TRACEABILITY -
A CONCEPT IN EVOLUTION
By Henry E. Sostmann
As our views of measurement become more sophisticated; as national and international calibration bodies struggle toward organization; as the regulation of quality assurance assumes a more prominent role in domestic and international trade; there is increasing interest in the objective validity of measurements at all levels. I discuss some of the trends I see in the formation of a new view of the meaning (and indeed, in the utility) of measurement traceability.
There is no doubt that the meaning of traceability is in evolution. The review, in this issue, of Nicholas' and White's book Traceable Temperatures, in which a new meaning of traceability is fundamental, is only one example of the call for re-examination.
The traditional meaning of measurement traceability is hierarchical; that is, it means that there exists an unbroken and demonstrable path whose upper end point is some national standard, and whose other end point is the specific calibration (or measurement) which is the subject of the exercise. This is a paper trail, that we might call an audit trail (Nicholas and White call it a filing cabinet trail). "Traceability" is required by some authority, which may be Company policy, the text of a contract between a supplier and a purchaser, an industry standard, or an official regulation. In the latter sense, specifically, such a regulation may be national (e.g., FDA regulation CF 21 820, governing good manufacturing practices for a medical device), or international regulations which require demonstration of compliance if one means to export or import.
In paper-trail sense, traceability is demonstrated by citing a certificate from one's National Laboratory (e.g., a NIST "test number") providing information and a date of the results of a test performed on one's primary standard(s) for that measurement; followed by some proof of the process through which this measurement is transferred to in- house standards, check and working standards, etc., and finally to the product. This is not meaningless, since it shows that the work has been done. It satisfies the definition of traceability of the First Edition (1984) of the International Vocabulary of Metrology [1]:
6.12: Traceability: the property of the result of a measurement whereby it can be related to appropriate standards, generally international or national standards, through an unbroken chain of comparisons.
In many situations the "appropriate standard" may be an intrinsic standard, such as the equilibrium states of pure materials which define the ITS-90, although some of us have had the difficult experience of defending this to an auditor who is ignorant of both the Scale and the concept of an intrinsic standard. An intrinsic standard is often permitted in contracts, by the inclusion of a phrase such as "...or to accepted values of natural physical constants". Such pathways to traceability are necessary because no national laboratory, including NIST, realizes or maintains an experimental basis in all the units which are of interest in present manufacturing. Appeal to a Standard Reference Material (SRM) obtained from such an authority as NIST may comprise an intrinsic standard. FDA in 21 CFR 820 even permits in-house standards, where no higher standard exists.
Absent, or at best implicit, in the earlier view of traceability is any suggestion that traceability should include a demonstration of the accuracy of the measurement through steps of the chain, and as it is applied to product. This absence began to be recognized in MIL-STD 45662A in 1980, with the requirement that standards for calibration have such accuracy and stability as their use requires, and the 1988 edition of this document demands that ...unless otherwise specified... the collective uncertainty of the measurement standard shall not exceed 25% of the acceptable tolerance for each characteristic being calibrated...
Since the 1984 edition of the Vocabulary, there has been an extraordinarily quantity of activity on the subject of quality assurance, including the quality assurance of measurements. This has been stimulated by advances in science and manufacturing technology, and in the political and economic climate of supranational and global markets. It is evidenced in such documents as the ISO-9000 series and ISO Guide 25 (incorporated verbatim in EN and other international-market consortia documents), and in the feverish activity they have generated in the commercial sphere. With this has come the realization that traceability is not a meaningful end in itself, but is a component of a quality assurance system. The revised edition of the Vocabulary (1993) includes a revised definition of traceability [2]:
6.10: traceability: property of a measurement or the value of a standard whereby it can be related to stated references, usually national or international standards, through an unbroken chain of comparisons all having stated uncertainties.
Either definition of traceability implies that a reliable calibration interval for the standard can be established, although none of the definitions above mention time. Readers of these pages will be aware that I have long argued that fixed calibration intervals do not make sense for standard resistance thermometers. Of a fixed-point calibration of a standard thermometer, nothing more may be said truthfully than, in the words of John Evans: "On this date and in my laboratory this thermometer did so-and-so". Standard thermometers are delicate things. It has been said, correctly, that to put one down on a laboratory bench with enough force so that one can hear it touch nay strain it out of calibration. Thermometers shipped for calibration may be strained during return shipment. Other hazards, of handling and of use, can be enumerated. Thus the interval during which the thermometer is in calibration may range from minutes (or seconds) to years. However, if one is prepared to check a thermometer at the triple point of water after every transportation, and after every use, and convey these measurements to a permanent record or control chart, assurance of calibration can be maintained as it can be for no other reference standard. Out-of-house recalibration should be done only when the water triple point check indicates that calibration is required. It is folly to take the risks involved (and spend the not-negligible money it costs) simply to satisfy an arbitrary and meaningless interval. Thermometrists are fortunate in having so reliable a way to control quality of measurement, and to assure that their instruments are traceable to the very definition of the Temperature Scale.
It is especially important to keep careful control over quality of measurement with the new high temperature platinum resistance thermometers (which are the stipulated interpolation instruments of ITS-90 to the freezing point of silver, 961 C). These thermometers are capable of highly accurate measurements, but are susceptible to a number of hazards in use, most of them reversible, some not, if the need for such treatment as annealing or deoxidation is not recognized (by a calibration) in time. Only frequent in-house calibration checks can assure the quality of measurements made with these thermometers.
The traceability pathway can be a long one. The longer the path, the more uncertainties are compounded. It should be a goal for everyone who will walk this way to shorten the path to a minimum number of steps. The ability to realize accepted values of intrinsic standards is the ideal and direct way toward this goal. A working temperature laboratory that is prepared to realize the triple point of water has an unarguable element of quality control over its reference and working thermometers; add the melting point of gallium, and additional control is obtained. The best situation is to have the equipment and skill to calibrate thermometers to the defining points of the ITS-90 over one's appropriate range of temperatures. With defining intrinsic standards on hand, and a measurement assurance system in place for using them properly and evaluating the results of using them, the path can be short, direct and immediate; and at the current price of national laboratory calibrations, can generally return the investment in measurable time.
REFERENCES
[1] International vocabulary of basic and general terms in metrology, (VIM), BIPM / IEC / ISO / OIML Geneva (1984)
[2] International vocabulary of basic and general terms in metrology, (VIM), Second Edition, ISO (with many other organizations participating) Geneva (1993)