this instruction a questionnaire was prepared and sent to all weights and measures officials on the Bureau mailing list, which comprised 356 names; in addition, there were 36 officials at the 1939 Conference whose names did not appear on that mailing list, and questionnaires were sent to these persons also. Two officials reported not receiving their questionnaires and additional copies were forwarded to them, thus making a total of 394 questionnaires sent out. The form of the question was as follows: IMPORTANT-PLEASE CONSULT ATTACHED MATERIAL MAIL BALLOT ON VEHICLE-SCALE TOLERANCES Are you in favor of continuing in effect the present tolerances on vehicle scales as adopted by the National Conference on Weights and Measures? The "attached material" referred to was an explanatory letter, together with a statement of vehicle-scale tolerances and related paragraphs from Definitions, Specifications, and Regulations. The result of the balloting is as follows: 186 ballots were returned, 45 of which were from State officials, 134 from local officials, and 7 from officials unidentified. Thirty-six State officials voted "Yes," 9 State officials voted "No," 111 local officials voted "Yes," and 22 local officials voted "No." This makes a total of 153 voting "Yes" and 33 voting "No," which would indicate that it is the desire of the great majority of this Conference to continue in effect the present tolerances on vehicle scales as adopted by the National Conference on Weights and Measures. Mr. LEVITT. I think there was a little misunderstanding on the questionnaire. Our original proposal and the original motion which was put from the floor was that the questionnaire be sent to interested persons, including others than members of the Conference. After this questionnaire came out, the National Scale Men's Association sent out a questionnaire to all members of the National Scale Men's Association, and the reports that we got from that were about seventyodd favoring the return to the 0.2 percent tolerance, and only about twenty-odd favoring the present 0.4 percent tolerance; so there is a little difference of opinion. The ACTING CHAIRMAN. Of course, the National Scale Men's Association is not the governing body here, and as long as the Committee is the official Committee of the National Conference, the vote that they have taken must prevail, as I see it. So you are voting on the acceptance or nonacceptance of the Committee's report. Mr. MCBRIDE. This is just a report of our action, so there is no need for a vote. Mr. LEVITT. While I am not objecting, I wanted to bring this up to show that there are other groups who do not agree. DEMONSTRATION OF RECENT DEVELOPMENTS IN WEIGHING AND MEASURING APPARATUS, BY REPRESENTATIVES OF MANUFACTURERS SECRETARY'S NOTE.-At this point the representatives of several manufacturers of commercial apparatus briefly addressed the Conference, in some instances referring to exhibits on display adjacent to the meeting room. A. W. Corwin, Sealer of Weights and Measures of Allegany County, N. Y., exhibited and described a carrying case and moistener for seals, which he had developed. CORRECTIONS TO STANDARDS By H. W. BEARCE, National Bureau of Standards Before any weights and measures inspection program can be effectively carried out, the responsible officials must be provided with adequate standards. It is a well-known fact that no weights and measures apparatus or other manufactured product is absolutely correct. This applies to the measures used in trade and commerce, to field standards used to test commercial measures, to office standards used to test the field standards, and to the standards tested and used by the National Bureau of Standards. In all this chain of interrelated standards not a single one is absolutely correct. This fact at once gives rise to the question: How nearly correct must a standard be in order that it may adequately serve its purpose? Because of the varied requirements and conditions of use of standards, it is impossible to give a single answer that will apply in all cases. It is possible, however, to lay down some general rules that will help to decide, in a given case, whether or not a certain standard is adequate for the purpose intended. In order that there may be no misunderstanding in our discussion of corrections to standards, it would be well to define certain terms that will be used throughout the discussion-tolerance, error, and correction. Tolerance, as applied to weights and measures standards and to precision measurements in general, is a value defining the amount of the maximum allowable error or departure from true value or performance; or, in short form, the amount of variation permitted. There is lack of complete agreement among laboratories and among individuals in the definition and use of the terms "error" and "correction," and particularly there is lack of agreement in the use of plus and minus signs as applied to these terms. As a starting point suppose we agree that when we consider a weight, a measure, the indication of a scale, the length of a measuring tape, a measured quantity, or a measured distance, the error, in each case, is the amount of departure from true value. Let us agree further that if the indicated or observed value is in excess of the true value, then the error is plus, and if it is less than the true value, then the error is minus. Let us agree further that the "correction" in each of the above examples is the amount that must be added, algebraically, to the weight, the measure, the indication of the scale, the length of the measuring tape, the measured quantity or the measured distance, to make it correct. At this point, serious complications enter the picture and if we are to avoid confusion in the use of corrections and signs of corrections, we must at all times be sure whether we are considering the weight or the object weighed; the measure or the material measured; the measuring tape itself, a distance laid off with the tape, or a distance measured with the tape. A few examples may be helpful. A weight that is heavier than its nominal value has, by our definition, a plus error and a minus correc tion. That is, since the weight is heavier than its nominal value, it will have to be made lighter to make it correct. Now let us see what happens to the sign of the correction when a weight is used to compare other weights, to test a scale, or to weigh out a quantity of material. If a correct standard weight is used to test other weights, a weight that is heavier than the standard will have a plus error and a minus correction. Please note that the correction we are now considering is the correction to be applied to the weight under test. Again, consider the situation when a standard weight is used to test an indicating scale. If the scale under test indicates a greater amount than is actually on the scale platform, then the scale is "fast," it over-registers, and it has a plus error. It has a minus correction, since an amount must be subtracted from the indication of the scale to give the correct amount; or, speaking algebraically, a minus correction must be added to the indication of the scale to obtain the correct weight. If, as is sometimes the case, it is necessary to take into account a correction to the standard weight in determining the correction to be applied to the indication of the scale, the problem becomes further complicated, and it is not uncommon for the correction to the standard to be applied in the wrong direction. For that reason, if for no other, it is advisable to use standard weights that are of such accuracy that in the testing of other weights and scales, corrections to the standards need not be applied. If the standard used in testing the scale has an error that must be taken into account, then the final error in the indication of the scale is the apparent error minus the known error of the standard, and the final correction to be applied to the indication of the scale is the apparent correction minus the known correction to the standard. For example: Final error in scale indication=(+2 oz)-(-1 oz) = +3 oz. That is, the scale indicates 10 pounds 2 ounces when the actual applied load is 9 pounds 15 ounces, or it has an error of plus 3 ounces and a correction of minus 3 ounces. Many other examples of errors and corrections will readily come to mind. Let us turn now to the application of corrections to length standards, such, for example, as steel measuring tapes. If the tape itself is longer than its nominal value, it will have a plus error and a minus correction. An interval laid off, or "generated" with such a tape will be too long and will also have a plus error and a minus correction. On the other hand, a distance measured with such a tape will appear to be less than its actual length by the amount of this error, and an amount equal to this error will have to be added to the observed length to obtain the true length of the measured interval. The correction to the observed length of the measured interval is therefore plus. For that reason, a tape or other length standard which is longer than its nominal length is usually said to have a plus correction, since the correction must be added to the apparent length of an interval measured with such a tape. Strictly speaking, the plus correction applies to the length value indicated by the tape rather than to the tape itself. The term "correction to interval" may properly be used in this case. The same thing is true in regard to weights. If a weight is heavier than its nominal value, then, as already agreed, it will have a plus error and a minus correction, and objects compared with it will, if correct, appear to be light, and will therefore have a plus correction; that is, something must be added to the observed weight to give the true weight. Here again, strictly speaking, the correction applies to the observed value of the object weighed, and is opposite in sign from that of the weight with which it is compared. In many cases the safe thing to do is to report actual values of lengths, masses, capacities, etc., and thus avoid the possibility of confusion. Turning now to the testing of weights and measures, let us take two typical examples, (1) capacity measures, and (2) weights. In the testing of capacity measures used in trade it is customary practice to use capacity standards supplemented by a glass graduate. Now it is obvious that if we are to test capacity measures by the use of other capacity measures that are accepted as standards, these standards should be so nearly correct that their errors may be neglected, or, if that is not the case, then the errors in the standards must be known and taken into account in deciding whether or not the measures under test are within the specified tolerance. For example, a correct measure, if tested with an oversized standard, will appear to be small; while a measure that is in error by the same amount and in the same direction as the incorrect standard, will appear to be correct. If we could at all times hold the error in the standard to 10 percent, or less, of the permissible error in the measure to be tested, there would be little occasion to worry about the possibility of passing apparatus that ought to be rejected, or of rejecting apparatus that ought to be passed, as a result of errors in the standards. There would, of course, even then, be a few border-line cases that would be in doubt, but these doubtful cases would be relatively few. Unfortunately, it is often impracticable to hold the errors in standards to 10 percent, or less, of the tolerance on the measure to be tested. The error in the standard is much more likely to be 20 percent, 25 percent, or even 50 percent of the tolerance. Such errors cannot, of course, safely be neglected. Corrections to the standards must then be known and must be applied. In passing, it may be noted that in the field of interchangeable manufacture of machined parts such as bolts and nuts, shafts and bearings, where accurate dimensions are highly important, it is customary practice to use very accurate "go" and "not go" limit gages to determine whether or not the parts are within specified dimensional limits. Here, again, it is impossible to make either the parts or the gages exactly to the desired size. It is necessary, therefore, to apply tolerances to the gages as well as to the product; and since it is good gaging practice to keep the gage tolerances inside the product limits, it is desirable to keep the gage tolerances as small as is practicable in order that the tolerance remaining for use on the product may not be reduced below practical working limits. The situation is closely analogous to that with which you are all familiar, in which you use one class of standards to test other standards of the same type, but of a less precise class. If you are to be sure that you will not reject some that ought to be passed, or pass some that ought to be rejected, then the errors in your standards must be small in comparison with the tolerances on the measures under test. In our further discussion of capacity measures suppose we confine our attention to two sizes, the 5-gallon and the 1-gallon, and consider these as typical. The Conference tolerances on these sizes, as given in Handbook H22, are as follows: These tolerances will perhaps mean more to us if we express them in another way. For example, 6 liquid ounces on 5 gallons is equivalent to 6 parts in 640, or 1 in 107, about. Eleven cubic inches on 5 gallons is equivalent to 1 part in 105. The tolerance on the 1-gallon measure is 1 part in 128. These are the tolerances in excess. Those in deficiency are half these amounts, and the tolerances on new measures are half of these, in each case. To test these commercial measures we would ordinarily use either conical measures of the glass slicker plate type, accurate to 1 part in 2,000,10 or the "field standard," having a small diameter, graduated neck, accurate to 1 part in 2,000 for the 5 gallon and to 1 part in 1,000 for the 1 gallon. Thus the tolerance on the standard is about onetwentieth of the tolerance on the measure being tested in the case of the 5-gallon measure, and about one-tenth of it in the case of the 1-gallon measure. These standards, in turn, are tested at the Bureau by determining the weight of water contained or delivered, and the capacity is then calculated from the known weight of water per unit of volume at various temperatures. The determination of volume from the weight of water contained or delivered by the capacity measure involves, of course, a knowledge of the corrections to the standard weights used in making the weighings, and a knowledge of the density of water at various temperatures. These corrections will be considered further on in our discussion of corrections to standard weights. For the present it is sufficient to state that, in general, the corrections to standard weights are small in comparison with the tolerances on capacity measures. For example, the tolerances on class C commercial test weights of 10, 5, 2, and 1 pounds are, respectively, 4, 3, 1.5, and 1 grains, that is, from 1 part in 17,500 on the 10-pound weight, to 1 part in 7,000 on the 1-pound weight; whereas, the tolerances on capacity measures vary from 1 part in 2,000 for the 5-gallon measure of the most accurate type, up to 1 part in 105 (the tolerance in excess) on the 5-gallon measure of the commercial type. 10 The NBS certificate gives the correction to 1 part in 10,000. |