944 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1953 



lated deviation arising in 1,200 amplifiers to a few tenths of a db is a 

 formidable problem. 



The second problem is concerned with the effect of equalization on the 

 signal-to-noise performance. As deviations creep into a repeatered circuit, 

 the transmission levels deviate more and more from the normal or design 

 levels. This effect is lumped in one term, misalignment. The result of 

 misalignment is degradation in signal-to-noise performance. Periodic 

 equalization helps to limit misalignment in the succeeding repeater sec- 

 tions but does not eliminate the increase in noise or modulation which 

 has occurred in the preceding repeaters. Thus, the objective is not only 

 to equalize the over-all circuit, but also to keep the deviations all along 

 the transmission line within the specified bounds in order not to exceed 

 signal-to-noise margins. 



1.2 NEED FOR CONTROLLING DEVIATIONS AT THEIR SOURCE 



Let us examine what accounts for the magnitude of the gain-versus- 

 frequency deviations arising in any one repeater. Let us assume that a 

 particular element deviates +1 per cent from the design objective. Is 

 this good or bad? The answer to this question may be had only if a 

 deviation study of the repeater is made and its sensitivity to the deviation 

 of the element under consideration is ascertained. There are two factors 

 which contribute to the equalization problem: (a) the deviation sensi- 

 tivity of the repeater to a given deviation of the element from its pre- 

 scribed value, and (b) the actual deviation of the element itself due to 

 all causes, including manufacture, temperature, aging, etc. To simplify 

 our terms, factor (a) will be called the sensitivity of the element, and 

 factor (b), the deviation of the element. The sensitivity is a function only 

 of the circuit design and is independent of the performance of the indi- 

 vidual element. The deviation of the element is a function only of its 

 design and manufacture. 



1^ USE OF STATISTICAL QUALITY CONTROL METHODS TO ASSURE A CON- 

 TROLLED DISTRIBUTION 



When the design of the L3 system was initiated, it was realized that 

 the effect of the variability of the component elements could be ma- 

 terially reduced by the application of statistical quality control tech- 

 niques to design and manufacture. Once a circuit design is available and 

 a deviation sensitivity study is made, it is comparatively easy to formu- 

 late the desired limits on the variability of components. Actually the 

 process of arriving at an individual tolerance objective is more complex 



