I have always been curious about the warnings on CD-R media about not exposing them to sunlight. I wondered how long the media would last when exposed to light, and also, what would be the best brand and type of media for archival storage under normal conditions. Exposing the media to light is considered an “accelerated life test”, that is, the performance of media under adverse conditions could be considered an indicator of performance under normal conditions. Similar tests are used in many industries, for example, semiconductors are tested at temperatures far above what they would normally operate at. Their lifetimes at high temperature can be used to predict their lifetimes and failure rates at normal temperatures.
Starting in Februrary of 2001, I hung 8 CD-R’s of various brands and types, recorded side out, on a somewhat-sheltered northwest-facing outside wall of my house in suburban Portland, Oregon, and tested them weekly until they failed.
All of the CD-R’s were audio recordings, because most of the CD-R’s I burn are audio recordings (all original; I have more about my recording techniques elsewhere on this web site), and because I felt best able to evaluate failures by listening for dropouts on audio playback. I don’t know how to get error statistics when reading data CD-R’s and I don’t know how much media deterioration would be corrected for by normal error correction. Of course, there’s error correction going on in audio playback as well. That’s one of the weaknesses of this study.
By July, all but one of the disks had failed by my standards. The survivor was a so-called gold-on-gold disk, and it had still not failed when I stopped the test altogether in August. Note that I started the gold-on-gold disk two weeks after the rest, but it survived long after the others had failed, so I don’t believe its late start skewed its performance.
Six of the disks were computer-data-type media, and were recorded at 8X on a Hewlett-Packard 9110i CD writer. Two of the disks were audio-type media (#4 and #6 in the table below) and were recorded at 1X in a Philips CDR-760 audio CD recorder. Seven of the disks were the so-called “long strategy” dye, and one, the survivor, was the “short-strategy” type.
I used a Sony D-E551 portable player for testing. I determined a CD-R had failed when the audible dropout rate made one or more tracks extremely irritating to listen to, or in some cases, when the CD-R would fail to play at all (the player was unable to read the table of contents). Most disks exhibited some dropouts weeks before I designated them failures. The D-E551 has Sony’s ESP (Electronic Skip Protection) which I always had turrned OFF. In fact, when dropouts started to show up on a disk, the disk couldn’t be played at all with ESP ON. Apparently the dropouts confounded the ESP.
The D-E551 is not specified for, nor will it play, CD-RW media. Even after I designated a CD-R a failure from testing with the D-E551, it could often still be played in a CD-ROM drive or audio player rated for CD-RW playback.
DISCLAIMER: The results of the study should perhaps be taken with a grain of salt because it was not a scientific study. I used only one sample of each media brand/type, they were recorded by two different CD writers at different times, and my evaluation of failure was subjective.
The results of the test are sumarized in the table below. The media are sorted by failure date, earliest first. Under Manufacturer is the brand and outward appearance of each disk. Under CRIdentifier and Recording Layer is the information read by Frank Wolf’s CDRIdentifier program ( the most recent versin of the program can be found at http://www.gum.de/it/download/history.htm). Disks #’s 2, 3, 6 and 8 were purchased at Fry’s Electronics. The others were purchased at CostCo or office supply stores.
| Number | Manufacturer | CDRIdentifier | Recording Layer | Start Date | Fail Date |
| 1 | Maxell (Gold) | Ritek | Dye (Long) | 2/10/01 | 3/23/01 |
| 2 | Hotan (Plain silver, no logo) | GigaStorage Corp | Dye (Long) | 2/10/01 | 3/28/01 |
| 3 | GQ (Hotan silver w/logo) | Lead Data, Inc. | Dye (Long) | 2/10/01 | 4/16/01 |
| 4 | TDK (White Audio) | TDK | Dye (Long) | 2/10/01 | 4/29/01 |
| 5 | TDK (Silver) | Ritek | Dye (Long) | 2/10/01 | 5/19/01 |
| 6 | Hotan (Tan audio) | CMC Magnetics | Dye (Long) | 2/10/01 | 5/28/01 |
| 7 | TDK (White non-Audio) | TDK | Dye (Long) | 2/10/01 | 7/7/01 |
| 8 | GQ (Hotan Gold on Gold) | Lead Data, Inc. | Dye (Short) | 2/24/01 | 8/7/01* |
* This disk had not failed at the time the test was terminated
For a given brand, the actual manufacturer and outward appearance may change from time to time. For example, the white TDK’s, such as #7, are no longer available. All the TDK disks I’ve purchased recently have been silver like #5.
As media started to fail, the first dropouts usually appeared in the outer (higher-numbered) tracks. Since the areal bit density of CD’s is constant by virtue of the use of constant tangential velocity, I’m surprised the dropouts weren’t more randomly distributed. An readers have ideas on this?
When most media were deemed to have failed they had visible fading in the recorded area, which was already lighter than the unrecorded area. On several media, the fading was in concentric rings (blurry, but quite visible).
Generally, the darker the unrecorded media, the longer the time to failure, except for the gold-on-gold, which have almost no visible contrast between the recorded and unrecorded sections.
Conclusions: As I said above, this can’t be considered a scientific study, but based on the results, I’m going to use gold-on-gold disks for archival storage (archiving the .WAV master files of my recordings, for example) but use less expensive disks for general distribution.