NCR-9415685, Johns Hopkins University, "An Investigation of the Interaction Between Codes and Decoding Machines", PI-Collins: The essential concept of this research is to study decoding, itself, as a communications problem. Electronics has advanced to the stage where many chips, each containing tens or even hundreds of thousands of gates, can advantageously be exchanged for a decrease in antenna aperture in a satellite terminal or for a lower transmitter power in a remote buoy or cellular telephone. These latter applications are well able to use extremely powerful coding because of the asymmetry of cost in the two directions of communication. In the case of cellular telephones, for example, power savings are much more important in the link from the mobile unit than in the link to the mobile unit. A sophisticated decoder in the base station drawing power from the AC mains can save battery power in the mobile transmitter. Similarly, a large scale environmental data collection network need incorporate only simple and cheap encoders in the transmitters; a single decoder in a satellite or ground station can service all of these encoders. Unfortunately, exploiting even a small fraction of the processing power of a single piece of silicon is hard, let alone that of a module containing many chips. The difficulty is providing sufficient internal communication among the different parts of the decoding machine. Today, wiring connectivity rather than gate count limits the sophistication of signal processing. This research presents bounds on the amount of information which has to flow in and out of the different sections of a decoding engine when it is partitioned into separate parts. At the moment, these bounds are confined to the Viterbi decoding of shift register based codes. This research will extend these bounds to other codes and decoding algorithms and, more importantly, explore codes (both convolutional and general finite state machine based) which allow decoding with minimum information flow. The same techniques used in analyzing decoder information flow will also be applied to other situations where partitioning is important, e.g., very large circuit switches. Dissecting decoder information flow may result in a new generation of codes of great practical use and will certainly yield insights into the ultimate limits of the type of codes in use today. ***************************************************************************** Aubrey M. Bush Program Director, Acting Deputy Divison Director Division of Networking and Communications Research and Infrastructure National Science Foundation
