C-Net Week #21 20 февраля 2000

Программирование

Sound - The paradoxes of digital sound: an article on analog-digital conversion / digital-analog conversion.

The paradoxes of digital sound.

DJ.SERG / ACL



    Hi all membersam
Grodno network C-No! I
then decided to highlight some of
the problem of sound processing. We
talk about the analog-digital
transformation, as well as
talk about the digital-analog
transformation.

        Type preface ...;-)



    Demap
audio signal into a digital
proceeds in several stages.
First, the analog signal
served on the analog filter,
which limits the bandwidth
signal frequency and eliminates
interference and noise. Then, from
analog signal using a
sampling scheme / Storage
stand counts: with
periodically
by memorizing
instantaneous
analog signal. Next
samples arrive at the ADC
which converts the instantaneous
value in numbers. Obtained
sequence of bits
digital code and is
audio signal into digital
form (about the size of this
sequences of talk
later). Still, I note that for
protect against possible errors, and
and to harmonize with
channel parameters recording
digital code can be entered
service information.


    Major role in the process
signal conversion from
analog to digital
is the ADC (Analog / Digital
Converter-ADC).


    Let us consider more in detail
analog-digital conversion


    Discretization (in
literature can be met and
such a thing as
oversampling, it is also
thing) is the process of taking
sample after a specified
time interval (the interval
sampling).


    Demap
Digital Signal
carried out by
encoding instantaneous
values ​​of its amplitude
(Counts) 16-bit
binary code (Fig. 1). Topics
************** Most
************** Provided
************** Opportunity
************** Map
************** Audio
**************( Program
Dynamic Range:

    D = 6n +1.8 = 6 * 16 +1.8 = 98db, DG
n-number of bits (16).
Readings followed by a frequency
44.1, 48, 88.2, 96 kHz. This
Frequency is the frequency
sampling.

    From the theory it is known that
for proper recovery
the original form of analog
signal is necessary to
sampling rate of Fd was
at least higher than the maximum
Fmax frequency spectrum
displayed a sound
signal. As a rule,
Fd = (2.1-2.4) * Fmax.

    In the CD format high
frequency range is 20 kHz,
therefore, the sampling frequency
44.1 kHz is well suited to
purposes of converting such
signal. Since the system
stereo, then this
transformation are
both channels.

    Now let's talk about
digital to analog conversion
and its main element, TsAPe
(DAC-Digital/Analog Converter)

    Strictly speaking, from
DAC block dominates the whole principal
sound parameters: the ratio of
signal to noise ratio, dynamic
range, bandwidth
Frequency and
Tone and his painting.

    If we talk about
Digital audio features
equipment, it's worth
remember that many of
They derive from the nature
the initial screening process -
analog-digital
transformation.

    Demap
Digital Signal
(Quantization) is performed in
CD system with 16-bit
accuracy. This means that
full range of possible
changes in the signal amplitude
divided by 2 ^ 16 = 65536
identical levels of-step
quantization. Through
intervals
time-sampling period
amplitude
estimated, and her
match the value placed
nearest quantization level
(See Figure 2). The error,
**************** Which, when
**************** This may
**************** Arise, not
**************** Should
**************** Exceed
****************) Half step
quantization + -1 / 2 * 65536, which
corresponds to approximately
distortions about 0.0015%.
Signal / noise
is obtained about 98dB.
Distortions of this type
called quantization noise.
If we compare its value with
magnitude of noise in analog
recording (70 dB) and the coefficient
Distortion (0.3%)
the magnitude of the quantization noise
can be considered
negligible. But this
value is valid
negligible only at high
level of musical
signal, when its amplitude
close to maximum
(Fig. 2a).

    If a phonogram
found fragments of the Pacific
sound - say, such
which is coded 03.04
discharges, the ratio
between useful signal and
quantization noise dramatically
changes (Fig. 2b). At the level of
signal encoding
which is enough 4
bits, the noise
quantization will be
1 / 2 ^ 4 = 6.25%. And this is a lot.

    Incidentally, note that in
analog recording weaker
signal, so it is cleaner, since
There is less nonlinear
distortion. With digital recording
all the way around, the more
signal, so it is better, but
weak signals transmitted from
distortions.

    Nevertheless,
gradation of the digital
signal should not inspire
fatal horror. After all, even from
sequence of rectangular
pulses can be distinguished
perfect sine wave (1-w
harmonic). So what really
talk about the steps in 65536
times smaller.

    And about TsAPah. Must
say that vysokorazryadnye
TsAPydovolno expensive
fun because of the complexity
their production. When
manufacture of such
converters required
very accurate fit
weighting resistors,
are the basis of
multibit DAC. For
16-bit DAC, this
accuracy is the same
amount as required
resolution of 1 / 65536. For
18-bit DAC is already
will be 1 / 262144, and for
20-bit - 1 / 1048576.
To obtain such accuracy in
current conditions difficult.
Because of this real DACs
always some imet
nonlinear transformations.

    Sometimes, instead of one DAC
in each stereo channel
use two. At the same time
one of the DAC converts
positive half
signal, the other -
negative. Effect of
its inclusion in the general
case is equivalent to an increase
resolution on
one bit.