The Bell System Technical Journal 



Vol. XX October, ig4i No. 4 



The Reliability of Holding Time Measurements 



By ROGER I. WILKINSON 



I — ThK PkoBI.KM 10 BK SOLVKD 



ONE of the fundamental quantities in tratBc engineering is the average 

 duration of subscribers' calls. This figure in seconds multiplied by 

 the average number of calls expected over a given route in an hour, and 

 divided by 3600, gives the traffic load submitted in average simultaneous 

 calls — or "the average" as it is commonly called. Tables and curves are 

 widely available which may then be consulted to find the number of paths 

 to be provided so that no more than a desired small percentage of the calls 

 presented will find all paths busy. 



The direct measurement of call lengths with a stop watch occurs to one 

 as being the simplest means for obtaining a sample of holding times. It is 

 seldom used, however, due to the relative slowness with which a large 

 number of observations are accumulated coupled with the not inconsiderable 

 expense of the small army of observers required, each looking at one call 

 at a time. 



A second direct method of obtaining holding time measurements is by 

 recording mechanically or electrically the length of each call passing over a 

 group of switches or trunks during a certain interval of time. Various 

 holding time recorders or "cabinets" following this principle have been used 

 more or less extensively in the Bell System. Their chief disadvantage has 

 lain in requiring considerable time and labor for summarizing the results. 

 Problems of the perfect maintenance of the measuring equipment have also 

 been present. 



To make possible the rapid accumulation of holding lime data on a 

 considerable number of calls at relatively slight expense, the method of 

 switch or plug counts has been introduced. This consists in scanning 

 mechanically, electrically, photographically or by eye the group of paths at 

 regular intervals, and recording each time the number found busy. Such 

 data give estimates immediately of the average load being carried, and by a 



' This numbt-r will be highly variable, and even on projjerly engineered groups greal 

 concern need not be felt should few, or even no, cases of "all paths busy" api)ear since 

 such peaks are of short duration and might easily be missed except in a very long series 

 of counts. 



365 



