BELLINGER: HIGH FREQUENCY AMMETERS 415 



i.e., metal wires or strips which are very thin and of high resis- 

 tivity. Another expedient is to change the design so that the 

 deflections depend on the entire heat production in the instru- 

 ment instead of on that in one branch of its circuit. It was found 

 possible to apply this idea to the thermocouple instruments, as 

 well as to those whose indications depend on other thermometric 

 principles. Still another method of improvement is to arrange 

 the working parts (either wires or strips) as equidistant elements 

 of a cylinder, so that each has the same set of mutual inductances. 

 This design approaches as a limit the circular tube, which theo- 

 retically has no change of current distribution with frequency. 

 In common with the others, however, it is subject to an error 

 which has hitherto been overlooked. It is difficult to obtain very 

 thin wires or strips of uniform cross-section, and this variation 

 results in the resistances of elements being unequal while the 

 inductances are substantially equal. As a result the current 

 distribution may be uniform at high frequency and not at low 

 frequency. 



Eddy currents induced in neighboring masses of metal are 

 found to cause no error. Inductive action of the leads near an 

 instrument in some cases appreciably affects the readings, and 

 must be guarded against. The distributed capacity of the cir- 

 cuits within the instruments is found to cause no error, but the 

 capacity of auxiliary parts produces an appreciable effect at the 

 highest frequency used, 1,500,000 cycles per second. This ef- 

 fect was very striking, two instruments in series carrying differ- 

 ent amounts of current. Apparently part of the current was 

 shunted out of one of the instruments by electrostatic induction. 

 A means of eliminating the effect was found for the purposes of 

 ammeter comparison. This phenomenon suggests that the cur- 

 rent in a conducting circuit may not be a definite quantity at 

 frequencies over 1,000,000. 



In conclusion, it may be said that all types of ammeters in 

 use for large currents of high frequency are subject to errors. 

 The sources of error have been isolated and studied. Some were 

 shown to be negligible, and others very serious. Means of hn- 

 proving the designs have been worked out. 



