main advantages of hard drives are the phenomenal storage capacity, rapid access times and high data transfer rates — all well in excess of any other medium currently available.
Data is organised on the disk according to the operating system of the controlling computer. That is the major reason for the incompatibility of disk drives employed in different audio recording and editing systems. However, the basic storage structure all relies on tracks, sectors, blocks, and cylinders. The disk is divided into concentric rings called tracks, each of which is sub‑divided into sectors, and within each sector, data is grouped into blocks or clusters. A vertical column of tracks across all disk surfaces is called a cylinder, and the positioning of data on the disk surfaces is logged in a special directory (known as the File Allocation Table or FAT in Microsoft‑speak), without which the stored data is meaningless.
At present, hard disk storage capacity is doubling (and the cost almost halving) roughly every three years. While this situation can't go on forever, the hard disk is likely to remain the most cost‑effective rapid‑access storage medium for some time to come.
Magneto‑optical (MO) disks, which derive from the computer industry, combine both magnetic and optical principles. Their advantage is portability and re‑recordability, combined with a virtually unlimited life.
The earliest systems were relatively slow compared with hard disks, both in terms of their access times and transfer rates, but the technology has improved considerably in recent years, and the latest generations of MO disks are certainly closing the gap.
The basic operating principle is that data is stored by a photo‑polymer layer within a glass or plastic substrate. For the chemists among you, the polymer is usually something like ferri‑terbium‑colbalt (FeTbCo), a substance which exhibits a property called the Kerr effect. This material is sandwiched between a reflecting layer and protective, heat shield layers, and when heated to a high temperature known as the Curie Point (between 185 and 250?C depending on its exact composition) its crystalline structure becomes flexible and it can be altered between two stable states. The heating is achieved by a powerful laser (much like the CD‑R) and the material's structure can then be changed by the application of a weak magnetic field. By switching between N‑S or S‑N fields, the required data can be stored in the physical structure of the polymer. Once cool, the material is perfectly stable with the data safely locked in place.
...a billion continuous replays of a one‑second segment of audio will take more than 31 years.
When polarised light is passed through the photo‑polymer, its angle of polarisation is changed slightly, with the direction and amount depending on the magnetic field the material was exposed to. A suitable photo‑detector, sensitive to light polarised in one specific direction, can be used to recover the stored data (bright where the polarisation matches that of the detector and dark where the polarisation is altered).
The life of a computer‑standard MO disk is usually of the order of a million record passes and a billion replay passes — in fact, the bearings will give out before the polymer does. To give some practical meaning to these figures, a billion continuous replays of a one‑second segment of audio will take more than 31 years.
The earliest computer MO drives were relatively slow, partly because of the way data was recorded. Typical systems required a three‑pass approach: first, the whole disk was bathed in a fixed magnetic field and the sector to be recorded heated with the high‑power laser. This effectively formatted the sector. Next, the magnetic field was reversed and specific data cells heated to change their state, thereby storing the required data. A third pass then verified the stored data. Due to this tedious process, recording took about twice as long as replay — a significant drawback in audio devices!
The big advance in MO disks has been the development of the direct‑overwrite system. This is used on MiniDiscs, as well as the latest computer MO systems, and is also known as LIMDOW (light‑intensity modulation direct‑overwrite).
The system abandons the need for the initial formatting by modulating the magnetic field with the required data directly. The process involves heating the required sector of the disk while modulating the magnetic field to encode the data. A verify pass then confirms that the correct data is in place. It may not sound much of an improvement, but it is far faster and represents a significant step forward, allowing MO drives to take on a far more practical role in audio editing and recording systems.
Current 130mm MO disks are available in 1.3, 2.6 and 5.2Gb capacities, but disks with capacities of 7Gb up to 11Gb are being developed and there are already 640Mb versions of the 64mm MiniDiscs.
Although it was originally designed to carry audio data, other industries were quick to pick up on the CD format. We now have a range of inter‑related formats, each identified by a different 'colour'. The basic audio disc is defined by the Red Book, with the Yellow Book dictating the specifications for the CD‑ROM. This provides storage of either computer data or audio and video data, and has an extra degree of error protection in place of the interpolation strategies of audio‑only systems.
The Yellow Book standard has been enhanced with the XA extension allowing audio and video data to be interposed with computer data on a sector‑by‑sector basis. This is the basis of the Kodak PhotoCD standard as well as the Green Book and White Book standards. The Green Book relates to Philip's proprietary CD‑i format — a home entertainment system which has met with limited success. The same can also be said of the full‑motion video format (using MPEG‑1 data compression) covered by the White Book.
Writeable CDs (CD‑Rs) are described by the Orange Book. This allows recordings to be made on a disc, either all at once or on a piecemeal basis — hence the term multi‑session.
One of the biggest problems with CDs is that of mistracking, resulting in that 'stuck in a groove' effect we have all experienced. This is normally caused by tracking problems and has nothing to do with the error correction system which was essentially designed to cope with 'pin‑holes' in the reflective layer which were a common problem with early discs. Try holding your oldest CD up to the light while looking at the playing side and see if there are any tiny pinpricks of light shining through. The error correction system of the CD format was designed to handle the complete absence of data caused by these kinds of gaps — a total gap of up to 2.5mm, which represents around 4000 bits of data, can be corrected perfectly.
Unfortunately, as we now know, holes in the reflecting layer are not the main problem with CDs. That honour goes to deep scratches and surface contamination by materials with optical properties such as grease... and marmalade! (So stop bringing your CDs to the breakfast table — Ed) What tends to happen is that the laser is happily tracking along the data when it comes across a scratch or patch of grease. The beam is then deflected (refracted) to some other part of the disc, typically a couple of tracks earlier or later. This does two things. First, it confuses the error detection system because the interleave structure of the data goes haywire. Second, the deflected beam probably doesn't fall exactly on the 'new' track, so the tracking servo has to re‑align the beam.
With the error detection system completely confused, raw errors are going to slip through. Hence the audibility of clicks and splats because the interpolation and muting mechanisms are also both helpless. But worse is the chaos caused to the tracking servo. No sooner has it caught up and corrected the laser beam's alignment on to its new track, when the scratch or grease passes and the beam reverts to the original part of the disc, whereupon it is now misaligned with the original track. And so it all goes horribly wrong — tracking servos go into oscillation, error detection systems go home in disgust, the same bit of music is repeated over and over, and you end up hitting the box and wishing for a solid state recorder to be invented. (They have been, by the way, but they are expensive and don't play for very long... yet!)