312 THE BELL SYSTEM TECHNICAL JOURNAL, MARCH 1953 



in dialing a full code. This behavior affects the holding time of registers 

 receiving the dialed digits and must be considered in throwdown studies. 

 Data collected on combined waiting and dialing time characteristics 

 show a frequency and time distribution that can readily be quantized 

 into a number of values, each equally hkely to occur. When the number 

 of values for a particular throwdown study are determined, each quan- 

 tized dialing time is represented by a number. Each ultimate good call 

 is then assigned a number from a random list of the representative 

 numbers to estabUsh the dialing time of the call. 



Ultimate good calls will develop one of three terminating conditions 

 attributable to subscriber behavior: 1) DA, called subscriber does not 

 answer, 2) busy tone because of called subscriber line busy, or 3) answer 

 by called subscriber. It is assumed from analysis of ''don't answer" 

 studies that, for certain throwdown evaluations of the smtching system, 

 approximately 10% of the ultimate good calls meet the DA condition. 

 The nimiber of busy tone terminations, of coiu*se, Avill develop during 

 the throwdown study as a result of the average originating and termi- 

 nating calling rate per subscriber served by the system. 



Most subscribers, upon encountering a line busy condition make subse- 

 quent attempts to reach the called line. The number and frequency of 

 attempts made depend upon the individual characteristics. A detailed 

 analysis of this characteristic, suitable for use in throwdown studies, 

 has appeared in a paper by Charles Clos.* 



When calls are answered by called subscribers, the connections will 

 be held for varying amounts of time. It has been determined from field 

 observations that the frequency distribution of these holding times 

 is closely approximated by an exponential distribution. For throwdown 

 purposes a simplifying assumption can be made that holding time is not 

 a continuous variable but is quantized so that a particular holding 

 time will have one of several values. To determine these values an ex- 

 ponential distribution having the proper average is plotted as shown in 

 Fig. 7. The area under the curve is then divided into the required number 

 of equal parts (ten, for this example). A central value of holding time is 

 determined to represent each subarea. Ten holding times are thus pro- 

 duced which are weighted according to the exponential distribution and 

 which are equally likely to occur. These holding times can be designated 

 to 9 and assigned to the calls by choosing one-digit numbers at random 

 for each call. 



* Charles Clos, "An Aspect of the Dialing Behavior of Subscribers and Its 

 Effect on the Trunk Plant," Bell System Tech. J., 27, pp. 424-445, July, 1948. 



