A compact disc is a single groove, about five kilometres long, coiled into a spiral so fine that forty of its turns would fit across a human hair. Along that groove sit microscopic pits. A laser follows it from the middle outwards, watching the reflection flicker. That is the entire physical idea — and it is the least interesting part of the story. The genuinely clever engineering is what sits on top: several layers of mathematics whose only job is to assume the disc will get damaged, and be right anyway.
One groove, five kilometres long
Unlike a vinyl record, a CD is read from the centre outwards, and unlike a hard disk it has no separate concentric tracks — it is one continuous spiral. The numbers are worth sitting with for a second:
- Track pitch: 1.6 microns. That is the gap between one turn of the spiral and the next. A human hair is around 70 microns across.
- Pit width: about 0.5 microns, with the shortest pits roughly 0.83 microns long.
- Total track length: around five kilometres, wound into a disc you can hold in one hand.
The disc spins at a changing speed — roughly 500 rpm when the laser is near the centre, slowing to about 200 rpm at the outer edge. This is deliberate: it keeps the track moving past the laser at a constant linear speed, so a pit near the rim takes exactly as long to fly past as one near the hub. Every part of the disc is therefore equally dense, which is why a CD holds more than it would if it spun at a fixed rate like a hard drive.
Pits and lands — and why they aren’t ones and zeros
The flat areas between pits are called lands. Almost everyone assumes a pit is a 1 and a land is a 0. Almost everyone is wrong, and the real answer is much neater.
A 1 is a change. Every time the surface transitions — land into pit, or pit back into land — that edge is a 1. Everything in between, however long the run, is a string of 0s.
Why do it this way? Two reasons, and both are about honesty over ambition. First, detecting a change in reflected light is far easier and more reliable than measuring an absolute level, which drifts as the laser ages, the disc wobbles or a smudge dims the beam. Second, it lets the player keep time. There is no separate clock track on a CD; the drive recovers its timing from the data itself by watching how often edges arrive. Long featureless stretches would leave it drifting — which, as we’re about to see, the encoding is specifically designed to prevent.
Why the pit is exactly that deep
A pit is about 125 nanometres deep, and that figure isn’t arbitrary — it is a quarter of the laser’s wavelength as it travels through the plastic. The laser is 780 nm in air, but slows and shortens to roughly 500 nm inside the polycarbonate.
Here is why that matters. The laser spot is wider than a pit, so when it sits on an edge, part of the beam bounces off the pit floor and part off the surrounding land. The pit half travels a quarter-wavelength down and a quarter back — half a wavelength further in total — and returns exactly out of step with the other half. The two halves cancel each other out. The photodiode sees a sharp drop in returned light.
So a CD does not read shadows. It reads interference. The pit isn’t dark; it is precisely the wrong depth, on purpose.
The disc can’t store your bytes directly
You cannot simply write your data onto the spiral. Real data contains patterns the optics can’t physically handle — long runs of identical bits, or edges packed closer than the laser spot can resolve. So every byte is translated first, by a scheme called eight-to-fourteen modulation (EFM).
Each 8-bit byte is swapped for a 14-bit pattern from a lookup table, chosen so that every pattern has at least two and at most ten 0s between consecutive 1s. In physical terms: no pit or land is ever shorter than the optics can see, and none is ever so long that the drive loses its timing. Three further merging bits are added between patterns so the rule still holds across the joins.
Do the arithmetic and it is slightly absurd: 17 bits go onto the disc for every 8 bits of your data — before any error correction is added. That is the first hint of this article’s real theme. An enormous fraction of a storage medium is not your data. It is the scaffolding that keeps your data true.
The clever bit: surviving a scratch
Now the good part. CDs use CIRC — Cross-Interleaved Reed–Solomon Code — and it is one of the most elegant pieces of engineering in consumer electronics.
It combines two ideas:
1. Reed–Solomon parity. Devised in 1960 by Irving Reed and Gustave Solomon, this adds extra parity bytes to each block, computed so that if a few bytes go missing or wrong, the originals can be reconstructed by algebra. It is not a copy — it is closer to how knowing a total and all-but-two of the numbers lets you work back to the missing ones. A CD runs two such stages, called C1 and C2.
2. Interleaving. This is the trick that does the heavy lifting. Before the data is written, consecutive bytes are deliberately scattered along the track, so bytes that belong together end up millimetres apart on the disc.
Consider what that means for damage. A scratch wipes out a physically continuous run of the spiral — thousands of consecutive bits. To any single error-correcting code, that is a catastrophe: a whole codeword gone at once. But because the data was interleaved on the way in, the player de-interleaves on the way out, and the catastrophe unravels. That one long gouge becomes one or two missing bytes in each of hundreds of different codewords — and losing two bytes out of a block is exactly the kind of small, boring problem Reed–Solomon eats for breakfast.
The damage isn’t repaired so much as diluted until it becomes repairable.
In practice, CIRC fully corrects a burst error of roughly 3,500 bits — about 2.4 mm of track, comfortably more than a typical scratch or a fingerprint. Beyond that, an audio CD falls back to concealment: it interpolates a plausible waveform across a gap of up to about 8.5 mm, and your ear never notices. (That fallback is also what audio CD copy protection later exploited — see why a copy-protected CD can’t be copied exactly.)
Music can fake it. Your spreadsheet can’t.
That last point matters more than it sounds. A wrong sample in music is an inaudible smudge. A wrong byte in an invoice, an executable or a database is corruption — and interpolation would be actively dangerous, because it invents data that looks fine.
So the CD-ROM standard adds a third protective layer on top of CIRC. Each sector on the disc physically holds 2,352 bytes, but a data disc gives you only 2,048 usable bytes per sector. The other 304 go to a sync pattern, a header, a 32-bit error-detection code that verifies the sector, and 276 further bytes of parity for a second round of correction.
Stack it all up and the ratio is striking:
| Layer | What it does | Cost |
|---|---|---|
| EFM | Reshapes bytes into pit lengths the optics and the clock can handle | 8 bits become 17 |
| CIRC | Reed–Solomon parity plus interleaving — turns one scratch into many trivial errors | 8 parity bytes per 24 of data |
| Sector ECC (data discs only) | A second correction pass, so a wrong byte is never silently accepted | 2,352 physical bytes yield 2,048 |
A disc that physically carries around 2 GB of pits and lands hands you roughly 650 MB. Only about a third of a CD is your data. The rest is the reason the third that matters can be trusted.
Pressed, burnt, and the ones that die
All of the above describes a pressed disc — the kind you buy. Its pits are physically stamped into polycarbonate from a metal master, then coated with a thin aluminium reflective layer and a protective lacquer. It is, essentially, a fossil: the data is a shape.
A CD-R you burn at home is a different animal. There are no real pits. The blank carries a layer of organic dye, and the writing laser heats spots of it until they darken and stop reflecting properly. The result reads like pits, but chemically it is a stain — and stains fade. Heat, humidity and ultraviolet light all degrade the dye. A CD-RW is different again, using a phase-change alloy the laser flips between reflective-crystalline and dull-amorphous states, which is what makes it erasable and also why it is the least reflective and fussiest of the three.
Then there is disc rot. If the lacquer is scratched, porous or poorly applied, air reaches the aluminium and oxidises it; the mirror turns transparent in patches and the data underneath simply isn’t there any more. Held up to a light, a badly rotted disc has pinholes you can see through.
Honest lifespans are genuinely unknown. Manufacturers quote decades; real-world CD-Rs have been observed failing in under five years in a hot car or a sunny window, while pressed discs from the 1980s still play. The dangerous part isn’t the average — it is that a dying disc gives no warning. Error correction hides the decay right up until the moment it can’t, and then the disc goes from “fine” to “unreadable” with nothing in between.
The practical lesson
Optical discs are distribution media, not backup. If your only copy of anything — client files, photos, old accounts — is on a burnt disc in a drawer, it is not backed up; it is on a countdown you can’t read. The same applies to a single external drive. Real protection means more than one copy, in more than one place, and — the step nearly everyone skips — something that actually verifies the copies are still readable.
Where else this shows up
Learn how a CD works and you have quietly learned how a great deal of modern technology works, because the same two ideas — add parity, then scatter the data so damage can’t concentrate — are everywhere.
- QR codes use Reed–Solomon too, which is why one still scans with a logo covering the middle or a corner torn off.
- Spacecraft lean on it hardest of all: the Voyager probes sent images across billions of kilometres through a whisper of signal, and error-correcting codes are the reason those images arrived intact.
- ECC memory in servers corrects single-bit flips in RAM using a related family of codes.
- RAID and modern filesystems apply the identical logic at the drive level — parity plus distribution, so one dead disk isn’t one dead array.
And the optical family itself just kept refining the same trick with a finer pen:
| CD | DVD | Blu-ray | |
|---|---|---|---|
| Laser | 780 nm infrared | 650 nm red | 405 nm blue-violet |
| Track pitch | 1.6 µm | 0.74 µm | 0.32 µm |
| Shortest pit | 0.83 µm | 0.4 µm | 0.15 µm |
| Capacity per layer | ~650 MB | 4.7 GB | 25 GB |
A shorter wavelength focuses to a smaller spot, and a smaller spot can read smaller writing. That is very nearly the whole difference between a disc that holds an album and one that holds a film in high definition.