Editors Note: Richard Steiner comments below on, "What constitutes a 'good' standard?" His response is specific to recommended practice guides and standards on physical measurement, but his ideas can be generally applied to all standards.

I've seen a recommended practice guide (that did many wrong things). For a good recommended practice document, it must:

1. Be correct. (At least without glaring errors due to misprints or inadequate scientific review.)

2. Be clear and simple in describing the theory of operation.

3. Describe the general specifications of the device design and/or operational apparatus. (Never list specific manufacturers' equipment as "required".)

4. Describe several typical procedures and associated error sources for an applied measurement. (Any standard is no good unless it can be used to calibrate something else. Do not list one method as THE way of doing things.)

5. Cover as many known cases of potential measurement errors or failure modes as possible. (Do not imply that there are few or no possible error sources.)

6. Mention tests to determine or ways to reduce uncertainty contributions. (Quoting manufacturer's specs is not acceptable.)

7. Be rewritten periodically to update all of the above. (Unexpected errors, outright device flaws, or aging failures are always popping up.)

8. Be reviewed by a wide range of actual users and accepted by a significant number or all of them. (It cannot be written by a few "experts" intent on promoting "their" method.)

I deal more with standard devices accompanied by recommended operation documents, so I'll limit my musings to that.

For a primary standard of a physical unit:

1. The basic theory of the effect for a properly working device has to be simple, well understood, and widely accepted. (A complete apparatus may be far more complicated.)

2. Any dependent variables must be easier to measure than the standard's values.

3. The device and operating apparatus needs to be reasonably acquired by several users, i.e., not one of a kind.

4. Results must be independent of the operator, not just consistent with the same operator. This should be obvious, but a lot of people think that because they always get the same answer, it must be right.

5. The degree of the standard must be understood by its users.

Because "good" is relative, my outlook on a "useable" standard is less restrictive than a "good" standard. Item #5 above depends on the number of dependent variables and desired uncertainty. For example, frequency standards would be zeroeth order, since frequency is a defined concept (for cesium), but it can range from atomic clocks to GPS receivers to crystal oscillators. The Josephson effect is first order, since it depends on frequency. The degree directly affects the acceptable uncertainty of the standard. When something depends on too many variables to control or measure, it may be a "useable" standard within the required specs (good enough for government work;-).

Go to Are Engineering Standards Developed Fast Enough?
Go to Creating "Good" Worldwide Standards
Go to Controversies and Comments About "Good" Standards