Current technology allows the recording of digital audio with 20 bits or greater resolution and nearly commensurate signal-to-noise ratios. Techniques have been proposed for reducing 20-bit or 18-bit audio to 16 bits ( for CD or DAT reproduction) using noise shaping, with minimal increase in perceived noise. The published proposals use an nth-order noise shaper to shape the truncation quantization noise to be minimally audible according to a selected noise weighting curve. Published simulations of 1 LSB (16 bit) of DAC differential nonlinearity, with ninth-order noise shaping of the quantization noise, show that much of the advantage of the noise shaping is lost. Can available DACs reproduce these signals? How little nonlinearity must the DAC have to retain the advantages of noise shaping? What other problems occur with DACs that would affect the utility of noise shaping? This paper addresses these questions.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.