What is Dithering and Noise Shaping?

These are some types of artificial methods for processing digital sound.

signal, intended to improve subjective sound quality
at the cost of an obvious deterioration of its objective characteristics (mainly
– the non-linear distortion coefficient and the signal-to-noise ratio).

Blurring is the addition of a small amount of
the amount of noise (pseudo-random digital signal) of different spectrum
(white, pink, etc.). In this case, the error correlation is noticeably weakened.
Useful signal quantization side (rounding errors “dissipate”)
and despite an increase in noise, the subjective sound quality
noticeably increases. The level of added noise is selected based on
of the task and goes from the middle of the least significant digit of the countdown to
various categories.

Noise Shaping is the transformation of
Noisy useful signal to displace purely noise components
nent in the supra-tonal region with a selection in the lower part of the spectrum of the main
useful signal energy. Basically, Noise Shaping is one
from the PWM view (Pulse width modulation – pulse width module-
tion, PWM) with discrete pulse width. The signal processed by this means
Therefore, it requires mandatory filtering with high frequency suppression:
this is done digitally or analog.

The main application of Noise Shaping is in the field of digital representation.
signals with lower bit depth counts more frequently
following. The delta-sigma DAC to increase the sample rate
the sampling rate increases tenfold, at which
initial multi-digit readings, a series of discharge readings are formed
ness 1..3. The low frequency part of the flow spectrum of these samples with high
how accurately the spectrum of the original signal is repeated, and the high frequency
it contains mainly pure noise.

In the case of converting a digital signal into lower resolution samples
row at the same sample rate Noise shaping is performed
along with the Dithering operation. Since in this case the hour increase
sampling is impossible, instead the added noise spectrum
is formed in such a way that its low and middle frequency part
repeated as accurately as possible the weak part of the signal enclosed in the
the least significant bits of the samples. Thanks to this, the main energy of noise
ma is shifted to the top of the operating frequency range, and in most
the audible area is still fairly legible traces of a weak signal,
otherwise it would be completely destroyed. Although
the objective distortion of the weak signal stored in this way is very
are large, your subjective perception is still quite acceptable, allows
listening to components, whose level is lower than junior
count of downloads.

Essentially, Dithering and Noise Shaping are special cases of one
technology – with the difference that in the first case, no
noise with a uniform spectrum, and in the second – noise with a spectrum, special
customized for a specific signal. This technology leads
to the “non-standard” use of digital format based on
characteristics of the human ear.