Description
Transmitting analog audio, video, or sensor data with a digital system requires sampling and quantization. While a sampled bandlimited signal can be reconstructed perfectly if the Nyquist-Shannon sampling theorem is met, quantization unavoidably adds irreversible errors. Given a certain bitrate for quantization (or source coding in general), the achievable signal quality is therefore limited. Transmission of the signal over a noisy channel causes additional errors. If some part of the bitrate is used for protecting the signal with a channel code, these additional errors can be reduced or even eliminated. However, this reduces the achievable maximum quality if the gross bitrate is fixed. Since this compromise between quantization and channel coding is often designed for the worst-case channel, the signal quality saturates early with increasing channel quality, and transmission over a better channel is far below optimum.
Instead of conventional digital systems, this thesis cover
