Digital dither: decreasing round-off errors in digital signal processing

Dithering is a known method for increasing the precision of analog-to-digital conversion. In this paper the basic problems of quantization and the principles of analog dithering are described. The benefits of dithering are demonstrated through practical applications. The possibility and the conditions of using digital dither in practical applications are examined. The aim is to find an optimal digital dither to minimize the round-off error of digital signal processing. The effects on bias and variance using uniform and triangularly distributed digital dithers are investigated and compared. Some DSP implementations are discussed.