Digital Signal Processing Reference
In-Depth Information
other domain is as follows:
y
ð
t
Þ!
y
k
!
y
k
c! ^
y
k
g
Continous Discrete Quantised Windowed
Discrete
Quantised
Discrete
In reality we get only windowed, discrete and quantised signal
fb
y
k
cg ¼ f^
y
k
g
at the
processor. Figure 1.10 depicts a 4-bit quantised, discrete and rectangular windowed
signal.
20
Signal
16
15
10
5
0
0
200
400
600
800
1000
1200
Sample
Figure 1.10 Windowed discrete quantised signal
1.5.4 Noise Power
The sample signal y
k
on passing through a quantiser such as an analogue-to-digital
(A/D) converter results in a signal
b
y
k
c
, and this is shown in Figure 1.9 along with
the quantisation noise. The quantisation noise
k
¼
y
k
b
y
k
c
is a random variable
2
, where n is
the length of the quantiser in bits. The noise power in decibels (dB) is given as
10 log
2
is given as
ð
1
12
Þ
2
n
(rv) with uniform distribution. The variance
=
ðÞ
2
¼
10 log
ð
12
Þ½
20 log
ð
2
Þ
n. If we assume that the signal y
k
is equi-
probable along the range of the quantiser, it becomes a uniformly distributed signal.
We can assume without any loss of generality the range as 0 to 1. Then the signal
power is 10 log
ð
12
Þ
. The signal-to-noise ratio (SNR) is 20n log
ð
2
Þ
dB or 6 dB per
bit.
Mathematically, the given original signal got corrupted due to the process of
sampling and converting into physical world, real numbers. These are the theore-
tical modifications alone.
Search WWH ::
Custom Search