In this test the carriage is moved up against the stops h several times, and several readings or tests are made, so as to see that the force of the contact of the carriage against the stops h is uniform at each test, and if any variation is found, the average of a number of readings is taken. It is found, however, that with practice the carriage may be moved against the head h by means of the hand-wheel with such an equal degree of force that an error of not more than one fifty-thousandth of an inch is induced. It is found, however, that if too much time is occupied in this test, the heat of the operator’s body will affect the temperature of the bars, and therefore expand them and vitiate the comparison. But in this connection it may be noted that if a bar is at a temperature of 40°, and is placed in an ice bath, it does not show any contraction in less than one minute, and that when it does so, the contraction is irregular, taking place in sudden movements or impulses.
Professor Rogers’ methods of testing end measures are as follows: To compare a line with an end measure, a standard bar is set upon the machine, its horizontal and vertical lines being adjusted true to the cylindrical guides by the means already described, and the microscope carriage is so adjusted that the spider web lines of the microscope coincide with the horizontal and vertical lines marked on the standard, while at the same time the stop (u, [Fig. 1350]) on the carriage k has contact with the fixed stop (v, [Fig. 1350].) Carriage k is then moved along the cylindrical guides so as to admit the bar (whose end measure is to be compared with the lines on the standard) between the two stops, and if, with the bar touched by both stops u and v, the microscope spider lines intersect the vertical and horizontal line on the standard bar, then the end measure corresponds to the line measure; whereas, if such is not the case, the amount of error may be found by noting how much movement of the micrometer wheel of the microscope is required to cause the lines to intersect.
It is obvious that in this test, if the cylindrical guides had a horizontal curvature, the test would not be perfect.
Fig. 1356.
The Horizontal Curvature.—The copy or bar to be tested may be set between the stops, and the standard bar may be placed on one side of it, as in [Fig. 1356], and the test be made as already described. It is then set the same distance from the bar to be tested, but on the other side of it, as in figure, and again adjusted for position and tested, and if the readings on the standard bar are the same in both tests, it is proof that the measurements are correct.
Suppose, for example, that the cylindrical guides were curved as in [Fig. 1356], it is evident that the vertical lines would appear closer together on the standard bar when in the first position than when in the second position.
In the Rogers machine the amount of error due to curvature in the cylindrical guides in this direction is found to be about 1⁄5000 part of an inch in 39 inches, corresponding to a radius of curvature of five miles.