Quantization noise reduction in oversampled analog-to-digital signal conversion

摘要

Quantization noise in oversampled digital data can be reduced by processing including alternate steps of lowpass filtering and so- called code projection, in cases where the set of discrete-time signals corresponding to quantized-signal data is a convex set. Corresponding specific methods are disclosed for simple analog-to-digital conversion, "dithered" analog-to-digital conversion, and first- and second-order single-path &Sgr;&Dgr;-conversion. For example, in the case of simple analog-to-digital conversion, code projection involves replacing sample values by quantization-threshold values in those cases where, after filtering, a sample value lies in a different quantization interval as compared with the corresponding value of quantized-signal data.