Signal conversion techniques in digital audio applications

The performance of digital audio equipment depends ultimately upon the signal transparency of analogue-to-digital and digital-to-analogue conversion gateways. Although the initial specification for a digital audio channel was 16 bit, modern systems are now seeking to achieve a subjective dynamic range approaching 20 bit. This requirement sets formidable engineering requirements especially where circuity must be of low cost. Modern VLSI and signal processing techniques allow an exchange of system complexity where analogue circuitry can be simplified at the expense of an increase in digital system complexity. The presentation will review a number of the signal processing techniques now being used including oversampling and noise shaping. In particular the techniques of bitstream and pulse-width modulation is described as alternatives to the more common multi-bit converter. Finally, consideration will be given to information capacity of a linearly quantized digital audio channel where it will be shown that noise shaping and dither together with appropriate models of human perception enable coding, than in a subjective sense, can achieve the dynamic range of a 20 bit system using only 16 bit data. >