THE CALIBRATION OF INSTRUMENTS 595 



be worked up while the test is in progress, which is highly desir- 

 able, as it insures that all necessary data are being taken and that 

 the procedure of the observers is correct. The check calibra- 

 tion eliminates questions as to the accuracy of the instruments 

 and as to whether or not they have been tampered with or 

 injured in any way. 



Choice of Instruments. In selecting the instruments for a 

 particular piece of work those should be chosen which will give 

 good deflections; that is, deflections in a favorable part of the 

 scale, and of such a magnitude that the required precision of 

 reading is readily attained. The choice, therefore, involves a 

 preliminary study of the conditions of the test in order to 

 determine approximately the magnitudes of the quantities 

 involved. It must then be decided whether or not the desired, 

 and obtainable, degree of accuracy in the final result is such that 

 careful calibrations are necessary. 



Often one knows from previous experience that his instru- 

 ments are correct to within 1 or 2 per cent, and instances are 

 continually arising where the conditions are such that an accu- 

 racy of 2 or 3 per cent is sufficient. In such cases, where 

 differences of nearly equal quantities are not involved, there is 

 no point in calibrating the instruments to 0.2 per cent, for 

 example. Again, there are many cases where the magnitudes 

 to be determined cannot be estimated a priori, and a certain 

 amount of rough preliminary work is necessary to determine 

 the most advantageous ranges of the instruments. In such a case 

 the calibrations should be deferred until the proper instruments 

 have been selected. 



Attention to these simple matters may save the beginner much 

 valuable time, and may possibly prevent his arriving on the 

 ground for a test without the proper equipment. 



Sources of Error in Instruments 



The various sources of error which have been referred to in 

 discussing particular instruments will be recapitulated. 



Errors of Reading. In general, the construction of the pointer 

 and the graduation of the scale should be such that under steady 

 conditions the position of the pointer may be read, by estima- 

 tion, to one-tenth of a scale division. This is readily attained 



