CD By the Book

ROM Ruminations

Before pocket calculators, I never sat down to idly figure out large numbers; now, I do. So I whipped my calculator out the other day and figured out how many bytes of audio data are encoded on a CD: 772,632,000. That's figured at 2 bytes (16 bits) per sample x 44,100 samples per second x 60 seconds per minute x 73 minutes of music per CD x 2 channels. The figure for a CD-ROM, the computer data storage version, is only about 550,000,000, however.

Why the difference? Because on the computer version, more data space must be given over to error correcting codes. Audio CD players correct what they can, but are still forced to conceal a few uncorrectable errors by intelligent fudging. Such error concealment is possible only because audio is somewhat predictable: If one data sample shows a signal with a value of + 12, and the value reaches + 16 two samples later, then you won't be very far off the mark if you assume the sample in between is +14.

With computer data, there's no such predictability; the byte between those samples could as easily be 0 or -127. Accuracy counts more with computer data or programs too: An error that would create a barely noticeable twinge of distortion on a musical recording could throw computer programs haywire, changing your fat bank balance into a deficit or changing an instruction from "run program" to "ruin programmer." No wonder that CD-ROM's error correction system, according to Denon, "supplements the normal audio-CD correction system and improves it by a factor of 100." It had better do that.

And I would only be surprised, not astonished, if audiophile record companies someday adopted this format for audio too. The disc's capacity would shrink from about 73 to 52 minutes of music (only a bit more than an LP's), and new players would be required, but error correction accuracy would go up, theoretically lowering noise and distortion.

Meanwhile, of what use is CD-ROM? The format is designed for computer use, but I don't expect to see too many programs distributed that way soon, at least for personal computers. The capacity is too high To fill a CD-ROM would require a lot of programs, and no two users could be expected to want the same assortment. Program versions change too fast, too-with floppies, you can send back the original disk and have the revisions recorded on it.

What we will see on CD-ROM are data bases, like the Grolier encyclopedia, whose 20-odd volumes are now available on a CD-ROM. The text fills only 58 megabytes of the disc-plus 60 megabytes for an index that shows where every significant word in the text can be found. At that, less than a quarter of the disc is used, leaving room for about 3,500 illustrations in a future edition. The CD-ROM's advantages here include lower "printing" cost, more compact storage, and such possibilities as illustrations with limited-animation sequences.

To put that in an audio perspective, this magazine's entire output to date, in 38 years of publication, comes to about 70,000 pages. If that were all text, it would amount to 70 million words or about 420 million bytes of data-less than one CD-ROM. In that format, getting your computer to search for some half-remembered article of 10 years back would be simple, though it might take the computer some time to find exactly what you wanted. Exhaustive indexing like Grolier's would double the storage requirements, turning Audio's back-issue file into a two-disc album, but would make searching simpler and faster. Including the illustrations would multiply storage needs considerably, probably bulging the file out to a disc per decade. (For magazines, it might be most practical to remove multiple appearances of repeated ads, to save space. But let's not omit them altogether-they're part of the fun and a good part of the information.) Other reference works might be of use to us as well. How about a single, master Schwann catalog, showing every disc they've ever listed, when it first appeared and when it was discontinued? That would tip you off to the original date of recordings reappearing on CD (if you care), help you chart artists' careers, or tell you the catalog number of that old record that you just decided to try and find.

Big companies could put their service manuals on CD, for distribution both to service shops and to those who want to do their own repairs. Transistor and IC cross-reference manuals could cover every semiconductor type back to the Bell Labs prototype. Complete audio reference libraries on disc could go back to Lord Rayleigh. Reference books for phonograph collectors could become truly exhaustive ... The possibilities are endless.

So far, of course, you can't use your audio CD player for CD-ROMs.

The odds are that you can't even use your home computer to access them-the special drives and interfaces are just beginning to appear. But tomorrow's CD players will, I strongly suspect, have multiple outputs: Standard audio (perhaps including quadraphonic surround sound), subcode data (such as graphics, lyrics or titles) and CD-ROM data.

But having a multi-use machine at home could create new problems. I can just hear it now: "Sorry to interrupt your opera, Mom, but I have to look something up on the encyclopedia disc ...."

The Audiophile, Defined

What is an audiophile? According to Keith Peterson, of Phoenix Gold cables, he's someone who "continues to upgrade his hardware in order to prove that all software is bad."

Signs of the Times

The New York Times' "Metropolitan Diary" column, a while back, chronicled an up-to-date avian note: A bird's nest in Central Park constructed mainly of cassette tape.

I encountered another sign of electronics' impact in northern California: A TV satellite dish rimmed with Christmas lights.

Confessions of a Gain-Rider

Even before digital recording gave us about as much dynamic range as we can use, the practice of "gain-riding"--varying recording levels as you tape--had gotten a bad name for itself. A recent "Tape Guide" item criticized the practice; some editors here consider it anathema. As for me, I used to do it all the time, and still do, on rare occasions. The reason I now do it rarely is not because it's wrong or I've outgrown it, out because today's equipment makes it far less necessary.

Gain-riding is criticized because it alters--compresses--the signal's dynamic range. It's senseless to do this when there isn't any need. But there are times you have to, times when the signal simply won't fit on the tape without a bit of active intervention.

When I first started recording, my deck's dynamic range was too limited even for dubbing the LP records of the time, let alone for taping live performance, without making some gain adjustment. Now, what with Dolby C and dbx NR, I can get enough dynamic range on tape for hands-free dubbing of LPs and even most CDs (a sign that CDs don't use all the dynamic range available, either), and I no longer make many live recordings. So I ride gain the way you'd ride a park bench-find a good place to sit, and stay there.

However, live recording still sometimes exceeds the range of my equipment. (For example, some performers I've taped habitually move up toward the mike when playing loud, and move back when playing quietly.) To compensate, you have to compress the music just a bit to make it fit the medium. At that point, the only choice left to you is whether to be an intelligent compressor or a stupid one.

The intelligent way to ride gain is to anticipate problems and work with the music. For example, start reducing level, very slowly, at the start of a big crescendo, and keep reducing level more slowly than the music is increasing it. This way, the music's dynamics keep on building (though not as much as if you'd left them alone), partially concealing your work.

As the music drops down toward a pianissimo or lower level, which might get lost in the noise, start increasing gain again, s-l-o-w-l-y. If you know the music well enough (following a score helps, if you can read scores and there is one), you can often make your adjustments so subtly that a bystander can barely tell your hands are moving.

What's given gain-riding a bad name, I feel, is doing just the opposite. If you yank the gain down just when the needle pins or the red LEDs light up, and then haul it back up when the meter's shown no life at all for several seconds, you're actually doing more harm to the music than the recorder's own noise and distortion would.

The trick is to react to what's coming up, not to the damage that's already done. You want your signal compressed j-u-s-t enough to avoid distortion at one end of the scale and burial by noise at the other. But you don't want it to sound compressed; you want to leave no fingerprints on the performance.

(adapted from Audio magazine, Jan. 1986)

= = = =