8.1.9 Errors, Life Expectancy and Testing and Analysis

8.1.9.1  The only way to know the condition of a digital collection is constant and comprehensive testing. This cannot be stated too strongly; no collection using CD-R or DVD-R/+R as an archival carrier should be without a reliable tester. The error correction capability of most replay equipment will mask the effects of degradation until the errors are well into the uncorrectable region.When this point is reached, all subsequent copies are irreversibly flawed. On the other hand, a comprehensive testing regime allows for best possible planning of preservation strategies by acting on the known, objective and measurable parameters that digital archiving make possible. In the well-documented digital archive, metadata will record the history of all objects, including a record of error measurements and any significant corrections.

8.1.9.2  Life expectancy of CD-R or recordable DVD is a many varied topic. To most end users, a CD-R or DVD-R/+R reaches the end of its life when the drive no longer reproduces the data written on the disc, but because drives are not governed by standards, a CD/DVD that will not play on one drive may well play on another. There are innumerable examples of this. The ANSI/NAPM IT9.21-1996 – Life Expectancy of Compact Discs (CD-ROM)- Method for Estimating Based on Effects of Temperature and Relative Humidity, discusses many of these issues. Alternately, some standards and suppliers specify an acceptable Block Error Rate (BLER). BLER is the number of erroneous blocks per second measured at the input of the C1 decoder (see ISO/IEC 60908) during playback at the standard (x 1) data rate averaged over a 10 second measuring period. Standards ISO/IEC 10149 and ANSI/NAPM IT9.21-1996, or Red Book standard, specify a maximum BLER rate of 220. The standard for recording general data on CD, otherwise known as Yellow Book standard, specify a BLER of 50. For data purposes this lower level is vital.

8.1.9.3  Studies have shown that BLER alone is not a very useful measure when determining LE, because defective discs may exhibit BLER well under 220, or indeed under 50. It is necessary to measure other test parameters, among them E22, E32 (uncorrectable errors), and frame burst errors (FBE, sometimes called Burst Error Length or BERL), which are valid end-of-life indicators.When these parameters exceed the limits specified below, it indicates a need for immediate duplication, assuming the disc containing archival information is still readable.

8.1.9.4  Errors in archival CD-Rs should not exceed that specified in the table below. These are maximum levels after which CD-Rs must be copied. In practice error levels much lower than this are achievable and preferable, and must be met in order for the disc to have any archival life before recopying becomes necessary. A BLER average of 1 and a peak level of less than 20 are easily achievable. Jitter is also a useful diagnostic indicator of the quality of the data recorded on a CD and should be measured after writing. The 3T jitter values should not exceed 35 nS (Fontaine and Poitevineau, 2005).

 

Frame burst errors FBE <6
Block error rate BLER average < 10
Block error rate BLER peak < 50
E 22 (correctable errors) 0
E 32 (uncorrectable errors) 0
3T Jitter <35nS

 Table 1 Section 8.1, Maximum error levels in an archival CD-R

8.1.9.5  The construction of a DVD is significantly different to that of the CD, and though there are many aspects in common the criteria that applies to CDs does not necessarily apply to the DVD. Jitter in DVDs is customarily measured in percentages. Though measured differently, the actual jitter measurement is largely equivalent in the two disc types, the main error measurements, however, are quite different. The two main DVD error measurements are Parity Inner Errors (PIE) and Parity Outer Errors (POE). Industry standards state that the POE should be zero. Other types of error measurement are defined, but at the time of writing no agreed threshold for archival purposes has been developed. The DVD specification also states that any eight consecutive ECC blocks (PI Sum8) may have a maximum of 280 PI errors and jitter should not exceed 9%. However, with regard to the use of recordable CD, archival experience and testing has led to a recommendation in maximum error levels that is approximately 25% of the red book recommendations. An extrapolation on the DVD figures would lead to a recommendation of a maximum of 70 PI errors in any eight consecutive ECC blocks. It is important to recognise that a distributed range of tests on DVD recordable in archival situations has not been undertaken to assess the validity of these figures.

8.1.9.6  Initial investigations indicate that recordable CDs do not necessarily proceed to failure in a linear way and that as a consequence small change in initial error rates could have a greater effect on useful life of the disc. There are several tests that have indicated this to be the case (Trock, 2000), (Bradley, 2001), however, there has not been an extended examination of this proposition. A “longitudinal” examination of recordings over time coupled with artificial aging experiments might bring better information on the factors of disc stability. A factor which continues to add to the lack of consistent research is the lack of an agreed standard for the production of CD/DVD-drives.

8.1.9.7  The comparison of the solid black line to the dashed line (see Fig 1 Section 8.1 overleaf) illustrates that the better the initial recording is the longer the expected lifetime will be. There are several tests that have shown this to be the case (Trock JTS 2000, Bradley IASA/SEAAPAVA 2001), however, there is no empirical proof that this is the case. The dashed line, starting at a higher error level, decays at the same rapid rate, but starting earlier reaches failure level in a much shorter period of time. A “longitudinal” examination of recordings with time, aging experiments, might bring better information on the factors of disk stability. A factor adding to the lack of consistent research is the fact that there is no standard for the production of CD/DVD-drives.

8.1.9.8  Being a composite item containing, amongst other components, organic dyes or other chemical compounds, these optical carriers are bound to deteriorate due to slow chemical reactions. Choosing optical discs as the target medium entails the requirement to set up a monitoring program for the discs and a procedure for recopying discs that approach the limit of LE. The use of recordable and rewritable CD/DVDs as archival carriers cannot be advocated unless a strict testing and monitoring program is set up. It should be noted that testing and analysing, though absolutely necessary, will be time consuming, adding long-term costs to the archival solution.When planning an archival strategy, these costs should be included. Logs of test results should be stored, and occasional testing, perhaps annually, can be carried out on statistically appropriate number of stored discs carrying archival information.When the error rate is shown to be increasing, a transfer to a new carrier can be undertaken of all the discs of that age or type.

accumulated CD-R errors over time

Fig 1. Section 8.1: Accumulated errors in a CD-R over time

8.1.9.9 Summary of Testing

8.1.9.9.1 Test all discs when writing.
8.1.9.9.2  Reject any discs which fail to meet specification.
8.1.9.9.3  Store the relevant test records of all discs.
8.1.9.9.4  Undertake a regular testing of a statistically significant number of stored discs of each different batch of products. 8.1.9.9.5  Undertake a recopying of discs when error rates increase.