Statistical Circuit Design: Confirmation of Design Using Computer-Controlled Test Sets

Communication systems, both analog and digital, often use linear networks to which system performance is highly sensitive. In cases where the relationship between network characteristics and system performance are complex, computers are being increasingly used for network and systems design. Examples of such designs are given in other papers in this issue. The design process is not completed until the design intent is verified, first by measurements of the breadboard models and finally by measurements of the manufactured networks. When a physical network is to be evaluated, direct measurement of its effect on system performance requires having a system available 1197 1198 THE BELL SYSTEM TECHNICAL JOURNAL, APRIL 1971 as a test instrument. To be a satisfactory test, one would also require that the physical system represent all necessary worst cases. This approach has been used, but it has inherent difficulties such as the difficulty in obtaining a nominal or worst-case system, the difficulty in maintaining the system, and the lack of information on a network that has failed a test. These difficulties are largely overcome with the test method described in this paper. Using this method, the network is measured on a general purpose test set controlled by a computer. From these measurements of the network, parameters are calculated which predict the performance of the network in a system and which provide direct information about the network. The sections which follow give examples of such tests and the steps necessary to implement the tests and establish their validity.