Digital Signal Processing Reference
In-Depth Information
which is another form of Discrete Fourier Transform (compare (
4.26
)) enabling for
transformation of the N
1
-element series in time domain into an equivalent N
1
-
element series in frequency domain (exact to the constant coefficient 2/N
1
).
8.1.1.4 Correlation Method
The digital correlation consists in representing the considered real signal with a
series of mutually orthogonal functions [
8
]. For power-system protection appli-
cations such representation is mostly limited to the components of fundamental
frequency only. The expansion components can be obtained from (
8.22
) and (
8.23
)
assuming calculation for the discrete frequency k = 1, which yields:
;
N
1
1
X
X
1C
¼
2
N
1
n
2p
N
1
x
ð
n
Þ
cos
ð
8
:
25
Þ
n
¼
0
:
X
N
1
1
X
1S
¼
2
N
1
n
2p
N
1
x
ð
n
Þ
sin
ð
8
:
26
Þ
n
¼
0
For a pure fundamental frequency signal the values resulting from (
8.25
) and
(
8.26
) are:
X
1C
¼
X
1m
cos
ð
u
Þ;
ð
8
:
27
Þ
X
1S
¼
X
1m
sin
ð
u
Þ;
ð
8
:
28
Þ
being constant, in contrast to the outputs of digital Fourier filters that are
time-depending functions. Nevertheless, also here the harmonic distortions (if
any) are completely rejected by the correlation algorithm and the obtained values
X
1C
and X
1S
are orthogonal, which enables calculation of signal magnitude, e.g.
with application of (
8.50
).
If the correlation algorithm is to be applied on-line, where time instant
n changes, it is more convenient to use the algorithms in the form:
;
N
1
1
X
X
1C
ð
n
Þ¼
2
N
1
x
ð
n
m
Þ
cos
ð
n
m
Þ
2p
N
1
ð
8
:
29
Þ
m
¼
0
:
X
N
1
1
X
1S
ð
n
Þ¼
2
N
1
x
ð
n
m
Þ
sin
ð
n
m
Þ
2p
N
1
ð
8
:
30
Þ
m
¼
0
The advantage of correlation approach is a possibility of obtaining simple
recursive algorithms. Observing that for the samples n and n - 1 the sums (
8.29
)
and (
8.30
) differ only by two elements, one can derive the following recursive
relationships:
Search WWH ::
Custom Search