October 2009

The Mechanism Of CD Rot
Article By Mark P. Fishman
The BAS Speaker Volume 31 No. 3

  Having read Brad Meyer's concerns about deterioration of CD-Rs (BASS V31 No2), specifically rising error rates with both radius and age, I think some explanation of mechanism is in order. For much of what I write below, credit should go to Mike Richter (www.MRichter.com), who has been at this far longer than I, has had access to better tools, and has burned many more CDs.

There are a couple of points of interest in Meyer's story. One is related to aging of CD players, which are not just electromechanical devices with special-purpose embedded computers, but are, crucially, electro-optical solid-state devices.

When transistors were introduced, one of their selling points was that they aged far more slowly than vacuum tubes, and failed far less often. In marketing-speak (NB: a pejorative term), they last forever. The lasers used in CD players (and other optical-media devices) are laser diodes,  not the gas or crystal lasers used in science fiction as death rays or cutting tools. They emit light when current passes through them, much like ordinary LEDs [light-emitting diodes], but laser diodes emit a narrower spectrum of light at a more consistent brightness.

That's when they are new.

Laser diodes age, and the characteristic with which we are concerned is the spreading of the emitted spectrum. Not only does the output at the center-frequency fall as the diode ages, that output becomes noisier: amplitude becomes less stable, and more energy goes into sidebands. The light becomes less pure. The physical dimensions of a stored bit on an optical medium are closely tied to the wavelength of the illuminating laser, so when that laser starts emitting more energy at longer and shorter wavelengths, and less energy at the wavelength it's supposed to emit, the contrast in the reflected signal drops. Basically, an old laser diode will produce a lower signal-to-noise ratio in the recovered waveform, with less-clearly defined edges on the ones and zeroes, than a new laser diode.

Thus it's harder for an old CD player to read the raw data off any CD in a form that can be reconstructed without errors. (I'm talking entirely about the physical process of reading the reflected light from the CD, which takes place before the digits are reconstituted. Once you get a readable data stream and correct the correctable errors, the conversion from digits to analog has no more noise in it than it should.) The other interesting point relates to the aging of the CD-Rs themselves. Pressed CDs and CD-Rs use the same kinds of materials for the actual reflective layer, and while claims are made for oxygen-resistance in that layer (gold, for example, rather than aluminum or silver), the reflectivity of the metal might be more important than any other characteristic, unless you live in a corrosive atmosphere that promotes oxidation (classic "CD rot").

A CD-R differs from a pressed CD in the way that it modulates reflected light. A pressed CD uses a varying depth, molded into the plastic, that is matched to the laser wavelength so that the tops and bottoms of this variation reflect different amounts of light. A CD-R can't have a molded variation, so it uses a dye that changes its opacity . its ability to block light . based on light and heat exposure. Ideally such a dye would stay receptive to change until you expose it to a special light, and thereafter stay in its altered state(s) forever.

By necessity, though, such dyes are unstable. Once manufactured they start to deteriorate, and can be modified reliably only when new. After some indeterminate but relatively short period measured in years, unused CD-Rs show a significant rise in error rates when written, because the necessary characteristics . frequency, exposure time, intensity . that would alter the dye state no longer match the laser or the behavior programmed into the CD-writer, and this deterioration is not uniform over a whole CD-R or throughout a whole batch of CD-Rs. At any point in the unwritten life of a CD-R, there's an optimal match between dye and writer. Not all media work equally well in all CD-writers. The dirty secret of the CD-R business is that you can't just spend more for a "better" blank. The best blank for your CD-writer, and the best speed for that blank in your CD-writer, is the one that produces the lowest (correctable) error rate. Any uncorrectable errors, of course, are a show-stopper.

Once written, the dye (although much more stable than previously) remains unstable. Properly written CD-Rs from the earliest days of such things are still readable . but no one makes those dyes anymore, so you can't buy the blanks, and no one makes 1x or 2x CD-writers anymore, so they wouldn't be usable if you could buy them.

All the data we have about expected lifetimes of current dye formulations and current write strategies come from so-called accelerated-aging studies, which are then extrapolated to provide those 10 to 100-year claims.

What we can learn from those studies, though, is that heat, light, humidity, solvents, and adhesives all shorten the storage life of a CD-R. If you want to safeguard your ability to recover the bits from CD-Rs, you have to store them in the dark, in a climate-controlled environment: somewhere around 65 to 68 Fahrenheit, about 40% relative humidity, and don't stick paper labels on them! Especially, do not write on them using solvent-based inks. If you use a thermal or inkjet printer to print on them, do so before you write the data, so that you can verify readability with the labeling already in place.

Most people listen to marketing-speak, so the odds are good that most CD-Rs are being left out on tables, in bookcases that don't protect them from ambient light (sunlight is especially bad), in jewel cases or sleeves that don't block light, and used in environments where temperatures routinely rise over 80F and/or 60% RH (such as in cars, or during the summer in New England). No matter what the archival-longevity claims might be for the media used, if you leave them in a closed car in the summer, or in direct sunlight, even if they don't melt they are likely to become unreadable in a very few weeks or days.

The upshot is that if you care about the data on your CD-Rs, there are several steps you can take to prolong the life of that data:

(a) Choose media that, when written on your drives, in your system, result in burns with very few correctable errors. You have to run the tests yourself, and you have to do it every time you buy discs from a new manufacturing run.

(b) Write at least two copies of every disc, preferably on different brands of media matched to different drives.

(c) If you are writing audio CD-Rs, keep copies of the audio as data files. Data CDs (CD-ROM) have additional error-correction overhead that means you can store about 13% less information as data than you can as streaming audio. That's why an "80-minute" CD-R holds about 700MB if data files, vs. about 800MB of audio. The extra error-correction information makes it more likely you'll be able to recover the original bits.

(d) Store them separately, in the dark, in a cool, dry place.

(e) Re-read them every five years; duplicate them every ten, or sooner if you notice a problem.

Welcome to the curatorial problem that's been facing libraries and museums for half a century already. The only way to make it worse is to let companies apply proprietary encryption schemes and access management systems, and then hope they don't go out of business or block your access to your own recordings.

Enjoy the Music.com highly encourages our readers to join the Boston Audio Society by clicking here).

This article is copyrighted © by the author or the Boston Audio Society. It is posted on Enjoy the Music.com with their permission, and with all rights reserved.