Integrated Data Systems

Abstract

Two theoretical data systems (Shannon's ideal channel with average power limitation and the ideal symmetrical binary channel) are compared with the recent results of HF-data transmission tests over a 5000-mile link. The 30-to-40-db increase in signal power required in the practical case is analyzed. NonGaussisn noise, multipath disturbances (fading), and lack of coding are reasons usually mentioned to explain this poor efficiency of presently available data systems. The present paper points towards one additional reason: low-order signal alphabets. Binary or quaternary sets of transmission signals can not achieve the ideal channel capacity as long as decision systems have to operate at 8-to-10-db "energy contrast" (E/N_{o}) for a sufficiently low error rate. To more nearly approach the ideal capacity, it appears necessary to follow most of ten important principles when designing an "integrated" data system. Operating with higher-order signaling systems and applying error-correcting codes are the two principles which promise the largest increase in transmission efficiency. All the rest of these ten "Cardinal Principles" are discussed briefly.