THE L3 SYSTEM — QUALITY CONTROL REQUIREMENTS 



959 



ACCEPTABLE / 



DISTRIBUTION -vr-^_ / 



OF INDIVIDUALS "^''^^- 



/ 

 / 

 / 

 / 



/ 



■-J-^'I— ^-< \ I I 



A LIMITS (N ± A) 



J. A50 LIMITS , 



(N±0.23A) 



PA LIMITS 

 (N ± O.IA) 



DISTRIBUTION 



OF SAMPLE 



AVERAGES 



(n = 50) 



-0.8A -0.6A -0.4 A -0.2 A 0.2A 0.4 A 0.6A 0.8A 



DEVIATION FROM NOMINAL 



Fig. 5 — Basis of A50 limits for sample averages, batch method. 



and sorted into three cells: lower, middle and upper, as shown in Fig. 6. 

 Units are then selected from the three cells in groups of 5 and packaged, 

 the five units in each package to be distributed among the three cells in 

 accordance with one of the two distributions shown at the top of Fig. 6. 

 These selected groups of five are maintained as packages of 5 in mer- 

 chandise stock and in deliveries to the user. 



Units in any cell which are in excess supply at any given time during 

 production and which, therefore, cannot be packaged may be included 

 with subsequent production provided each package of five satisfies one 

 of the two distributions shown in Fig. 6. 



For the three-cell method several matters were open to choice — the 

 relative width of the three cells, the number of units in a package, and 

 the required distribution of units in a package. Sampling experiments 

 were run with groups of 4, 5, 6 and 12 and the net effects studied prob- 

 ability-wise for various possible kinds of quality situations that might 

 be encountered in production. 



With appropriate choices of these items it is possible to provide as- 



