This talk aims to
present a mathematical study of the robustness issue in
analog-to-digital (A/D) conversion. Any A/D converter has to operate
with analog devices, and is therefore highly susceptible to component
variations. A simplistic point of view assumes that the accuracy of a
converter is dictated by the accuracy of the arithmetical operations
implemented in its analog circuitry. However this is far from the
truth, as the encoding mechanisms of these converters can be designed
to be robust against specific types of imperfections by introducing
redundant codewords. These encoders operate very differently compared
to classical error-correcting codes as their 'corrupted' codewords
differ greatly in the Hamming metric, but very little in their decoding
neighborhood. We will present some explicit examples of encoding
strategies based on dynamical systems ideas.