Z Systems rdq-6 Instruction Manual - Dither
rdq-6 instruction manual
z-systems audio laboratories
18
F
IGURE
19
•
dither
In order to understand dither, we must briefly discuss the problems associated with
quantization and re-quantization. First, quantization is the process whereby a continuous-
valued analog signal is converted into a discretely valued digital signal. In the case of
compact disc digital audio, an audio signal is represented by 16-bit samples taken 44,100
times per second. A 16-bit system comprises 2
16
=65,536 quantization steps. Thus, 16-bit
quantization consists of sampling the analog waveform uniformly in time and approximating
each sample by the nearest of the 65,536 quantization steps. Conventional quantization
begins to show its weakness on source material with wide dynamic range (i.e., a very high
ratio between the levels of the loudest sounds and the softest sounds, typical of orchestral
music). In order to make room for the loudest sounds, the quietest sounds are "pushed"
below the smallest quantization step, leading to signal-dependent quantization noise. This
distortion can manifest itself in many different ways including harshness, edginess, loss of
reverberant detail, and alteration of timbre. Loosely speaking, then, a digital audio system's
performance is bounded by a level that is established by the size of the quantization step
used.
This same type of distortion can occur when doing digital-to-digital operations, as well. For
example, to ensure a high-quality end product, many compact discs are mastered at 20 bits or
24 bits as an intermediate format for editing, equalization, compression, and level adjustment.
One of the final steps of the mastering process is to convert the 20- or 24-bit digital signal to a
16-bit signal for final release on the CD format. The problems here are more or less the same
as those associated with quantization of an analog signal. The 20- or 24-bit signal has much
wider dynamic range than the 16-bit signal and signal-dependent distortion can occur when
making the transfer to 16 bits. In this case, the mastering engineer must deal with the
problems associated with re-quantization noise.
Any type of DSP algorithm such as volume control or tone control involves mathematical
operations on an input signal which result in the accumulation of extra bits. For example, the