Variable Rate Transmission Over Noisy Channels

Abstract

Hybrid automatic repeat request transmission (hybrid ARQ) schemes aim to provide system reliability for transmissions over noisy channels while still maintaining a reasonably high throughput efficiency by combining retransmissions of automatic repeat requests with forward error correction (FEC) coding methods. In type-II hybrid ARQ schemes, the additional parity information required by channel codes to achieve forward error correction is provided only when errors have been detected. Hence, the available bits are partitioned into segments, some of which are sent to the receiver immediately, others are held back and only transmitted upon the detection of errors. This scheme raises two questions. Firstly, how should the available bits be ordered for optimal partitioning into consecutive segments? Secondly, how large should the individual segments be? This thesis aims to provide an answer to both of these questions for the transmission of convolutional and Turbo Codes over additive white Gaussian noise (AWGN), inter-symbol interference (ISI) and Rayleigh channels. Firstly, the ordering of bits is investigated by simulating the transmission of packets split into segments with a size of 1 bit and finding the critical number of bits, i.e. the number of bits where the output of the decoder is error-free. This approach provides a maximum, practical performance limit over a range of signal-to-noise levels. With these practical performance limits, the attention is turned to the size of the individual segments, since packets of 1 bit cause an intolerable overhead and delay. An adaptive, hybrid ARQ system is investigated, in which the transmitter uses the number of bits sent to the receiver and the receiver decoding results to adjust the size of the first, initial, packet and subsequent segments to the conditions of a stationary channel.