The Analog Compact Disc:
Page 6
A parameter called a "dropout" is defined as an instance in which the signal coming off the disc drops below 30% of its nominal value. Large errors combined with a dropout indicate physical damage to the disc. Large errors without a dropout could be caused by localized areas of poor pit geometry.

Although these errors are completely corrected, a disc with high error rates will have less tolerance for scratches, dirt, fingerprints, and poor-quality CD players before producing an uncorrectable error. This is especially important in CD-ROM, where bit-for-bit accuracy is essential.

Errors
I encountered no uncorrectable errors (E32s) during 15 hours of analysis. Even Sam Tellig's discs, with the brown blotches around the edge, produced no E32 errors. Not surprisingly, however, Sam's discs had the highest error rates of the discs tested: BLER of 267, E11 of 188, E21 of 47, and E22 of 29. Note that the BLER of 267 doesn't meet the Red Book specification (maximum BLER of 220). The HF signal from Sam's disc was very poor: severely unstable and amplitude-modulated.

By contrast, the best disc I measured was Lesley Olsher's Anyone in Love on Vital Records. I chose this disc for measurement because it had been recently manufactured by Disc Manufacturing, Inc., the company I used to work for, and who I know make high-quality discs. In addition, this disc was fresh out of the wrapper. The BLER was just 5, E11 was 4, E22 was 0, and E31 was 0. The HF-signal quality was excellent. I also measured very good performance from discs made by Sony's Digital Audio Disc Corporation (eg, Music for Trumpet and Orchestra, SSK 6245), and rather poor technical performance from discs made by Matsushita (eg, Telarc Sampler, Volume One). This small sample size doesn't necessarily represent all discs produced by the two plants—CD quality varies from week to week, and even from master to master, within a factory.

Discs that have been in my collection for nearly ten years, and discs I've played hundreds of times, didn't have any higher error rates than new discs. The only exceptions were a few discs that had once been caught between the drawer and the front panel of the Esoteric P-2 transport. The damage, visible to the naked eye, produced some E22 errors but, amazingly, no E32 errors.

Curiously, the gold-metallized Zeonex Chesky disc had more errors and a poorer-quality HF signal than did the aluminum polycarbonate version. I attribute this not to the gold and the Zeonex, but to the master-to-master variability described earlier. This shows that a visual inspection of the HF signal doesn't say anything about the disc's sound. Instead, a high-resolution HF-signal jitter analyzer is needed to really know what's going on in the HF signal.

It's easy to tell where a CD has been manufactured: Look on the inside band between where the music starts and the center hole. Some discs will say right on them where they were made. Others give more subtle clues. A band with a bar code and the letters "DIDX" mean it was made by Sony; the letters PDO indicate that it was made by Philips-DuPont Optical; Nimbus has a distinctive double row of letters; discs with a band of large block letters and no other identification were probably made by Sanyo.

An engineer I worked with, Alan Hamersley, could not only identify the plant where a disc was made, but the mastering machine it was cut on. Nearly every mastering machine has periodic variations in the track pitch (distance between tracks) that occur at the same disc radius. The track pitch variations, caused by periodic imperfections in the mechanical system that moves the turntable underneath the optics during mastering, can be seen with the naked eye as changes in the diffraction pattern.

What it all means
It's becoming incontrovertible that CDs containing the same 1s and 0s produce varying levels of sound quality. These sonic differences are not caused by data errors, as commonly assumed, but by some other mechanism. The most likely culprit is jitter in the recovered HF signal, which varies greatly between CDs. As we've seen, CD manufacturing isn't a black-and-white process of putting binary 1s and 0s on a piece of plastic, but a series of highly malleable procedures, each of which introduces an analog-like variability in the signal read from the disc by your CD transport.

Unfortunately, there's no way for you to judge the quality of the discs you buy. Until the record companies put pressure on CD manufacturers to make better-sounding discs, purchasing CDs will remain a crapshoot.