966 



THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1953 



175 were packaged. Of the 25 units not packaged, 24 were in the upper 

 cell and 1 unit did not meet the A limits. At the end of the experiment 

 295 of the 300 units had been packaged. Of the 5 units not packaged, 3 

 remained in the upper cell, and 2 did not meet the A limits. The averages 

 of the packages are shown in the lower chart of Fig. 9. For comparison, 

 the PA limits are also shown on this chart. It is apparent from these two 

 charts that the three-cell method yields packages whose averages are 

 held closer to the nominal than are averages for packages from the con- 

 trol chart method. 



The corrective effect of the three-cell method is further illustrated by 

 Fig. 10, which shows the average of product packaged by the three-cell 

 method as a function of the process average. This curve is for "long 

 term" conditions, that is, it represents the expectancy for any given level 

 of process average. This corrective effect is purchased at the expense of 

 not packaging a portion of the product while the process average is not 

 at the nominal value. However, as already noted, the unpackaged por- 

 tion may be packaged with subsequent product if the process average 

 subsequently deviates from the nominal in the opposite direction. 



The percentage that can be packaged is also shown in Fig. 10 as a 

 function of the deviation of the process average from the nominal. It 

 should be noted that this curve also represents the expectancy for any 

 given level of process average. Of course, continued production at a 

 fixed level other than nominal would result in a steadily growing ac- 

 cumulation of unpackaged units, a situation that would call for corrective 



O.IA 0.2A 0.3A 0.4A 



PROCESS AVERAGE, X', (DEVIATION FROM NOMINAL) 



0.5A 



Fig. 10 — Expectancy curves for three-cell method. 



