The Reliability of Telephone Traffic Load 

 Measurements by Switch Counts 



BV W. S. IIAY\VAI{1), Jli. 



(Maiiuscii|il rcccivcMl ( )cl()l)i'i' 15, 1051) 



The .s'(r//r7(. coiinl mcllwd of Iclcplwuc lidjlic mcafiurcmeitl /.s suhjcci to sam- 

 pling errors. The nature of these errors is discussed and formidas are de- 

 rived which describe the extent of the errors under normaHy encountered 

 traffic conditions. 



IXTRODUCTION 



Of prime importance to the telephone traffic engineer is the deter- 

 mination of the busy season busy hour load carried by groups of trunks 

 or other circuits of a telephone switching system. Three direct methods 

 of measuring such loads are found in the field today. These are: 



a. Peg Count and Holding Time Method 



The number of calls carried by the circuit group during the observa- 

 tion period is counted. This number multiplied by the average holding 

 time per call (in hundreds of seconds) and divided by the length of the 

 observation period (in hours) gives an estimate of the group load in 

 units of hundred-call-seconds per hour (CCS). The major drawback to 

 this peg count method is that it requires a separate determination of the 

 average holding time per call for the group under observation. R. I. 

 Wilkinson^ has analyzed the sources of errors of holding time measure- 

 ments. In addition, correlation between load and holding time introduces 

 an error which has not been studied. 



b. Switch Count Method 



At fixed intervals the circuit group is scanned and the nvnnl)cr of busy 

 circuits is counted. The total number of busy conditions cijunted divided 

 by the number of scans is, then, an estimate of the load on the group in 

 units of average simultaneous calls or erlangs*. This estimate is generally 

 converted to CCS (1 erlang = 3G CCS) by ti-affic engineers since the 



* The name "erlang" for average simultaneous call was adopted at a plenary 

 meeting of the CCIF at Montreux in October, 1946. 



357 



