A Hybrid Analog-digital Coding Scheme for Digital Sources
Author: Lu Li
Publisher:
Published: 2018
Total Pages: 115
ISBN-13: 9780438600409
DOWNLOAD EBOOKTraditional analog and digital coding systems present their own advantages and constraints. Digital systems show excellent BER performance, but optimizing them over a wide range of signal to noise ratios (SNR) requires switching between different systems (e.g., adaptive coding). Analog systems allow high throughput communications over a wide dynamic range, but their bit error rate (BER) performance is not as good as in digital schemes. The aim of this work is to explore how to overcome these drawbacks by properly integrating both analog and digital systems into a hybrid one. ☐ The proposed hybrid coding scheme is realized by parallel concatenation of Rate Compatible Modulation (RCM) and a Low Density Generator Matrix (LDGM) code. RCM generates its output symbols by standard linear combinations of input bits. In addition, an LDGM code produces a few output bits. Because of dense constellation and fine-grain energy accumulation, RCM is able to achieve smooth rate adaptation, but it experiences performance degradation due to the presence of error floors. The introduction of the LDGM code allows to reduce most of the residual errors, substantially improving the system performance. ☐ This work considers the application of the hybrid scheme to point-to-point AWGN channels and to multiple access channels (MAC). For AWGN channels, our study focuses on optimizing the design for uniform and non-uniform memoryless sources. Decoding is implemented by building an appropriate graph and performing belief propagation. The influence of several design parameters on the system performance and the existing trade-offs are discussed. The proposed hybrid coding scheme achieves better performance than traditional joint source channel coding techniques in terms of gaps to the theoretical limit. We also introduce a simplified decoding technique for the proposed hybrid scheme. The idea is to treat the coded symbols as real values, and to use a Gaussian approximation, so that very simple analog message passing can be applied. This substantially reduces the decoding complexity, while the system performance does not experience significant degradation. ☐ For the MAC, a synthetic decoder structure and decoding algorithm has been developed and studied to tackle the ambiguity problem resulting from the existence of multi-level symbols proceeding from the RCM sub-block. The synthetic decoder also reduces the computational complexity by jointly considering messages propagated in the graph. As we will see, the hybrid scheme can achieve high transmission rates with good BER performance in MAC environments, even when the simplified decoding technique is applied.
