Research notes / Preservation evidence
Optical disc lifetime: error thresholds, percentiles and uncertainty
Published · Literature synthesis and arithmetic check
A disc's estimated lifetime is not its warranty. Ask which error threshold, population percentile, storage environment and uncertainty calculation produced the number.
An error count needs a correction boundary
The 2004 NIST paper distinguishes errors entering a decoder from errors it cannot correct. Neither is a count of damaged files. CD BLER and DVD PI Sum8 also use different observation units. [1, section 2]
| Measurement | Unit and boundary | Avoid this inference |
|---|---|---|
| CD BLER | Blocks with detectable errors per second, entering C1 correction. | Not uncorrectable bytes per second. |
| CD E32 | Errors uncorrectable at the C2 stage. | Not interchangeable with BLER. |
| DVD PI Sum8 | Rows containing errors, accumulated over eight consecutive ECC blocks. | Not a per-second rate or a count of lost files. |
The studies use 220 blocks/s for CD BLER and 280 PI errors over eight ECC blocks for DVD as warning/end-of-life thresholds. Crossing a study threshold and observing unrecoverable user data are different events. Record the metric, aggregation window and test setup before comparing scans. [1] [2, sections 3 and 5]
What the longevity study actually sampled
The 2007 final report covers seven CD-R and fifteen DVD products purchased in 2005, with 90 discs per product across six temperature/humidity conditions. It extrapolates toward 25 degrees Celsius and 50% relative humidity using an Eyring model and log-normal lifetime distributions. This is not decades of observed survival. [2, sections 1, 2 and 5]
A less visible limit matters: analyzers sampled five radial/time bands, not every location on every disc. Recorder/media pairing was selected for compatibility. Light, contaminants, handling and playback-system variation were outside the aging model. The result is conditional evidence, not certification of all files or today's retail products. [2, sections 1 and 4.6.4]
Two different meanings of 95%
95% survival refers to a fitted population: the time at which 5% have reached the defined end point. 95% confidence concerns uncertainty in an estimate. It is not a second population survival percentage, and the two numbers should not be multiplied into a reliability score.
The appendix's rounded inputs are a mean of 12.28 and standard deviation of 0.45 in natural-log hours, with n = 90. Hesela checks the arithmetic below using 8,760 hours per year. The report gives about 24.7 years for the median and 10.7 years after its adjustment; rounded inputs produce the values shown here. [2, Appendix I, steps 6-7]
| Quantity | Years | Calculation in hours |
|---|---|---|
| Fitted median | 24.6 | exp(12.28) |
| Fitted fifth percentile | 11.7 | exp(12.28 - 1.644854 * 0.45) |
| Report-style adjustment | 10.7 | Fifth percentile multiplied by exp(-1.96 * 0.45 / sqrt(90)) |
This reproduces a historical recipe, not a general statistical guarantee. It does not refit the accelerated-aging model, propagate uncertainty in its parameters or establish coverage of the confidence procedure. A numerical match cannot validate those assumptions.
Reproduce it from the source-located inputs and dependency-free calculation module. Published results and limitations include the unit conversion.
Harsh conditions are not a calendar conversion
Slattery and colleagues' earlier study explicitly says its sample was too small to estimate lifetime. Stability under direct light and under heat/humidity also differed for some samples. A good result against one stress cannot silently stand in for all deterioration mechanisms. [1, sections 3-4]
Natural aging offers a different comparison. The Library of Congress reports that most discs in its first fifteen years of CD observations changed relatively little, while 4% of that specific population exceeded the standard error-rate limit. This is neither an annual failure rate nor a probability for all CD-R, DVD or Blu-ray media. [3]
Hesela's inference: an archive should retain observations and the assumptions behind predictions. Replacing a measured error trend with an advertised number discards information needed to decide when evidence has changed.
Blu-ray is not a single recording chemistry
M-DISC versus dye-based Blu-ray is a false binary. Ordinary BD-R includes inorganic HTL media as well as organic-dye LTH media. JVC's terminology describes the direction in which reflectivity changes after writing: high to low for HTL, low to high for LTH. These names describe the recording response, not a lifetime rating. [5]
LTH needs explicit drive support; JVC notes that some compatible models need a firmware update. CCI separately warns that format incompatibility can look like degradation. Keep the exact format and reader in the preservation record before interpreting an unreadable disc. [5] [6, recorder compatibility]
The M-DISC FAQ itself acknowledges inorganic non-M-DISC Blu-ray. It distinguishes the special writing support needed by its DVD from Blu-ray compatibility; that statement should not be stretched to every capacity or drive. Verbatim's cited 100GB product specifies BDXL-compatible drives. [7] [10]
CCI uses LDC and BIS for Blu-ray quality scans. Do not apply the CD BLER or DVD PI Sum8 thresholds above to a Blu-ray report, or treat a corrected-error scan as proof that every file matches a trusted copy. [6, recording techniques]
Follow the M-DISC citation, not just the headline
NIST IR 8387 Table 1 groups M-DISC DVD and Blu-ray under a longevity entry of at least 100 years. Its footnote frames longevity as an offline-readability planning horizon before copying, not a fitted survival percentile. The row cites both a 2009 Navy comparison and Perdereau's 2012 LNE report. [8, printed page 5]
Following the latter reference reveals a narrower experiment. LNE tested three M-DISC DVD samples, not Blu-ray, at 90 degrees Celsius and 85% relative humidity. They exceeded the study's PI8 criterion at the 250-hour inspection and were unreadable at 500 hours. Measurements for this media used CATS at 1X because the other analyzer's drive was incompatible. [9, sections 2.3, 3.2 and 5.3]
This is a severe-stress comparison, not a shelf-life estimate. Inspections 250 hours apart do not locate the exact failure instant. It neither establishes a 100-year life for Blu-ray nor proves that contemporary M-DISC fails in ordinary storage. The small historical DVD sample and single stress condition cannot settle either claim. [9, sections 4-5]
Hesela's source-audit conclusion: this LNE report does not independently substantiate the grouped calendar-life claim in the NIST table. We have not re-audited the separate 2009 Navy report here and do not present this as a complete reassessment of M-DISC testing. A citation is a route to evidence, not a substitute for checking what was tested.
Verbatim describes a several-hundred-year projection based on ISO/IEC 16963 testing for its cited BDXL product, alongside a ten-year limited warranty. These are different claims; neither is centuries of observed field survival. We do not derive a new lifespan or a modern brand ranking from them. [10]
Structured records: BD-R HTL, BD-R LTH, M-DISC and Blu-ray data archives.
A useful preservation record
The playback chain is part of the archive. The Library of Congress notes that optical media can outlast the equipment and software needed to use them. [4] Our suggested record keeps four kinds of evidence separate:
- Identity: format, batch where known, recording date, recorder, firmware and recording settings.
- Measurement: analyzer/reader, speed, scanned area, error metric, units, aggregation window and unreadable regions.
- Content: complete-read results, comparison with a trusted earlier hash, and checks that recovered files open correctly.
- Continuity: independent copies, migration history and the availability of a tested reading path.
This is an editorial checklist, not a tested optimal preservation schedule. No new physical aging experiment, modern product ranking or brand-specific lifetime is claimed. A partial scan cannot establish that unscanned content is intact; a complete successful read cannot establish future survival.
Related: scrubbing and verification scope and verification versus restoration. Structured terms: optical archival storage, CD BLER, DVD PI Sum8 and optical-disc lifetime estimate.
Primary sources
- Slattery, Lu, Zheng, Byers and Tang. Stability Comparison of Recordable Optical Discs: A Study of Error Rates in Harsh Conditions. J. Res. NIST 109(5), 517-524 (2004), sections 2-4.
- NIST/Library of Congress. Optical Disc Longevity Study, final report (September 2007), sections 1-5 and Appendix I.
- Library of Congress: CD-ROM Longevity Research, natural-aging observations and accelerated study.
- Library of Congress: Audio/Visual Materials FAQ, preservation and playback-system longevity.
- JVC (7 October 2008): BD-R LTH TYPE announcement, terminology and compatibility notes.
- Joe Iraci. CCI Notes 19/1 (2020): Longevity of Recordable CDs, DVDs and Blu-rays; BD-Rs and recorder compatibility.
- Millenniata M-DISC FAQ: format-specific writing compatibility and inorganic BD-R, manufacturer explanation.
- NIST IR 8387 (2022), Digital Evidence Preservation: Considerations for Evidence Handlers, Table 1 and Perdereau reference.
- Jacques Perdereau, LNE (June 2012). Durability of recordable DVD and glass DVD at elevated temperature and humidity, sections 2.3, 3.2, 4 and 5.3.
- Verbatim M-DISC BDXL 100GB, product 98912: projected lifetime and warranty, manufacturer claims.
Source review: . No third-party figures reproduced.