10 



BELL SYSTEM TECHNICAL JOURNAL 



extremely large samples, the size is given to the nearest 10. This basis 

 of rounding sample sizes is followed for all of the sampling tables presented 

 with this paper. 



On each table are listed values of AOQL to indicate the upper bound to 

 the long term average per cent defective in product after inspection that 

 may be reached under the most adverse conditions. 



PROCESS AVERAGE RANGE IN PER CENT (FOR p,=5°/o) 



Fig. 3 — Basis of choosing the n and c values given in the sampling tables 



Double Sampling — Lot Quality Protection 



The solution for this plan is carried out in substantially the same way 

 as for single sampling. Protection is defined, as before, by specifying values 

 of lot tolerance per cent defective and Consumer's Risk. As for the single 

 sampling procedure, a Consumer's Risk value of 0.10 is adopted. In 

 double sampling, a lot is given a second chance of acceptance if the first 

 sample results are unfavorable, so that the Consumer's Risk is the sum 

 of two parts: (a) the probability of accepting a lot of tolerance quality for 

 the first sample, and (b) the probability of its acceptance for the second 

 sample, if the first fails. For example, if the two allowable defect numbers, 

 Ci and Co, are 1 and 7, respectively, the Consumer's Risk is the sum of the 

 probabilities for all of the following possible ways in which these criteria 

 may be met, as shown in Table 2. 



As in the case of single sampling, for any given process average value 

 there are a large number of acceptance criteria — pairs of C\ and Ci in this 



