244 S. P. Langley — Bolometer. 



grams, and its six needles, of proportionate weight and dimen- 

 sions. 



This system is now made to serve with a much shorter 

 swing than that formerly employed. If we reduce it to a time 

 of single vibration of 20 seconds, only for the purpose of com- 

 paring it with the values already given in the earlier form, we 

 obtain the results submitted below. 



Before giving them, however, it is to be mentioned that the 

 apparatus at Washington is most unfavorably situated, owing 

 to its being subject to tremor from the traffic of neighboring 

 streets and to other causes, which it has been the object of 

 years of struggle to conquer. This has been so far done that 

 the values presently to be given (which, it will be remembered, 

 are only attainable in a chamber of constant temperature, with 

 special precaution against disturbance from external tremor), 

 can be counted on as real values, always obtainable under 

 proper conditions, and, in fact, rather within than without the 

 average working capacity of the instrument. 



I here consider the bolometer as at present employed. 



(1) With regard to its precision, or exactness of pointing. 

 The old thermopile could be set on a portion of the spectrum 

 only with an error of a considerable fraction of a degree. The 

 linear bolometer as employed in 1886 could be set with a prob- 

 able error of a fraction of a minute of arc. The bolometer as 

 employed to-day, and moved through the spectrum by clock- 

 work, can be automatically set with a probable error of a single 

 observation of little over a second of arc, can be set, that is, 

 in the dark with a precision little inferior to the capacity of 

 the eye in setting a micrometer thread in the light. 



(2) As to its accuracy. I have had occasion recently to take a 

 series of measures of successive throws of the galvanometer, 

 using as a source of heat an Argand petroleum flame in a com- 

 mon student's lamp. I had no photometer at hand, but taking the 

 usual statements of the text books as to the accuracy of vision, 

 it might be expected that such measures with the eye would 

 give a probable error of about one per cent. (This is where 

 sources of light of similar quality are compared.) The 

 probable error of a single galvanometer reading was between 

 0*03 and 0*04 of one per cent., and this included the 

 fluctuation of the intensity of the source of radiation, and the 

 error of estimating tenths by the reader on the scale, both 

 quantities of nearly the same order as the error in question. 

 It seems safe to say, then, that no error attributable to inac- 

 curacy of the bolometer could be detected by the means 

 employed. 



(3) As to sensitiveness. In the early work, for a time of 

 single swing of 20 seconds, a deflection of one millimeter with 



