Permanence of the Rutherfurd Photographic Plates. 20'7 



4. We shall now consider the errors caused by a deviation from 

 perfect straightness in the guiding cylinder of the machine. It 

 was not at first intended to take these errors into account in the 

 present researcli, for the reason that it was certain a priori that 

 they would be extremely small. Moi'eover, there was some delay 

 in securing the apparatus needed for the investigation of this 

 matter. The computations were therefore carried out without the 

 application of any corrections for lack of straightness of the 

 cylinder. The c^dinder was, however, very carefully investigated 

 afterwards, and it was found that no appreciable errors could be 

 observed. The method employed for the investigation of the 

 cylinder was that used by Gill in the investigation of the Cape 

 of Good Hope measuring machine. It consists in comparing the 

 cylinder with a stretched spider thread. For this purpose a 

 stretched spider thread was carefully mounted on an iron plate, 

 which could be inserted in the machine in the same way that the 

 ordinary photographic negatives are put in. The spider thread 

 being then made very nearly parallel to the cylinder by means of 

 the position circle, it was merely necessary to move the plate 

 along the guiding cylinder, and bisect the spider thread at various 

 positions with the micrometer microscope. We thus measure 

 directlj' the departure of the cylinder from a straight line, pro- 

 vided we assume that the spider thread itself is exactly straight. 

 But as the thread was stretched with considerable tension, it is 

 not possible that it should deviate at all from a sensibly straight 

 line. Its own weight would of course make it hang in a catenary 

 curve, but it is the projection of this curve upon a horizontal 

 plane that we bisect. And this projection would of course be 

 perfectly straight. As a further precaution, all the measures 

 were repeated with the thread-plate turned through an angle of 

 1 80°. In this way any error of the cylinder will be determined 

 separately b};- two entirely different parts of the thread. If there 

 be real errors of the cylinder, tbey should come out sensibly the 

 same from the two positions of the thread-plate. 



In order that it may be possible to fix the points of the c^dinder 

 for which the errors are determined, there is an auxiliary scale 

 attached to the machine. This scale is parallel to the cylinder, 

 and is attached to the casting that carries the plate. A rough 

 auxiliary microscope serves to read this scale to the nearest tenth 

 of a millimetre. The numbering on this scale runs from 200 to 



