Development of a Testing Methodology to Predict Optical Disk Life Expectancy Values

Special Publication 500-200, Development of a Testing Methodology to Predict Optical Disk Life Expectancy Values by Fernando L. Podio of the National Institute of Standards and Technology, is a technical research report that describes a methodology to predict the longevity of optical media. The National Archives and Records Administration supported this three-year project because there are no national/international standards for the longevity of optical disks that can assist managers in the federal government to select optical media for the storage of permanent records with a reasonable assurance of how long they may be stored on the media.

The study, which consists of five chapters, begins with a Program Overview that reviews the history of the NIST computer storage optical media research program and summarizes methodological issues involved in developing the report. Initially, the research plan called for testing several 300 mm WORM (Read Once Read Many) disks using vendor-supplied drives and other supporting equipment. When this became impossible due to costs, only Sony 300 mm WORM disks, which the National Archives was using in another research project, were tested. Other cost considerations also dictated the level of detailed information captured about the test disks.

A key methodological consideration was the Arrhenius Model, a set of mathematical procedures and computations for accelerated aging, which assumes that temperature and relatively humidity are the crucial independent variables that over time affect the longevity of optical media. The NIST use of this model involved storing optical disks in three different high-stress environments (70° C, 80° C, and 90° C with a constant relative humidity of 90 percent) for an extended period of time (ranging from 4120 hours to 5711 hours). The disks were read periodically to monitor the effect of temperature and humidity on the error rate. A linear increase in error rates as harsh conditions of the stress environments increased from 70° C to 90° C over the test period made it possible to extrapolate from these errors and to predict error rates at nominal room temperatures.

Chapter Two focuses upon such technical matters as definition of the end of life of optical media, the measurement of the byte error rate (BER), the optimum number of bytes required for reliable statistical analysis, the pattern of test data (sequential, random, and high frequency), statistical analysis of data, measurement of signal degradation, and procedures for conducting the accelerated aging tests.

The end-of-life definition used in the study was 5 10-4. This means an error rate of five bytes out of every 10,000 bytes, which exceeds the capacity of error correction codes to correct.

Of course, this does not necessarily mean catastrophic failure so that a disk is totally unreadable. Another important consideration was the selection of three different sections of the disk (inner, middle, and outer) of the disk surface area on which to record the test data.

Chapter Three presents the results of the tests, which in most instances are displayed in five tables and fifty graphs. The most significant findings that these tables and graphs convey are summarized below.

• The overall byte error rate (BER) increases as the temperature increases, which confirms the applicability of the Arrhenius Model.
• The byte error rate for the random and sequential test patterns is similar while that of the high-frequency test pattern is larger. The high-frequency test pattern represents the highest areal storage density.
• There is a significant difference in the byte error rate for the three different test sections, with the middle section showing the lowest byte error rate. Consequently, any test that predicts the life expectancy of optical media should calculate the byte error rate for each section and include this information in a report.
• Visual inspection of the three different test patterns revealed damage to the surface of the inner and outer edges, possibly caused by damage to the sealing on the inner and outer edges that may have occurred as a result during placement of the disks in ovens or from the very high stress conditions in the ovens.
• The extrapolated life expectancy with storage at nominal room temperature and 90 percent humidity for each of the three areas varies. The most conservative estimate is 57 years, while a more liberal estimate is 121 years. In either case, a relative humidity between 40 percent and 50 percent should lead to an even longer life expectancy.
• These test procedures are repeatable so that 300 mm optical disk manufacturers can employ them to predict the longevity of their media.

Chapter Four presents a number of conclusions, recommendations, and follow-on activities that flow out of the above discussion of the results of the tests. They include the following:

• Broad test parameters without precise specification can yield equally accurate but uncomparable results (e.g., the average byte error rate for the entire test surface area rather than the byte error rate for each of the surface areas).
• Vendor claims for life expectancy should be accompanied by detailed specification of the test parameters.
• A standardized test methodology for predicting the life expectancy should take into account at least five factors. The entries in boldface are the options used by NIST in their research program.

  
       1.   Test Method Used (substrate independent
            versus substrate dependent)
            a.   glass substrate
            b.   polycarbonate substrate
  
       2.   Quality Measurement Approach (byte error
            rate, bit error rate)
            a.   Data Patterns Used (random, sequential, high
                 density)
            b.   Amount of Data Tested (900 sector blocks on
                 each surface)
            c.   Area of Media Tested (outer,
                 middle, inner)
  
       3.   Mathematical Model Used for Extrapolation
            a.   Arrhenius Model
            b.   Erying Model
  
       4.   Criteria for Data Analysis (300 sector blocks
            for read/write)
  
       5.   Experimental Stress Conditions on Media
            a.   Relative Humidity 90%
            b.   Test Temperatures
                 (1)  60°C
                 (2)  70°C
                 (3)  80°C
            c.   Temperature Ramp-Up/Ramp-Down Rates
  
  

• NIST will use the findings of this study to develop a common test framework for predicting the life expectancy of optical media that has the support of the optical media industry. This common test framework could lead to the development and approval of a national/international standard test methodology to predict the life expectancy of optical media.