510 



HENRY A. ROWLAND 



of parallelism between the screws and the ways or eccentricity in the 

 screw mountings thus scarcely affects the forward motion of the car- 

 riage. The guide against which the steel pieces of the nut rest can be 

 made of such form as to correct any small error of run due to wear of 

 the screw. Also, by causing it to move backwards and forwards peri- 

 odically, the periodic error of the head and mountings can be corrected. 

 In making gratings for optical purposes the periodic error must be 

 very perfectly eliminated, since the periodic displacement of the lines 

 only one-millionth of an inch from their mean position will produce 



m 



FIG. 2. 



" ghosts " in the spectrum. 1 Indeed, this is the most sensitive method of 

 detecting the existence of this error, and it is practically impossible to 

 mount the most perfect of screws without introducing it. A very prac- 

 tical method of determining this error is to rule a short grating with 

 very long lines on a piece of common thin plate glass ; cut it in two with 

 a diamond and superimpose the two halves with the rulings together 

 and displaced sideways over each other one-half the pitch of the screw. 

 On now looking at the plates in a proper light so as to have the spec- 



1 In a machine made by the present writer for ruling gratings the periodic error is 

 entirely due to the graduation and centering of the head. The uncorrected periodic 

 error from this cause displaces the lines ^^fa^ih of an inch, which is sufficient to 

 entirely ruin all gratings made without correcting it. 



