Digital Signal Processing Reference
In-Depth Information
y
1
ð1Þ¼v
1
ð1Þþjv
1
ð0Þ¼2 þ j 1 ¼ 2 þ j
v
1
ð2Þ¼v
1
ð0Þþxð2Þ¼1 þ 3 ¼ 2
y
1
ð2Þ¼v
1
ð2Þþjv
1
ð1Þ¼2 þ j 2 ¼ 2 þ j2
v
1
ð3Þ¼v
1
ð1Þþxð3Þ¼2 þ 4 ¼ 2
y
1
ð3Þ¼v
1
ð3Þþjv
1
ð2Þ¼2 þ j 2 ¼ 2 þ j2
v
1
ð4Þ¼v
1
ð2Þþxð4Þ¼2 þ 0 ¼2
y
1
ð4Þ¼v
1
ð4Þþjv
1
ð3Þ¼2 þ j 2 ¼2 þ j2
Then the DFT coefficient and its squared magnitude are determined as
X ð1Þ¼y
1
ð4Þ¼2 þ j2
jX ð1Þj
2
¼ v
1
4
þ v
1
3
¼ð2Þ
2
þð2Þ
2
¼ 8
Thus, the two-sided amplitude spectrum is computed as
r
jX ð1Þj
2
A
1
¼
1
4
¼ 0:7071
and the corresponding single-sided amplitude spectrum is A
1
¼ 2 0:707 ¼ 1:4141.
From this simple illustrative example, we see that the Goertzel algorithm has the following
advantages:
1.
We can apply the algorithm for computing the DFT coefficient
XðkÞ
for a specified frequency
bin
k
; unlike the fast Fourier transform (FFT) algorithm, all the DFT coefficients are computed
once it is applied.
2.
If we want to compute the spectrum at frequency bin
k
, that is,
jXðkÞj
, Equation
(8.71)
shows that
we need to process
v
k
ðnÞN þ
1 times and then compute
jXðkÞj
2
. The operations avoid complex
algebra.
If we use the modified Goertzel filter in
Figure 8.54
,
then the corresponding transfer function is
given by
G
k
ðzÞ¼
V
k
ðzÞ
1
1
2 cos
2
pk
N
XðzÞ
¼
(8.77)
z
1
þ z
2
Search WWH ::
Custom Search