Dual quantization oversampling digital-to-analog converter

Abstract

An oversampling D/A converter system achieves high linearity and low quantization error by using dual internal D/A converters: one with high linearity but high quantization error and the other with low linearity but low quantization error. A signal path in the system includes an L-bit quantizer for quantizing a digital input signal into an L-bit signal plus a quantization error and a D/A converter for converting the L-bit signal and quantization error into an analog signal. A correction path that parallels the signal path includes an M-bit quantizer, where M is greater than L, and an M-bit D/A converter. A signal generated by the correction path cancels the shaped quantization error of the signal generated by the signal path by adding the negative of the shaped quantization error to the signal of the signal path. The resulting analog signal is filtered by a low-pass analog filter to extract the analog equivalent of the digital input signal. Various arrangements of filter stages are also disclosed for constructing preferred implementations of the oversampling D/A converter.