358 THE BELL SYSTEM TECHNICAL JOURNAL, MARCH 1952 



load entries of most traffic tables are in terms of CCS. For theoretical 

 studies the erlang is a more convenient unit and wdll be retained here. 



c. Continuous Method 



The busy condition of each circuit is represented by a fixed increment 

 of electrical current through an ampere-hour meter. The instantaneous 

 current is then analogous to the calls simultaneously present so that the 

 meter, which integrates the current, may be cahbrated to indicate hun- 

 di'ed-call-seconds or erlang-hours directly. Although this method is po- 

 tentially the most accurate, practical difficulties have limited its use. 



In addition to these direct methods, there are several methods of in- 

 direct load measurement which, relying more heavily on traffic theory, 

 make use of partial load indications, such as duration of group busy or 

 the number of calls finding the group busy. Such measurements are less 

 reliable than the direct measurements particularly when apphed to un- 

 derloaded groups. 



This paper is concerned with the reliability of switch count load meas- 

 urements since this method appears to have prospects of considerably 

 wider adoption in the future. Main emphasis will be placed, both quali- 

 tatively and by the appHcation of error formulas, on the relative effects 

 of various measurement and traffic parameters on the accuracy of s'^itch 

 count measurements. Where long derivations of formulas are required 

 they are deferred to the Appendix. 



SOURCES OF ERROR 



As has been described, switch count measurements yield the average 

 number of calls found present when a group of circuits is scanned at 

 fixed intervals during an observation period. Usually only that period 

 of the day during which the load is greatest is of interest to the traffic 

 engineer. Because the load during such periods also fluctuates from day 

 to day, measurements of the loads for several days must be averaged to 

 provide a useful load estimate. 



There are two main sources of error, therefore, in sudtch count esti- 

 mates of telephone traffic loads: 



1. Each individual count of busy circuits is separated from the next 

 by a time interval during which changes in load are not detected. Con- 

 sequently, the load indicated by measurement may differ appreciably 

 from the actual load carried. This difference can be decreased by de- 

 creasing the interval between scans. 



