A generalized accuracy chart of the type used at NBS to make basic decisions regarding priorities within the measurement system.

areas. The figure above is an idealized version of such a chart. The heavy line indicates present NBS capability; the next line down shows what good industrial laboratories can do; the lower line gives accuracies at the ultimate user's level-at the factory bench and in the finished product. The dots indicate the accuracies NBS customers say they need and the dashed lines show where NBS activities now underway are leading. Finally, the stars represent the ultimate needs for capability expressed by important segments of the scientific and industrial community.

NBS staff use this type of chart to show graphically where major efforts are being applied, to indicate goals, to display needs of the national measurement system, and to determine where to concentrate further efforts. The charts aid in resolving such questions as whether it is more important to raise the line representing NBS capability, and thereby bring up the line representing industrial capability, or whether to try to bring the industry line up closer to the NBS line by tightening up the system, perhaps by reducing the number of echelons between the standard and the ultimate user.

Unfortunately, users of the system who send instruments to NBS for calibration are not likely to obtain any knowledge of their own capability to use the instruments to the accuracy with which they have been calibrated, nor can users be certain that accuracy has not been degraded in the shipping process. With this in mind, NBS scientists are now considering ways for tying all laboratories in the country into the measurement system on a self-calibration basis. The aim is to make it

277-070 0—67 -2

Graphic representation of interactions between various groups in the instru mentation network.

possible for industrial laboratories to do much of the work of calibrating with their own instruments, staffs, and procedures. Having done so, they would then have a measure of their own capabilities and would know how well their instruments could be traced back to the national standards.

Ways in which self-calibration can be carried out include the use of standard reference materials, key data on properties of materials, and circulating standards.

Standard reference materials are highly characterized materials that NBS certifies either for chemical composition, physical state, or with respect to a specific physical or chemical property. The Bureau disseminates materials on request at a fee which covers the cost of preparation and certification. Purchasers use these reference materials to calibrate their measuring process.

A new approach is now being used with great success in the field of mass calibration. The primary aim is to provide techniques for ascertaining that the measurement processes used in different laboratories are compatible. Under this program laboratories do not send their mass standards to NBS for calibration; instead they receive two calibrated mass standards from the Bureau and use these weights to calibrate their own sets of weights, and also their measurement procedures. Raw data are transmitted to NBS for statistical analysis by high-speed computers. Thus, emphasis is placed on performance of the measurement process and its ability to produce consistent measurements, rather than on the assignment of values to individual test items.

Data Network

The data network offers another important means for enabling the user to perform his own calibrations, for when sufficient data have been obtained to characterize a substance, it then can serve as a reference material for the calibration of instruments that measure the properties of other substances. The use of freezing and boiling points of various substances the "fixed points" of the temperature scale-to calibrate a thermometer is a good example. Thus, in this network the NBS role is to select certain key materials, to characterize these materials carefully, and to make precise measurements of their properties. In some cases the Bureau makes available not only the data but the materials themselves as standard reference materials. The system can then couple to the properties of these materials and use them as points of reference in building a reservoir of data to meet the remaining needs. This central function provides a basis for NBS leadership of the system. At the same time it supplies the system with the basic information needed for self-calibration of instruments and measurement procedures, and gives scientists and engineers the data they depend on in designing and building apparatus and equipment. To do the job properly, NBS must obviously have broad competence in materials research and in measurement science.

The corresponding data network for design specifications or performance characteristics of devices and systems is at present quite broad and diffuse. At this stage of the network's development NBS devotes its effort to providing the technical basis for setting meaningful design or performance standards, leaving the actual setting of the standards to the voluntary cooperation of the other elements of the system (except for mandatory standards that NBS is legally required to set). The data network is of great value in providing a basis for decisions that must rely on measurement. If, for example, an engineer were setting out today to design a new competitive light bulb, there are a great many things he would need to know. Obviously he would need instruments to make direct measurements of the diameter of the bulb, the pitch of the thread, the weight of the materials, the diameter of the wires, and so on. But once he had the capability of making these measurements in production, he would still be a long way from the design of a light bulb. He would need a vast store of such ready reference information as the electrical resistivity and spectral emissivity of tungsten and other competitive materials, the melting point and thermal expansion of glass-in fact, a whole library of data of this kind (which incidentally would also be of value to designers of vacuum tubes and other products). If he has to stop and measure all these properties, he will be investing substantial sums in a research program before he can start his design. On the other hand, if data are already available. because someone else has already measured the properties, then he

Graphic representation of interactions between various groups in the data

network.

can save this large investment. Once he has found the numbers, he can proceed with the design, provided that he can trust the numbers to be correct.

There is a second important aspect which points up the need for critically evaluated data in the decision process. When an engineer goes to the literature in search of design data, he is likely to get a wide range of value for each property he looks up. If he is designing an industrial process that involves the heat of formation of hydrogen sulfide, for example, he will find in the literature an array of values ranging from 2.0 to 4.9 kilocalories per mole. If he accepts the value "2.0" for the heat of formation of hydrogen sulfide, he might conclude that his planned process will not work and there is no point in going further. On the other hand, if he accepts the value "4.9" he may find that his process will be highly productive and should be pushed. In the absence of critically evaluated data on the heat of formation of hydrogen sulfide, he can do only what is usually done in industry today-seek expert advice if he can find it, make an educated guess, or measure it again himself, adding another value to the list. Unless he is an expert in the measurement of heats of formation, the value he obtains will probably be no better than those already in the literature, and may be much worse.

The solution to this type of problem is to assemble a group of experts who know the field and who can evaluate the various measurements from the literature and obtain a "best value" the most acceptable and trustworthy value-and can make this value generally available. This is the process of critical data evaluation and compilation. In the data network the primary need is for a core of carefully measured key data that can serve as reference data for the determination of other data throughout the system.

The process of critical evaluation and compilation of data has lagged far behind the generation of data in the literature. As a result, a large backlog of unevaluated data has been built up, and as this backlog continues to grow it has become increasingly difficult for scientists and engineers to find the data they need. Lack of critically evaluated data in conveniently available form has thus become an important and wasteful deficiency of the national measurement system. To remedy this deficiency, the Office of Science and Technology in 1963 established a National Standard Reference Data System and charged NBS with the responsibility for its administration and coordination.

Techniques Network

The techniques network is that part of the national measurement system through which all users of the system can be told how to make optimum use of the measurement capability developed in the instrument network and the data network. Thus the techniques network provides users with the procedures and techniques that make for meaningful measurement.

The term "meaningful measurement" has two distinct aspects. The first is concerned with the ability to measure what one sets out to measure. For example, in attempting to measure the temperature of a particular body at a certain stage of a physical or chemical process, it is sometimes very difficult to be sure that one is in fact measuring the desired temperature, rather than some other temperature within the system. Even though the measurement may be very precise and reproducible, the temperature thus obtained may be quite different from the actual temperature of the body, and thus the measurement can be highly inaccurate.

The other aspect of meaningful measurement is the matter of determining what one should set out to measure in order to accomplish a particular measurement goal. The problem is to find what properties, or combination of properties, or set of physical quantities can be measured, so precise and accurate indications can, by some agreedupon process, be put together to give a number characteristic of the aspect of the system being measured. It must then also be shown that the chosen procedure does in fact lead to an objective, reproducible measurement.

To make meaningful measurements the materials involved must be characterized in terms of properties that are relevant to the measurement goal being sought and in terms of the environment in which the measurements are to be made. Characterization becomes more difficult and more sophisticated as the interactions of materials multiply and become more intricate, and as the environments become more extreme. Characterizing the bulk properties of a material for various measurement objectives is complex. And obviously the characterization of