1004 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1953 



was introduced, although no change in the assembly process was made. 

 The standard deviation of 0.00090'' and 3-sigma limits of 0.00270'' al- 

 though better than obtained with the tool maker's microscope were still 

 unsatisfactory. Based on the new average X for the process, a shift in 

 welding mandrel was made directed at producing an average nearer to 

 the 0.0750" value desired. The new welding mandrel was provided with 

 heavy spring clips to hold the plate sections in position during assembly. 

 A very uniform product resulted from this tool. The third portion of Fig. 

 12 represents production of plate assemblies using the new welding 

 mandrel. The average X of 0.07495" was almost exactly that desired and 

 the standard deviation of 0.00035" and 3-sigma limits of 0.00106" proved 

 satisfactory and a change in the specification was introduced. 



3.64 Final Test Characteristics 



The controls kept on the critical parts as well as on many other parts 

 of comparatively lesser importance have resulted in distribution of end 

 requirements that have been well within the tolerances in the test speci- 

 fications. In several cases the process average was found to differ from 

 the nominal value specified. An analysis of the control charts enabled 

 the design and manufacturing engineers to arrive at mutually acceptable 

 adjustments of either the process or the test specifications. Where satis- 

 factory performance would be obtained and the shift could be justified 

 on the basis of control chart records, the specified nominal was adjusted 

 to conform with the process average obtained. At the same time narrower 

 limits, assuring a more uniform product to the L3 amplifiers, were 

 adopted wherever it became evident that these were within the process 

 capabilities. 



4.0 Conclusions 



The quality control methods described in this paper provide practical 

 assurance that the selected characteristic of any component will have 

 an acceptable distribution centered close to the specified nominal value. 

 In this application, various statistical quality control techniques are an 

 actual part of the product specification. The additional effort required 

 in the application of such procedures is compensated for in part by the 

 valuable assistance they provide in controlling the manufacture of the 

 product. The procedures provide: 



1. Means for determining compatibility of the product specification 

 and the manufacturing process capabilities including the accuracy and 

 stability of measuring facilities. 



