424 THE BELL SYSTEM TECHNICAL JOURNAL, MARCH 1953 



points is found which could be used for the concentrating switch. In this 

 case the new approach to the switching network problem may prove to 

 be of value. 



Comparisons with existing arrays having blocking are likely to be 

 unfavorable because the grades of service are not the same. For in- 

 stance, a No. 1 crossbar district-to-office layout of 1,000 district junctors 

 and 1,000 trunks requires 80,000 crosspoints. This layout can handle 

 708 erlangs with a blocking loss of 0.0030. The minimum number of 

 crosspoints with a non-blocking array is slightly less than 138,000. This, 

 however, can handle 1,000 erlangs without blocking. By introducing 

 blocking into the design methods described in this paper, a more favor- 

 able comparison with existing arrays having blocking can be made. 

 This can be done by omitting certain of the paths. If done to an array 

 requiring 1,000 inputs and 1,000 outputs a layout can be obtained re- 

 quiring 79,900 crosspoints with a blocking loss of 0.0022 for a load of 

 708 erlangs. For this example, at least, it appears that the new design 

 methods may prove to be valuable especially for use in the development 

 of electronic switching systems where the control mechanism may not 

 be dependent upon the particular switching array used. 



CONCLUSION 



In present day commercial telephone systems the use of non-blocking 

 switching networks is rare. This may be due to the large number of 

 crosspoints required. With the design methods described herein, a wider 

 use of non-blocking networks may occur in future developments. For 

 the usual case of networks with blocking, new systems have generally 

 been designed by an indirect process. Several types and sizes of switch- 

 ing arrays are studied until the most economical one for a given level 

 of blocking is found. With the new design methods, a straightforward 

 approach is possible. Fig. 5 indicates that a region of minimum values 

 exists. By first designing a non-blocking system with a reasonable number 

 of switching stages and then omitting certain of the paths, the designer 

 can arrive at a network with a given level of blocking and be very close 

 to a minimum in crosspoints. The possibility of the adoption of this 

 direct design method is important. 



ACKNOWLEDGEMENT 



In addition to those specifically mentioned in this paper, the author 

 is also indebted to E. B. Ferrell, B. D. Holbrook, C. A. Lovell and E. 

 F. Moore for suggestions and encouragement in the preparation of this 

 paper. 



