THROWDOWN MACHINE FOR TRAFFIC STUDIES 331 



call, and so on for the remaining types of calls. If divisions of less than 

 one per cent are to be made, three or more digits could be used for this 

 purpose. 



The meanings of certain of the remaining digits will depend upon 

 the type of call. If the call is originated within the office, the next five 

 digits will give the identity of the calling line in terms of its frame 

 location. If the call is incoming from another office certain of these five 

 digits will be used to specify the office of origination and the trunk num- 

 ber. This again is on a percentage basis depending on the fraction of 

 total incoming traffic expected from each office. Five other digits give 

 the identity of the called line if the call is completed within the office; 

 if outgoing they specify the terminating office. Other digits give the 

 percentage of calls which will result in partial dialing by the calling 

 subscriber. Since, as will be discussed later, there is a possibility that 

 these will be re-originated later as good calls, all of the information for a 

 good call is also determined for these calls. Additional random numbers 

 determine various other aspects of the call such as the identity of the 

 number group which will be used. As a suggestion to those who would 

 undertake a throAvdown study, it is advisable to include a number of 

 surplus "utility" digits in the original random number. It invariably 

 happens that some factor is overlooked in the initial planning or arises 

 during the course of a test and these digits can be used in making de- 

 cisions in these cases. 



It should be noted that the random circuit is used for making certain 

 random decisions in the course of a call at the time that a need for these 

 decisions arises. For example, the holding time of an established con- 

 nection and the probability of the called subscriber not answering is 

 determined by this circuit. This circuit could have been eliminated by 

 including digits in the original number for every possible situation of 

 this type which might be encountered. This would cause much useless 

 work in preparing the data since all situations are not encountered on 

 every call. 



A quantity of random numbers must be drawn to provide the desired 

 load on the office. This is not a simple process of determining the number 

 of calls expected during a given period and drawing this number of 

 random numbers. One factor is that in generating data by the above 

 procedures it will be found that certain numbers will represent calls 

 originated by lines Avhich are busy at the indicated time on a connection 

 established previously. The number of such cases can be estimated from 

 the expected number of busy fines in the office and a corresponding 

 number of additional numbers drawn. When this situation is encountered 



