124 GENERAL CONSIDERATIONS. 



With all the instruments which have been described this 

 accuracy can be relied upon to a considerable degree of 

 precision. In all of them, readings can be taken to one 

 ten-thousandth of an inch in ten inches. This is ecmivalent 

 to a degree of precision of one in one-hundred-tnousand, 

 which is quite sufficient for any ordinary purposes. In 

 the case of some of the extensometers an even greater 

 degree can be depended upon. The writer has tested ;m 

 extensometer of the lever multiplying kind, by means of 

 a Whitworth measuring machine, and found it perfectly 

 accurate within the degree of precision mentioned. 



An extensometer should be so constructed as to be 

 easily and quickly fixed to the specimen, ready for the 

 test. Some instruments are very deficient in this respect, 

 and much time is wasted in getting the instrument 

 properly set. 



Lightness of construction is an advantage so long as 

 strength and rigidity are not sacrificed. As in testing 

 machines, simplicity of design is of the utmost impor- 

 tance. A multiplicity of parts adds to the weight as 

 well as to the cost. There is greater liability to derange- 

 ment, and the instrument is much more difficult to adjust. 



The accuracy of extensometer readings should be 

 made to depend not upon direct measurements of lever 

 ratios, but the instrument should always be directly cali- 

 brated in the manner already described ; and, moreover, 

 this calibration should be repeated occasionally so as to 

 make sure that the adjustment has not been disturbed. 



In using an extensometer it is important that 

 it be fixed on the proper part of the test bar; other- 

 wise the readings obtained may be inaccurate. Many 

 test pieces are initially curved to a slight extent, and, as 

 the load upon one of these is increased, it not only 

 stretches but tends to straighten at the same time. The 

 effect of this straightening is an unequal strain on the two 

 sides of the bar ; the straightening will have the effect of, 

 to some extent, counteracting or lessening the stretch of 

 the bar on its convex side and increasing the stretch on 

 the concave side. If, therefore, the measurements be 

 taken on either concave or convex side, they will be rendered 

 inaccurate by the straightening of the bar ; a similar effect 

 is sometimes produced by the load upon a bar not being 

 applied centrally, and the consequent curving of an initially 

 straight test piece. It has been shown by Professor Unwin 

 that either of the effects mentioned will be accompanied by 



