THE L3 SYSTEM AMPLIFIERS 881 



greater precision than we can economically control the individual. This 

 is exactly the effect which modern methods of statistical quality control 

 aim at achieving, and since the entire amplifier is merely the sum of its 

 parts, quality control must start at the roots of the manufacturing proc- 

 ess. In order to apply quality control intelligently, and to be sure that 

 all important causes of gain variation are understood, it has been neces- 

 sary to carry out, side by side with empirical laboratory work, a pro- 

 gram of computing the insertion gain of the amplifier, starting with 

 fundamental element values. These computatuions have also proven of 

 value in obtaining satisfactory stability margins in the design of the 

 low^ and high frequency cut-offs of the feedback loops, and in obtaining 

 preliminary information on amplifier gain deviations for use in equali- 

 zation planning. 



The severe gain and delay reproducibility objectives also have their 

 effect on the mechanical design of the amplifier. At first sight the unit 

 appears to be a lumped constant structure rather than one in which 

 distributed effects would be of paramount importance. Usually the cir- 

 cuit designer in such a case is interested in the mechanical design only 

 for reasons of neatness and economy, but when we look deeper we find 

 that in the amplifier structure as a whole as well as in the case of certain 

 components, the effects of distributed capacity and inductance could 

 easily defeat our objectives if the mechanical design were not such as 

 to assure precise control of element placement and wiring lengths. 



For these reasons, an order of mechanical accuracy is specified beyond 

 that w^hich can be justified by the accuracy of transmission measurements 

 on individual amplifiers. This striving for mechanical accuracy is carried 

 all the way through, from element piece parts through subassemblies 

 to the assembly on the final amplifier framework. The logical expectation 

 is that by reproducing the amplifiers as exactly as possible we will con- 

 trol not only the known elements but also the many parasitic effects, and 

 thereby minimize the appearance of shifts in amplifier gain and delay 

 which would be substantial in the system even though diflftcult to detect 

 in the measurement of individual amplifiers. 



Another design feature of the amplifier based on the system equaliza- 

 tion point of view is the omission of all adjustable elements, or trimmers, 

 to control the transmission. By eliminating gain adjustments, the pos- 

 sibility of adjusting one element to compensate for the short-comings of 

 another, and the possibility of systematic errors of setting due to faulty 

 or inaccurate adjustment techniques, are both eliminated. Either of 

 these possibilities would tend to convert relatively large random effects 

 into smaller but systematic effects, a conversion which would penaUze 



