366 



THE BELL SYSTEM TECHNICAL JOURNAL, MARCH 1952 



2 4 6 8 10 20 40 60 100 200 400 1000 2000 4000 10,000 



LOAD IN HUNDRED-CALL-SECONDS PER HOUR 



Fig. 4- — Hours of measurement required for 90 per cent assurance that error in 

 estimating source load does not exceed plus or minus 5 per cent when measuring 

 interval is 120 seconds. 



requirements may then be read directly from such graphs. In the second 

 type of question, z, NT, scan rate and holding time are fixed so that 

 zVs is proportional to l/\/a. Plotting log zVs against log s/a again 

 gives a linear plot as sho^\ai on Fig. 6. 



In the numerical example above, the limits of error corresponding to 

 90 per cent assurance may be read from Fig. 6 which is plotted for the 

 appropriate assurance, average holding time and scan interval. Reading 

 the error limits at the point where the 10 hours measured line crosses 

 180 CCS (5 erlangs) gives ±8.1 per cent as before. Fig. 5 may be 

 entered to find the total number of hours required to reduce this error to 

 5 per cent. Reading at the point where the 180 second holding time line 

 crosses 180 CCS gives 26 hours. 



QUALITATIVE EXTENSION OF THEORETICAL APPROACH 



The original traffic assumptions made in deriving the theoretical re- 

 sults above are: 



a. Calls originate collectively and individually at random. 



