Digital Signal Processing Reference
In-Depth Information
y
ð
n
Þ¼
CU
m
cos
ð
nX
þ
u
þ
b
Þ:
ð
8
:
143
Þ
A ratio of output and input magnitudes is equal to the zero-order Walsh FIR
filter gain for given frequency:
Y
m
U
m
¼
C
¼
W
0
ð
jX
Þ
j
j:
ð
8
:
144
Þ
The filter gain is given by:
j
W
0
ð
jX
Þj ¼
sin N
1
2
2
;
ð
8
:
145
Þ
X
sin
while the frequency characteristic is shown in Fig.
6.2
.
The gain is equal to zero for nominal frequency (X
¼
X
1
¼
2p
=
N
1
) and the
same is for magnitude of the output signal. As it is seen from Eq.
8.144
the gain
value can be measured using ratio of output Y
m
and input U
m
magnitudes. Having it
we can rearrange complex function of frequency to find it or its deviation. The last
possibility leads to simpler equation. One can notice first that:
N
1
X
2
¼
N
1
ð
X
1
þ
DX
Þ
2
¼
p
þ
N
1
DX
2
ð
8
:
146
Þ
and substituting it to (
8.145
) and then to (
8.144
) yields:
N
1
DX
2
C
¼
sin
:
ð
8
:
147
Þ
X
1
þ
DX
2
sin
If DX and N
1
DX are very small then one obtains:
:
2
N
1
sin
X
1
2
DX
¼
C
ð
8
:
148
Þ
The measurement error is a result of simplifications made. Its value is a fraction
of percent for small frequency deviations.
Simpler formula is obtained when X
1
þ
DX is very small (by high sampling
frequency) and both sine functions in (
8.147
) are substituted by their arguments.
Simple rearrangements give:
DX
X
1
C
N
1
C
:
¼
ð
8
:
149
Þ
This is the simplest equation allowing to calculate small frequency deviations.
To apply it is necessary to know the ratio of output and input magnitudes of zero-
order Walsh filter. Having it one can calculate frequency deviation from (
8.149
).
For greater frequency deviations the measurement error increases significantly.
One can avoid it using finite difference of output signal (
8.143
) at discrete time
n and n
1. Then:
Search WWH ::
Custom Search