70 



Messrs. Paterson and Duddino- on the 



with sub-standards eventually covering a range of from 

 1600° C. to about 2700° C. The authors are at present 

 experimenting with gas-filled standard lamps with the inten- 

 tion of carrying the work a stage further should they prove 

 satisfactory. 



It is found that four more steps are required to connect 

 up the existing set 6 with the standards required for the 

 testing of present-day gas-filled lamps. These are indicated 

 by horizontal dotted lines in fig. 1, marked 7, 8, 9, and 10. 



Lamj)s used in the Investigation. 



The types of lamps used in the investigation are shown in 

 fig. 3. Some of them, such as the Fleming-Ediswan lamps 



rr~r 



TU 



Representative Lamps from each Set of Sub-standards. 



(set 2) date back from 1902, and are the type of carbon fila- 

 ment standards due to Dr. J. A. Fleming, F.R.S., which have 

 been used at the Laboratory since then for maintaining the 

 unit of candle-power. Set 4 are " gem " lamps with so- 

 called metallized carbon filaments, and were the best lamps 

 which could be found at the time they were installed for 

 giving constant results at approximately 3*5 watts per mean 

 horizontal candle. The remaining lamps have tungsten fila- 

 ments and have been specially made for the Laboratory by 

 the Osram Lamp Works and presented at different times 

 by the General Electric Co. Reliable tungsten standards 

 which can be depended on for constancy when operated at 

 high efficiencies are not easy to obtain. Many tungsten 

 lamps are liable to small erratic fluctuations of candle-power, 

 the reason for which is obscure. Such fluctuations being 

 small, do not affect the lamps for ordinary lighting purposes, 

 but are a serious drawback for standard work. Each set of 



