The Analog Compact Disc:
Page 3
All three discs sounded different, the Zeonex sounding the best. The standard disc had a drier, more forward sound, with less depth. The treble was more forward, yet the sound lacked the air, openness, and extension heard on the Zeonex disc. The Zeonex disc's bass was also better defined, and I heard an overall increase in resolution, with more musical information presented.

These sonic differences may be regarded as slight or even imperceptible to the casual listener through a low-quality stereo system. But to the audiophile with a keen ear, an open mind, and a high-resolution playback system, the differences are musically significant.

I can imagine a skeptic playing these discs casually for a few seconds through a low-quality, poorly set-up system and concluding that they all sound the same. It's easy to scoff at the possibility of sonic differences between CDs; after all, CD-ROMs work perfectly, no matter where they were manufactured. This argument—that computer data and audio data are identical—forgets that computer data are never converted into an analog signal and perceived with analog instruments (our ears). Computer data and audio data are identical and can be treated identically—if the data are never converted into music.

Another example of how two discs with identical data sound different is the strange case of copying (in the digital domain) a CD to a CD-R (a CD made on a CD recorder); the CD-R sounds better than the disc from which it was made. Although the data are identical, the CD-R's HF signal looks much better than that of the mass-manufactured CD (footnote 5).

Finally, CD tweaks such as disc stabilizers, CD Stoplight, and fluids applied to the CD surface don't affect the data integrity, meaning that other factors are at work in changing the sound of CD playback.

But if the data are the same on all these discs, why do they sound different?

What goes wrong in CD manufacturing
Creating the pits on a CD is a highly variable process. Nearly every step in the CD manufacturing process affects the pit shape, which affects the HF signal recovered from the disc.

Let's take a closer look at the CD manufacturing processes described earlier, and what factors could introduce this analog-like variability in CD sound quality.

First, the glass master's photoresist thickness must be correct to within a small fraction of a micron; photoresist thickness determines the pit depth, which determines the HF signal's amplitude. The photoresist must also be a uniform thickness around the disc. Otherwise, the pit depth will vary and cause amplitude modulation of the HF signal. Any contamination in the photoresist can wipe out large areas of pits, causing errors in the recovered datastream.

The pit shape—and thus the HF signal's quality—is determined by myriad variables in the mastering process, including the recording laser beam's intensity, the ambient humidity in the mastering room, developer solution concentration, development time, and how vigorously the master is agitated in the developing solution—to name a few. A slightly too-high laser power or a few extra seconds under the developer will give the pits on the glass master a poor shape and result in a low-quality HF signal from the final CD. Moreover, if the master isn't developed uniformly, some pits will be longer, deeper, and/or wider than others.

Another factor that changes the pit lengths from their theoretical ideal is even the tiniest rotational instability of the turntable spinning the glass master. A slight decrease in turntable speed produces longer pits; an increase in turntable speed produces slightly shorter pits. A flutter-type instability produces pits of rapidly varying length.

With this concept of turntable flutter in mind, let's go back to what I said earlier about how the digital data are encoded in the HF signal recovered from the disc. We saw that the 1s and 0s are contained in the zero crossing points of the nine discrete-frequency sinewaves of the HF signal. The zero crossing points are the logic thresholds at which the system differentiates between 1s and 0s, and the point from which the clock is recovered. Varying pit lengths will shift those zero crossing points in time, introducing jitter in the HF signal.

---

Footnote 5: At the 1992 Winter CES, Meridian's Bob Stuart copied a CD to a CD-R of music that engineer and high-end retailer Peter McGrath had recorded. Bob played the original CD, then the CD-R. Seconds into the CD-R, Peter jumped from his chair and exclaimed, "That's impossible!"