Dr. S. P. Thompson on Photometry. 125 



As the intrinsic light of the arc is not far from 70 candles 

 per square millimetre, a circular aperture 1 millimetre in 

 diameter will afford a light of about 55 candles. Hence 

 an aperture smaller than this is preferable in a photometric 

 standard for all ordinary photometric work. For the purpose 

 of a special photometric standard with which to compare 

 other arc lamps, a standard light of 1000 candles or some 

 such magnitude is doubtless advantageous. This would 

 require about 14' 3 square millimetres of crater-surface, or a 

 circular aperture of about 4*25 millimetres in diameter. For 

 ordinary photometric work an aperture of 0'674 millimetre 

 diameter would give a light of about 25 candles. It is easy 

 to ensure the illumination of the apparent aperture by employ- 

 ing a* magnet to deflect the arc to the front face of the 

 carbon. 



There is an advantage in using as a standard source of 

 light one whose light is greater than that of the light under 

 measurement. The light under measurement may obviously 

 be balanced either against a smaller light brought nearer to 

 the screen, or against a greater light placed at a greater 

 distance. But if the greater light at the greater distance be 

 employed, there will, for any given inequality of illumination 

 on the screen, be required a greater actual displacement of 

 the moving part, whether screen or light, in order to arrive 

 at the position of balance. Hence any error in reading the 

 scale will be a lesser fraction of the quantity to be measured. 



It might have been supposed, a priori, that for a given total 

 length of photometric bench (as in a Bunsen photometer from 

 lamp to lamp) the position of the photometric screen which 

 would make errors a minimum would be the position in the 

 centre ; balance being sought between two approximately 

 equal lights. Geometrically it is true that a given displace- 

 ment along the scale produces a 1 minimum change in the 

 difference of the two illuminations, if the point selected be at 

 the centre of the scale. On the other hand,, if the total 

 length of the bench be not fixed, it is obvious that the length 

 of bench w T hich will, for a given change in the difference 

 between the two illuminations, yield the greatest actual 

 displacement along the scale will be an infinite length. But 

 in that case, unless infinite lights are used, the two illumina- 

 tions will both be zero. In fact there is another element to 

 be considered, namely, that in order to make a satisfactory 

 comparison, the eye requires a certain actual brilliancy of 

 illumination in the two surfaces which are to be compared 

 together. Doubtless the habitual patterns of photometer in 

 use have come by natural selection to be of the size that they 



