Digital Signal Processing Reference
In-Depth Information
and similarly, when the current is delayed:
X
N
1
Q
1
ð
n
Þ¼
1
N
u
1
ð
n
k
Þ
i
1
ð
n
k
N
1
=
4
Þ;
ð
8
:
46b
Þ
k
¼
0
where N
¼
mN
1
=
2
:
It can be noticed that it is possible to get very simple algorithms using aver-
aging, which results in Eqs.
8.39
,
8.43
,
8.45
and
8.46a
,
b
. In the first case only the
measurement error depends on sampling frequency. On the other hand, the algo-
rithms are sensitive to noise of either high or low frequency. The resulting errors
can be removed by prefiltering.
Example 8.3 With the averaging approach provide the equations of half- and full-
cycle algorithms for signal magnitude and power components measurement for the
sampling rate equal to 1000 Hz. Assess influence of sampling frequency on the
measurement errors. Specify sampling frequency and/or measurement period to
get errors less than 0.1%.
Solution For the assumed sampling frequency the number of samples per full
cycle of fundamental frequency component equals N
1
¼
f
S
=
f
1
¼
1000
=
50
¼
20.
Thus, for half-cycle algorithms one obtains N
¼
N
1
=
2
¼
10 and from Eqs.
8.39
,
8.43
,
8.45
and
8.46a
,
b
one gets:
X
1m
ð
n
Þ¼
0
:
157
X
9
j
x
1
ð
n
k
Þj;
k
¼
0
t
X
9
x
1
ð
n
k
Þ
X
1m
ð
N
Þ¼
0
:
447
;
k
¼
0
P
1
¼
0
:
1
X
9
u
1
ð
n
k
Þ
i
1
ð
n
k
Þ;
k
¼
0
Q
1
¼
0
:
1
X
9
u
1
ð
n
k
Þ
i
1
ð
n
k
5
Þ:
k
¼
0
Full-cycle algorithms are obtained by taking twice longer averaging window, i.e.
for N
¼
N
1
¼
20
:
Comparing to the above the scaling coefficients and summing
limits are changed, yielding:
X
1m
ð
n
Þ¼
7
:
85
10
3
X
19
j
x
1
ð
n
k
Þj;
k
¼
0
t
X
19
x
1
ð
n
k
Þ
X
1m
ð
n
Þ¼
0
:
316
;
k
¼
0
Search WWH ::
Custom Search