Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Chapter 5
Frequency Transformation
We need to apply a suitable frequency transformation, if we wish to design
bandpass, bandstop and high-pass filters, using the low-pass approximating
function analysis previously covered. The block diagram, shown in figure
5.25, illustrates the procedure, which produces, from the specification
supplied, the required high-pass approximating function.
Before we consider frequency transform techniques, lets consider the second order
series-tuned LCR circuit shown in figure 5.26. Here we obtain expressions for the
standard second-order transfer functions for low-pass, high-pass, and
bandpass filter configurations.
1/sC
Vc(s) = Vin(s) (5.1)
1/sC + R +
sL
Reworking the equation, multiply above and below by sC and divide both sides by
V c(s) 1
T(s) = 2 (5.2)
Vin (s) s LC + sCR +1
1 L
p2 = and Q p (5.5)
LC R
s R/L
T(s) = 2
(5.6)
s +(R/L)s + 1/CL
In standard form
K( p /Q)s
T (s) (5.6a)
s2 + ( p /Q)s + p
2
Consider the output voltage across the inductance L. The transfer function is:
2
s
T(s) = 2 (5.7)
s +(R/L)s + 1/CL
Ks2
T (s) (5.8)
s2 + ( p /Q)s + p 2
The specification for a high-pass filter includes the passband edge frequency, hp,
and the stopband edge frequency, hs The maximum passband attenuation is Amax
and the minimum stopband attenuation Amin. The transformation of
is
high-pass specification to an equivalent normalisedthe low-pass specification is
achieved by applying the frequency transform SL = 1/s, where SL is the low-pass
normalised complex frequency variable. However to account for the process
of normalization we must replace SL by 1/(s/ p) =p /s.
p
SL (5.9)
s
p
S L (5.10)
1 / np
$ (5.14)
s
In the Chebychev case, we apply the substitution
p
$ (5.15).
s
Example 5.1
Obtain the transfer function H(s) for a Chebychev high-pass filter, which satisfies
the following specification:
Passband edge frequency p = 3000 rs -1
Stopband edge frequency s = 1000 rs -1
Maximum passband attenuation Amax = 0.5 dB
Minimum stopband attenuation Amin = 22 dB
Solution
The order of the filter is calculated:
1/2
100.1 Amin 1
-1
cosh
100.1 Amax 1
n
cosh (LS )
-1
Substituting values
1/ 2
100.1.22 1
cosh 0.1.0.5
-1
10 1
n -1
2.42
cosh (3/1)
p
$=
s
A(s) = A($)
s s2
H1(s) = 2
3000+ 0.626s 9.10 + 1.879.2s + 1.142s
6
1.602s s2
H1(s) = 2 2
s + 4807 s +1644s +
7.878.106
Circuit implementation:
100.1Amin 100.1.20 1
log10 10.1 A max log 10 0.1.2
n 10 1 10 1
4 2log( LS ) 2log(3000/1500)
Substituting values:
0.764 1/ 4 3000 2800
$
s s
Substituting this into the required high-pass approximating function which, when
inverted, gives the required H-P gain Inverting.
H(s) = 1/A(s)
s
2
s
2
H(s) = 2 6 2 6
s 2142s 7.84.10 s + 5174.4s 7.84.10
Where
s1 = the lower stopband edge frequency
s2 = the upper stopband edge frequency
p1 = the lower passband edge frequency
p2 = the upper passband edge frequency
If we consider the normalised low-pass specification in figure 5.32 see that
we has a mirror image A(-) i.e. its an even function of A(). This is considered
A()
a bandpass function whose centre frequency, is at = 0 and the
frequency range from - to + . The frequency transform required to
carry out this
frequency shift is: s o
$ 2 2 (5.16)
B.s
B p2 p1 (5.17)
o p1 p2 s1s2 (5.18)
s 2 2o 1/
$ (5.21)
n
Bs
To obtain the normalised values, we consider evaluating the transform at the
following frequencies: s1, s2, p1 and p2.
( j)2 02
$
(p2 p1 ) j
But = j so that
2 0 2
j j
(p2 p1
Or
0 2 2
(p2 p1 )
Substituting the frequency values s1, s2, p1 and p2 into this equation.
Figure 5.34
1) At = o
0 2 0 2
Lo =0
(p2 p1 )0
2) At =s1 and 2 o s1s2
s12 s 1
Ls1 Ls2 s1 s 2 = s2
( p2 p1) s1 (
p2
)
3) p1
At = s2 and 2 o s1s2
s 22 s 2 s 2 s1
Ls2 s 1 s 2 = s1 =
( p2 p1) s2 (
p2 p2)
) p1 (
p1
4) At =p1 and 2 o p1
p2 p 1 p 2 p 12
Lp1 Lp2 1
( p2 p1)
5) At =p2 and 2 o p1 p2 p1
p 1 p 2 p 22
Lp2 1
( p2 p1)
p2
Example 5.3
The specification for a bandpass filter is
s1 = the lower stopband edge frequency = 1000 rs-1
s2 = the upper stopband edge frequency = 9000 rs-1
s 12 s 2 s1 9000 1000
Ls1 Ls2 s1 s 2 = 3.2
( p2 p1) s1 ( p2
) 4500
2000
p1
1/ 2
100.1x Amin 1
-1
cosh
100.1.xAmax 1
n
cosh -1 (LS1 )
Substituting values
1/ 2
10 0.1.6 1
cosh 0.1.5
-1
10 1
n -1
cosh (3.2)
1.24
Example 5.4
A bandpass filter is required for a particular application, which will tolerate ripple
in the passband equal to 0.5 dB. The specification for a bandpass filter is:
s1 = the lower stopband edge frequency = 688 rs-1
s2 = the upper stopband edge frequency = 1930 rs-1
p1 = the lower passband edge frequency = 970 rs-1
p2 = the upper passband edge frequency = 1370 rs
-1
Solution
The maximum allowed ripple present in the passband is 0.5 dB so a Chebychev
design is considered. The order n is calculated as
1/ 2
100.1 Amin 1
-1
cosh 0.1 Amax 1
n 10
cosh -1
( S )
s 12 s 2 s 1 1930 688
Ls1 Ls2 s1 s 2 = 3.1
( p2 p1) s1
p2 p1
1370 970
Substituting values
1/ 2
-1 100.1.15 1
cosh 0.1.5
n 10 1 1.93
-1
cosh (3.1)
s 2 o2
Using the frequency transformation $ in the loss function. Here B is the
B.s
3 dB bandwidth, which can be expressed:
B p2 p1 = 400 rs-1 .
.
s 2 2o s 1153
2 2
$
B.s 400s
s2 1153 2 s2 1153 2
2
1 160000s2
H (s) 4
A(s) s 570s3 2.95.106 s2 7.5.108 s 1.74.1012
s 2 2o 1/ n
As an exercise use the Butterworth $ substitution on this
Bs
problem.
Exercise
Obtain the transfer function for a low-pass CR filter and a high-pass CR filter for C =
1F and R = 1 . Substitute s = 1/s in the low-pass filter and compare with the hpf
transfer function.