An exact copy of a copy-protected audio CD is impossible when the protection works by deliberate errors, because part of the music was never written to the disc correctly. At chosen points the disc holds wrong numbers, arranged so that its own error correction can spot them but not repair them. A CD player covers each one with a guess. A copy can only contain the wrong numbers or somebody’s guess, and there is nothing on the disc to check either against.

What an exact, verified copy normally means

An audio CD is a long list of numbers: 44,100 measurements of the sound every second for each of two channels, each one a 16-bit value. Copying a disc properly means getting that list off it with every number intact, and then proving it.

The proof comes in two parts. A careful ripping program reads each stretch of the disc more than once and checks that the reads agree. Then it compares a checksum of the result with a public database of checksums from other people’s rips of the same pressing, made on thousands of different drives. When strangers with different equipment all arrive at the same checksum, the list of numbers is right, and it can be stored in a lossless file that will stay right.

Both checks rest on one assumption: that the right numbers are on the disc to be found.

Correction recovers. Concealment guesses.

Every CD expects to be damaged. The music is stored with extra parity bytes and scattered along the track, so that a scratch becomes a sprinkling of small errors that can be rebuilt by algebra. Our lesson on how a CD stores data covers how. When the decoder chip in a player meets a suspect stretch, it works through three stages.

  1. Correct it. If the damage is within what the parity can rebuild, the original bytes are recovered exactly. Not approximately: a corrected disc and a flawless disc produce identical numbers.
  2. Flag it. If there is too much damage, the decoder can tell the bytes are wrong but can’t work out what they should have been. It marks those samples as unreliable.
  3. Conceal it. A player can’t stop the music to think, so it replaces each flagged sample with a made-up value and carries on.

The first stage is error correction. The third is error concealment, and it is a different thing entirely. Correction gives back the truth. Concealment gives back something plausible.

Data discs don’t allow the third stage at all. A guessed byte in a spreadsheet is corruption, so a CD-ROM adds another layer of checking and correction and reports a failure if that isn’t enough. Audio CDs were given concealment instead, because a guessed sample in a song is almost never heard.

How the firmware guesses

Concealment is built into the decoder chip and firmware of every CD player and every computer disc drive. The commonest method is interpolation: a missing sample is replaced by the average of the good samples on either side of it. The CD format was designed to make that possible. Odd and even samples are stored apart, so a failure tends to take out every second sample and leave its neighbours standing.

For a longer gap there are no neighbours to average, so the firmware holds the last good value. For a longer one still it fades to silence, which is the dropout you hear on a badly scratched disc.

Interpolation: a player guessing one missing sample Nine audio samples follow a smooth curve. The fifth has been flagged as unreliable. The disc holds a wrong value far above the curve. The player ignores it and puts in a value halfway between the fourth and sixth samples, which lands close to the true sample but not on it. on the disc: a wrong value the true sample the player’s guess
One sample, three versions. The decoder knows the stored value is wrong, throws it away and puts in the midpoint of its two neighbours. The guess is close to the truth and is not the truth. The gap between them is exaggerated here.

On music, interpolation works very well. Sound is smooth at the scale of one 44,100th of a second, so the halfway point between two neighbours is seldom far from the real value. But it is a guess. The number that comes out was invented by the player, and a different player with a different chip may invent a slightly different one.

What copy protection did with that

In the early 2000s record labels tried to stop CDs being copied on computers. The most widely used scheme was Cactus Data Shield, from an Israeli company called Midbar, later bought by Macrovision. EMI and BMG sold discs carrying it under the Copy Control name from about 2001 to 2006, in Australia among other countries.

Its central trick, set out in the company’s patent (US 6,425,098), turns the concealment stage against copying. When the disc is mastered, selected samples are overwritten with wrong values, while the error-correction codes are left as they were calculated for the right values. The decoder finds a mismatch it can’t resolve, flags the sample, and the player interpolates.

The samples aren’t chosen at random. The patent picks the ones whose true value already sits close to the average of their two neighbours, and ones that fall in near-silent passages. Those are exactly the samples interpolation guesses almost perfectly. In a CD player, the listener hears nothing wrong.

The damage is aimed at the copy. A computer drive that hands over the stored values without concealing them produces a file with the wrong values in it. Burn that file to a blank disc and the burner calculates fresh error-correction codes for whatever it is given. Now the wrong values are certified as correct. A player has no reason to conceal them and plays them as written: a click at every one.

The same wrong value on the original disc and on a copy Two rows of four steps. On the original disc the wrong value comes with codes that say it is an error, the decoder flags it, the player fills in a guess and it sounds right. On a copy the same wrong value comes with fresh codes that say it is fine, the decoder sees nothing wrong, the player plays it as written and the result is a loud click. The original disc, in a CD player Wrong value, codes say: error Decoder flags the sample Player fills in a guess Sounds right A copy burnt from a computer rip Same wrong value, fresh codes: fine Decoder sees nothing wrong Player plays it as written A loud click
The protection is the mismatch. On the original, the wrong value travels with codes that expose it. A burner writes new codes that vouch for it.

The scheme had other parts: a misleading table of contents and a second session on the disc, both aimed at confusing computer drives. Other products of the time left the audio alone and relied on software that installed itself on Windows, the best known being the XCP software behind the Sony BMG recall of 2005. Those interfere with reading the disc, not with what is stored on it. This lesson is about the deliberate errors, because they are what makes an exact copy impossible in principle.

None of these discs was a Compact Disc as the standard defines one. Philips, which created the format with Sony and licenses its logo, said that such discs broke the specification and were not entitled to carry the Compact Disc Digital Audio mark.

Why the copy can’t be exact

Ask what an exact copy would be a copy of, and there are three candidates. None of them can be had.

  • The music as mastered. At each altered point, the disc doesn’t hold enough information to work out what the true sample was. The error correction has been pushed past what it can rebuild, on purpose. The right numbers exist in the label’s studio master and nowhere on the disc.
  • The disc as pressed. A true clone would reproduce the wrong values together with their mismatched codes, so that a player would flag and conceal them just as it does on the original. Ordinary hardware can’t do that. A drive gives the computer the output of its decoder, not the raw pattern of pits, and a burner always calculates its own codes for whatever it writes.
  • The disc as played. What comes out of a player is the true samples plus a scattering of guesses, and the guesses depend on the player. There is no single “played” version to copy.

Why the copy can’t be verified

Even a copy that sounds perfect can’t be shown to be right, because each of the usual proofs fails.

  • Reading twice proves less than it seems. Deliberate errors are stable: they are in the same place on every read. A drive that conceals them produces the same guess each time, and two identical reads look like success. They show the drive is consistent, not that it is correct. A stricter ripping program asks the drive which samples its decoder flagged, sees the same flags on every pass, re-reads until it gives up and reports the spot as unreliable.
  • Comparing with other people’s rips fails too. The database holds whatever each contributor’s drive produced. Drives that conceal differently produce different numbers, so one track can sit in the database under several checksums, or under none that agree. If yours matches someone else’s, you have learnt that two drives guessed alike.
  • The disc carries no answer sheet. A data CD stores a checksum with every sector, so a drive knows whether it read that sector correctly. An audio CD has nothing of the kind. Its only self-check is the error correction, which is the very layer that has been tampered with.
One disc, three drives, three different rips A single disc is read by three drives. Drive A averages the two neighbours of each flagged sample, drive B repeats the last good sample, and drive C passes the wrong value through. Each produces a file with a different checksum. One disc Drive A averages the two neighbours Drive B repeats the last good sample Drive C passes the wrong value through Checksum A Checksum B Checksum C Three files that disagree, and nothing on the disc to say which is right
No reference to settle it. On an ordinary disc all three drives recover the same numbers and the checksums match. Here each drive fills the gaps its own way.

There is one more catch. The concealment happens inside the drive, before the computer sees anything. Unless the drive reports its flags, and reports them accurately, the ripping program receives a tidy list of numbers with nothing to show which ones were invented.

How much does it matter?

To the ear, with the original disc in a good player, hardly at all. The altered samples are few and were picked for being easy to guess. A rip made on a drive that conceals well will sound the same as the disc.

What is lost is certainty. The copy is close, not exact, and nothing can show how close. On a drive that doesn’t conceal, it is a file full of clicks. The deliberate errors also spend some of the disc’s safety margin: a real scratch landing near one has less error correction left to absorb it. And these discs were known for misbehaving in car stereos, DVD players and games consoles.

A scratched disc does the same thing, with one difference

A bad scratch on an ordinary disc also ends in concealment, and a rip of that stretch can’t be verified either. The difference is that a scratch is a marginal fault. The right data is still underneath it, and a second read at a different speed, a different drive or a careful clean will sometimes get it. Deliberate errors are pressed into the disc. No amount of re-reading finds data that was never written.

How to tell if a disc has it

  • A Copy Control logo, or wording such as “copy protected” or a warning that the disc may not play on computers, on the back cover or the disc itself.
  • No Compact Disc Digital Audio logo anywhere on the packaging.
  • A release date between about 2001 and 2006, on an EMI, BMG or Sony BMG label.

Macrovision withdrew its CD protection products in 2006 and the major labels had abandoned the idea by about 2007, so the great majority of audio CDs, older and newer, are ordinary discs that copy and verify without trouble.

The short version

  • A CD’s error correction rebuilds damaged data exactly. When it can’t, an audio player conceals the gap with a guess.
  • Some copy protection planted wrong samples that the disc’s own codes expose, so that players guess past them and copies don’t.
  • The true samples at those points aren’t on the disc, so no copy can be exact.
  • Drives guess differently and audio CDs carry no checksum, so no copy can be verified.
  • The difference is rarely audible. What is missing is proof.

What this means for a transfer

When we copy audio CDs to files, every track is verified. A disc protected this way can’t pass that check, for the reasons above. If one of yours turns out to be, we’ll tell you rather than work around it.

Media transfer service How a CD stores data

Common questions

Can a copy-protected audio CD be copied exactly?
Not when the protection works by deliberate errors. At chosen points the disc holds wrong sample values that its error correction can detect but can’t rebuild. The true samples exist only in the label’s master. A copy can hold the wrong values or a player’s guesses, and neither is the music as it was mastered.
What is the difference between error correction and error concealment?
Correction uses the parity stored on the disc to rebuild damaged data exactly. Concealment is what an audio player does when there is too much damage to correct: it replaces each unreliable sample with a made-up value and keeps playing. Correction recovers the truth. Concealment is a guess. More above.
How does a CD player guess a missing sample?
By interpolation: it replaces a flagged sample with the average of the good samples before and after it. For longer gaps it holds the last good value, and for longer still it fades to silence. The exact method varies between players and drives. See how the firmware guesses.
How did Copy Control and Cactus Data Shield work?
Selected samples were overwritten with wrong values at mastering, while the error-correction codes were left as calculated for the right values. A CD player flags and interpolates each one inaudibly. A computer drive that passes the wrong values through gives a file full of clicks, and a disc burnt from it certifies the wrong values as correct. The full account is above.
Why can’t a rip of a copy-protected CD be verified?
The deliberate errors are in the same place on every read, so a drive repeats the same guess and looks consistent without being correct. Different drives conceal differently, so their rips disagree. And an audio CD carries no checksum of its own. See the three failures above.
Does copy protection make a CD sound worse?
In a good CD player the difference is inaudible. The cost is elsewhere: the deliberate errors use up some of the disc’s margin for real scratches, and these discs were known for misbehaving in car stereos, DVD players and games consoles.
How can I tell whether a CD is copy-protected?
Look for a Copy Control logo or a warning about computers on the case, and for the absence of the Compact Disc Digital Audio logo. Most protected discs date from about 2001 to 2006 and came from EMI, BMG and Sony BMG labels.
Are all copy-protected CDs affected the same way?
No. Some schemes altered only the table of contents, and some relied on software that installed itself on Windows. Those interfere with reading the disc but leave the audio as mastered. The ones that plant deliberate errors in the audio are the ones this lesson is about.

Keep learning

Storage & media How a CD stores data Pits, lands and the error correction that copy protection turned against copying. Data & encoding How a computer knows a file isn’t broken Parity, checksums and hashes: what verifying a copy actually involves. Data & encoding WAV, FLAC and MP3 Where the samples go once they are off the disc, and why lossless is the copy to keep.