Audio and Video Digital Signal Conversion

by Elliptic Filter and MATLAB Software

Kai-Long Hsiao
Department of the Digital Recreation and Game Design, Taiwan Shoufu University, 168, Nansh Li,
Madou Jen, Tainan, Taiwan, Republic of China,
Email: [email protected]

Abstract— In this study, towards analog and digital frequency, so the other frequency selective
elliptic filters, use of transfer function through the choice filter is also applicable, such as high-pass,
of a particular filter structure from a set of design
specifications to achieve the step change in transfer band pass and band stop filter. For the
function analysis. Elliptic filter design with two frequency conversion of analog design with a
optimization criteria (1) maximally flat amplitude diameter cut-off frequency at different
response and (2) small amplitude ripple response, and
the excellent performance of the filter available. There degrees of order low-pass Butterworth filter
are many different filter design methods discussed in this is normalized Butterworth low-pass filter
paper is the first for the analog method, used in analog start and uses the filter cutoff frequency and
filters. The specified frequency response with amplitude
response and phase response which obtained through the order of replacement which is s functional.
rational transfer function approximation. In addition to This is an example of frequency conversion.
the order of the filter for different analysis, the use of Using the same method can be normalized
continuous signals into discrete Z-transform system by
using (1) Pulse constant conversion method (2) Bilinear low-pass filter, converted to other types of
transformation method two conversion methods. This frequency selective filters. Finally, the output
paper and using MATLAB software to calculate the of the system to produce a continuous time
analysis results, and the results plotted graphically
confirmed that the two algorithms are calculated from signal reconstruction filter, on behalf of the
the Z-transform is very similar to results with the original input signal is filtered results. This
correct solution, using MATLAB software is used to paper and using MATLAB software to
evaluate the convenience, and By the results of its
analysis to support the correctness of the conversion analyze the results of calculations [6], and
result, and confirmed that the bilinear transformation the results graphically plotted for reference
method than the pulse of the same conversion method comparison. Since elliptical filters better than
has a better conversion results.
the Butterworth filter's performance, so this
Keywords— Audio and Video, Elliptic filter, Signal
processing, Matlab, Bilinear transformation method, article gives special seminars. Recently a
Impulse invariance method more linear elliptic filter, designed to be
more flexible and has a wide bandwidth [7,8],
I. INTRODUCTION
on the other filter current transmission
Elliptic filter widely used in the realization control is also a new development [9], and
of speech processing. Because in speech have very low band-pass sensitivity and
analysis and speech synthesis that has the compensation effect [10], usually require
excellent processing power ability, the high-order elliptical filter before they can
mathematical model of analysis by a transfer achieve the above effect [11-14]. Through
function can be described, or you can use the references [1-14] and to consolidate research
amplitude response, phase response and discussion to make a design of high-
described. Through literature Ma et al. [1] order elliptic filter is one of the points this
applied to the circuit signal analysis, Romero study, this study first use of the Laplace
et al. [2] engaged in the ladder filter of the transform algorithm, followed by its digital
test, Gordon et al. [3] the reduction in filter technology, making it the digital filter and
design and analysis of their response, use the calculation (computation) to perform
Alarcon et al. [4] the use of three-pole continuous-time signal filtering action.
Butterworth analog filter for ECG signal of Design frequency-selective filter, is used to
the research. The study of algorithms [5], convert the continuous-time signal
first to use the Laplace transform, from the corresponding to the series. Digital filter in a
regularization of the classical analog low- series of sampling points on the basis of
pass filter. For the pre-distorting the cutoff processing sequence to generate new series,

22
new series followed by a digital to analog (D regarded as the prototype filter and the
/ A) converter, conversion to the system functions fairly easy to obtain.
corresponding continuous time signal. Furthermore, to design high-pass or band-
Finally, the output of the system to produce a pass or band stop filter, frequency conversion
continuous time signal reconstruction filter, can be used easily to obtain. Frequency
on behalf of the original input signal is conversion can be done in two ways. In the
filtered results. In this study using Impulse analog frequency conversion, the analog
invariance method and Bilinear prototype filter H P ( s ) into a desired system
transformation method to deal with the function of the digital filter of another analog
digital signal transfer, change of continuous system function H ( s ) . Then use the
signals into discrete Z-transform system. corresponding skills, into a system
function H ( z ) of the digital filter. In the
II. MAIN CONTAINS digital frequency conversion, the analog
Signal processing algorithms often occur in prototype filters into a first digital domain,
a series of complex operations can easily with the system function H P ( z ) . Then use of
make the calculation error, re-verification frequency conversion, so that can be
must spend a lot of time in order to calculate converted into the digital filter.
the convenience and accuracy of using
MATLAB to write program makes saving a 2-2 Analog frequency transformation
lot of time on operations, and therefore Our
main areas of this monograph is often used in Analog frequency conversion formula used
engineering to math up development, such as; in a prototype low-pass filter into a low-pass
linear algebra, polynomial solving, (with a different cut-off frequency), high
probability calculations ... etc., developed the pass, band pass or band stop filter, as shown
program features allow users more below.
convenient operation. Elliptic filter is
sometimes called Cauer filter. This filter has (i) low-pass filter cut-off frequency C to a
the same ripple for band pass and band stop. new low pass filters cut-off frequency C * :
Discussed in the previous filter in the known
c
order of the filter, band pass and stop with s 
error, Elliptic filter with minimal transient C*
bandwidth. The size of the square of the (2)
response is Therefore, if the prototype filter response of
1 the system is H P ( s) , the low-pass filter of the
H( j)
2
=
1   2 U N ( / c ) system response will
(1)  c 
H(s)  H P  * s 
Where U N ( x ) is the N order elliptic function,  
 C 
while  is the pass band ripple of the (3)
constant. (ii) low-pass filter cut-off frequency C to a
2-1 Frequency transformation new high-pass filters cut-off frequency C * :
C C*
s
There are four basic types of frequency s
selective filters, for low pass, high pass, band (4)
pass and band stop. Discussed in the previous the system function of high-pass filter is
design skills, we only take into account the
low-pass filter. Low-pass filter can be

23
  cC* 
H(s)  H P  
 s
 
Therefore ， to the all frequency ， must
(5) be f (e j )  1 。 So that, the
(iii) low-pass filter cut-off frequency C to correspondence all pass filter can be written
the cut-off frequency  1 and high cut-off as
n z 1   k
frequency of the band pass filter  2 : f (z 1 )    1
k 1 1   z
k
s 2   1 2
s  C (11)
s   1  2 
(6)
the system function of band-pass filter is Table (1) Analog frequency transformation formulas
 s   1 2  2
H(s)  H p  C  Analog Transformation
 s   2  1  

(7) c
(iv) cutoff frequency C for low pass filter Low pass s s
C*
cutoff frequency to the low  1 and high C C*
High pass s
cutoff frequency  2 of the band stop filter: s
s   2  1  Band pass
s  C 
s 2   1 2 s   1 2 
2
s   C 
 s   2  1  
(8) 
the system function of band-stop filter is Band stop
 s    1   s   2  1 
H(s)  H p   C 2 2  s  C
 s   1 2  s 2   1 2
(9)
In order to stabilize the prototype filter to be
2-3 Digital frequency conversion stable filter must set k  1 . For the
conversion prototype low-pass digital filter
As in the analog domain, frequency to the digital low-pass, high-pass, band-pass,
conversion in the digital domain is feasible. or band-stop filter, the conversion formula
In the digital domain into the frequency from (10) obtained, as shown in Table (2).
converter by replacing the function  1 to
achieve, that is, f(  1 ). This corresponds to
TABLE (2) CONVERSION PROTOTYPE LOW-
take into account the stability criteria. All
PASS DIGITAL FILTER TO THE DIGITAL LOW-
poles must fall within the unit circle
PASS, HIGH-PASS, BAND-PASS, OR BAND-STOP
corresponds to its own, and also corresponds
FILTER
to the unit circle itself. To achieve the
corresponding unit circle to itself, for r = 1, Z Types Transformat Design
= r * e jw . ion parameters

 j  j  j jarg[f (e  j )] Low- z1  a a

sin[(c  c* ) / 2]
e  f (e )  f (e )e z1  sin[(c  c* ) / 2]
pass 1 az1
(10)
High- z1 
z1  a a
cos[(c  c* ) / 2]
cos[(c  c* ) / 2]
pass 1 az1
Analog frequency transformation formulas as
shown in Table (1)

24
 Band- z2 az1 a
z1  2 1 1 2
a1  2 K /( K  1) Magnitude Response

pass az
2 az
1 1 a 2  ( K  1) /( K  1) 1

cos[(2 1)/2]
0.8913


cos[(c 1)/2]
      
K  cot  2 1  tan 2 

|H|
 2  2
Band- z2 az1 a a1  2 /( K  1)
z1  2 1 1 2
stop a2z az1 1 a2  (1 K)/(1 K)
0.1585

cos[(2  1 ) / 2]
a
cos[(2  1 ) / 2]
0
0 0.2 0.3 0.5
Analog frequency in pi units
      
K  tan 2 1  tan c  Magnitude in dB
 2  2 0
1

decibels
Frequency conversion can be done either two 16

available techniques, however, important to

note that what techniques to use. Move out,
the pulse invariant transformation is not 30
0 0.2 0.3 0.5
suitable for high frequency high-pass Analog frequency in pi units

resonant or band-pass filter. In this case, 1

Phase Response

assuming a low-pass prototype filter which

uses the analog frequency conversion into 0.5

high-pass filter, and then using impulse-

invariance techniques into the digital filter.
radians

This will cause aliasing problems. However,

if the same low-pass filter prototype, the first −0.5

use of pulse techniques to convert non-

variant digital filter, then after using the −1
0 0.2 0.3 0.5

digital frequency conversion into high-pass Analog frequency in pi units

filter, phase will not have any aliasing

Impulse Response

0.25
problem. Regardless of when to use the
bilinear transformation, whether using analog 0.2

or digital frequency conversion frequency 0.15

converter is not important. In this case,

ha(t)

0.1

analog and digital frequency conversion 0.05

techniques will get the same results.
0

−0.05

III. MATLAB EXAMPLES 0 5 10 15 20

time in seconds
25 30 35

Example 1: In accordance with the following Figure 1. The relationship for analog elliptical low-pass filter
specifications makes analog elliptical low- frequency to amplitude and phase

pass filter design. Wp = 0.2π; Ws = 0.3π;

Rp = 1dB; As = 16dB; From above Example 1 we can find out that
analog elliptical low-pass filter at the confine
specifications Wp = 0.2π; Ws = 0.3π; Rp =
1dB; As = 16dB will produce a small ripples
at the low pass region and obtain a clearly

25
slope for the next region, it is a good 40
Group Delay

transform results for a low pass elliptic filter.

35

Example 2: In accordance with the following 30

specifications makes digital elliptical low- 25

pass filter design. Wp = [0.4π 0.6π]; Ws =

samples
20

[0.3π 0.75π] ; Rp = 1dB; As = 40dB; 15

Magnitude Response
1
10
0.8913

0
0 0.3 0.4 0.6 0.75 1
frequency in pi units

Figure 2. The relationship for digital elliptical low-pass filter

frequency to amplitude and phase

0
0 0.3 0.4 0.6 0.75 1
From above Example 2 we can find out that
frequency in pi units
analog elliptical band-pass filter at the
confine specifications Wp = [0.4π 0.6π]; Ws
Phase Response
1

= [0.3π 0.75π] ; Rp = 1dB; As = 40dB will

0.8

0.6

0.4 produce a small ripples at the band pass top

region and obtain a clearly slope for the other
phase in pi units

0.2

regions, it is a also good transform results for

0

−0.2

−0.4 a band pass elliptic filter.

−0.6

Example 3: In accordance with the following

−0.8

−1

specifications makes digital elliptical low-

0 0.3 0.4 0.6 0.75 1
frequency in pi units

0
Magnitude in dB
pass filter design. Wp = 0.3; Ws = 0.4; Rp =
1dB; As = 20dB;
ellip filter frequency response

0
|magnitude|:dB

−10
−20
40 −30
−40
0 0.3 0.4 0.6 0.75 1
frequency in pi units −50
0 0.2 0.4 0.6 0.8 1
freq (rad/sec)

10

5
phase

−5

−10
0 0.2 0.4 0.6 0.8 1
freq (rad/sec)

26
Figure 3-a. The relationship for digital elliptical low-pass filter pole zero location

frequency to amplitude and phase 1

pole zero location 0.8

1 0.6
0.8
0.4
0.6

Imaginary Part
0.2
0.4
0
Imaginary Part

0.2
−0.2
0
−0.4
−0.2
−0.6
−0.4

−0.6 −0.8

−0.8 −1

−1 −1 −0.5 0 0.5 1
Real Part
−1 −0.5 0 0.5 1
Real Part
Figure 4-b. Digital elliptical high-pass filter pole and zero points
Figure 3-b. Digital elliptical low-pass filter pole and zero points map
map
From above Example 4 we can find out that
From above Example 3 we can find out that digital elliptical high-pass filter at the confine
digital elliptical low-pass filter at the confine specifications Wp = 0.4; Ws = 0.3 ; Rp =
specifications Wp = 0.3; Ws = 0.4; Rp = 1dB; As = 20dB will produce a small ripples
1dB; As = 20dB will produce a small ripples at the high pass region and obtain a clearly
at the low pass region and obtain a clearly slope for the next region, it is a good
slope for the next region, it is a good transform results for a high pass elliptic filter.
transform results for a low pass elliptic filter. On the other hand, from Figure 4-b has
On the other hand, from Figure 3-b has shown that all the pole and zero points are
shown that all the pole and zero points are convergence at the unit circle.
convergence at the unit circle.
Example 5: In accordance with the following
Example 4: In accordance with the following specifications makes digital elliptical Band-
specifications makes digital elliptical high- pass filter design. Wp = [0.3 0.5]; Ws = [0.2
pass filter design. Wp = 0.4; Ws = 0.3 ; Rp = 0.55] ; Rp = 1dB; As = 20dB;
1dB; As = 20dB; ellip filter frequency response
ellip filter frequency response 0
|magnitude|:dB

0 −10
|magnitude|:dB

−10
−20
−20
−30
−30
−40
−40
−50
−50 0 0.2 0.4 0.6 0.8 1
0 0.2 0.4 0.6 0.8 1 freq (rad/sec)
freq (rad/sec)
10
10

5 5
phase
phase

0 0

−5 −5

−10 −10
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
freq (rad/sec) freq (rad/sec)

Figure 4-a. The relationship for digital elliptical high-pass filter Figure 5-a. The relationship for digital
frequency to amplitude and phase
elliptical band-pass filter frequency to
amplitude and phase

27
 pole zero location pole zero location

1 1

0.8 0.8

0.6 0.6

0.4 0.4
Imaginary Part

Imaginary Part
0.2 0.2

0 0

−0.2 −0.2

−0.4 −0.4

−0.6 −0.6

−0.8 −0.8

−1 −1

−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1

Real Part Real Part

Figure 5-b. Digital elliptical band-pass filter Figure 6-b. Digital elliptical band-stop filter
pole and zero points map pole and zero points map

From above Example 5 we can find out that From above Example 6 we can find out that
digital elliptical band-pass filter at the digital elliptical band-stop filter at the
confine specifications Wp = [0.3 0.5]; Ws = confine specifications produce a good
[0.2 0.55] ; Rp = 1dB; As = 20dB will transform results for elliptic filter. The most
produce a small ripples at the band pass important thing is finding that the bilinear
region and obtain a clearly slope for the next transformation method is better than the
region, it is a good transform results for a Impulse-invariance method for the high pass
band pass elliptic filter. On the other hand, filter status.
from Figure 4-b was shown that all of the
pole and zero points are convergence at the Example 7. Impulse-invariance method and
unit circle. bilinear transformation method will be
converted analog elliptical low-pass filter to
Example 6: In accordance with the following high-pass digital filter. Filter specification
specifications makes digital elliptical Band- are
stop filter design. Wp = [0.2 0.55]; Ws =
[0.3 0.5] ; Rp = 1dB; As = 20dB; pL  0.2, pH  0.2, R p  1dB
ellip filter frequency response
s  0.3, A s  15dB
0
pL :Low-pass digital frequency
|magnitude|:dB

−10
−20 pH : High-pass digital frequency
−30
−40
Rp : Ripple of band-pass
−50
0 0.2 0.4 0.6 0.8 1 s : Band-stop digital frequency
freq (rad/sec)
As : Band-stop digital frequency decay
10

5
This article by the above theoretical analysis,
phase

0
first create a simulation system by Laplace
−5
transform method to expand as part of the
−10
0 0.2 0.4 0.6 0.8 1 fractions, and using MATLAB software to
analyze the selected time interval of its
freq (rad/sec)

Figure 6-a. The relationship for digital calculation and the results plotted graphically
elliptical band-stop filter frequency to in Figure 1 . Impulse-invariant method and
amplitude and phase Bilinear transformation method will be
converted to analog elliptical low-pass filter

28
and high-pass digital filter. Wp = [0.2 0.55]; Ws = [0.3 0.5] ; Rp =
1dB; As = 20dB; will produce a small
1
Lowpass Filter Magnitude Response
ripples beside the band stop regions and
0.8913
obtain a clearly slope for the next region, it is
a good transform results for a band stop
elliptic filter.
|H|

IV. RESULTS AND DISCUSSIONS

0
0 0.2 1
This article by the above theoretical
. frequency in pi units
analysis, first by the analog frequency
0
1
Lowpass Filter Magnitude in dB
conversion from the regularization of the
classical analog low-pass filter transfer
function H a (s) start, pre-distorting the cut-off
frequency, and then apply the bilinear
decibels

transformation equation to produce the

equivalent digital filter H(z). For the
frequency conversion of analog design with a
30
diameter-degree low-pass cut-off frequency
in the elliptic filter is normalized elliptic low-
0 0.2 1
frequency in pi units

1
Highpass Filter Magnitude Response
pass filter 0 from the start, and with
0.8913
replacement of s / 0 . Using the same method
can be normalized low-pass filter, converted
to other types of frequency selective filters,
|H|

such as high-pass, band-pass, and band-stop.

This series of computer simulation using
MATLAB software for analysis of related
0
0 0.6 1
functions and a graphical plot of the
frequency in pi units

Highpass Filter Magnitude in dB

conversion result, as shown in Figure 1-6.
0
1 Finally, there is the discrimination, by (a) the
same impulse-invariance method (b) Bilinear
transformation method, Laplace equation can
be converted to the Z-transform and discrete
decibels

systems using MATLAB software to analyse

the selected time interval of its calculation
and the results plotted graphically in Figure 7,
30
0 0.6 1
the resulting two kinds of digital signal
frequency in pi units
conversion of a graphic comparison of the
Figure 7 Using Impulse-invariance results are very similar. This calculation by
transformation method and bilinear elliptic the simulation system using MATLAB
low-pass filter method to high-pass filter for software for analysis of related functions and
the frequency and amplitude relationship the results plotted graphically, and to get
good simulation results. Found in Figure 3 of
From above Example 7 we can find out that the high-pass digital filter of the amplitude of
digital elliptical filter using Impulse- the Bilinear transformation by Impulse-
invariance transformation method and Invariance produce more amplitude than
bilinear elliptic low-pass filter method to higher-up about 25%.
high-pass filter at the confine specifications

29
 V. CONCLUSIONS [7] Roberts, G. W.; Sedra, A. S.: All current-mode frequency
selective circuits. Electron. Lett. 25 (1989), 759–761.
[8] Wilson, B.: Recent developments in current conveyor and
(1) This analog elliptic filter system, through the establishment of current-mode circuits. Proc. IEE, Part-G 137 (2) (1990), 63–77.
model. Analog signals with the proposed digital signal equation, [9] Fabre, A.; Saaid, O.; Wiest, F.; Boucheron, C.: High frequency
obtained using MATLAB analysis and calculation of the order of applications based on a new current controlled conveyor. IEEE
the impulse response of different situations, and once again Trans. Circuits and Systems-I 43 (1996), 82–91.
confirmed that the number of higher order than those of their [10] Orchard, H. J.: Inductorless filters. Electron. Lett. 2 (1966),
turning point clear. 224–225.
[11] Tan, M. A.; Schaumann, R.: Simulating general-parameter LC-
(2) for discrete processing and digital filter system to get, by (1) ladder filters for monolithic realizations with only
Pulse constant conversion method (2) Bilinear transformation transconductance elements and grounded capacitors. IEEE
method and other two conversion methods, Laplace equation can be Trans. Circuits and Systems 36 (1989), 299–307.
converted to use the Z- conversion of discrete digital system, and [12] Hwang, Y. S.; Liu, S. I.;Wu, D. S.;Wu, Y. P.: Table-based
selected using MATLAB software to analyze its time interval is linear transformation filters using OTA–C techniques. Electron.
calculated and obtained good simulation results, and bilinear Lett. 30 (1994), 2021–2022.
transformation method to get the same conversion than the pulse [13] Tan, M. A.; Schaumann, R.: A reduction in the number of
method has a better conversion results. active components used in transconductance grounded
capacitor filters. Proc. of IEEE Int. Symp. Circuits and Systems,
REFERENCES (1990). 2276–2278.
[14] Hwang, Y. S.; Chiu, W.; Liu, S. I.; Wu, D. S.; Wu, Y. P.:High-
frequency linear transformation elliptic filters employing
[1] Hong-Guang Ma, Xiao-Fei Zhu, Jian-Feng Xu and Ming-Shun minimum number of OTAs. Electron. Lett. 31 (1995), 1562–
Ai, Circuit state analysis using chaotic signal excitation, Journal 1564.
of the Franklin Institute, In Press, Corrected Proof, Available
online 5 July 2007
[2] Eduardo Romero, Gabriela Peretti, Gloria Huertas and Diego
Vázquez, Test of switched-capacitor ladder filters using OBT,
Microelectronics Journal, Volume 36, Issue 12, December 2005,
Pages 1073-1079
[3] D. Gordon E. Robertson and James J. Dowling, Design and
responses of Butterworth and critically damped digital filters,
Journal of Electromyography and Kinesiology, Volume 13,
Issue 6, December 2003, Pages 569-573
[4] G. Alarcon, C. N. Guy and C. D. Binnie, A simple algorithm for
a digital three-pole Butterworth filter of arbitrary cut-off
frequency: application to digital electroencephalography,
Journal of Neuroscience Methods, Volume 104, Issue 1, 15
December 2000, Pages 35-44
[5] Signals and Systems, Hongwei Yao, M., Chian-Hua Book
Company(2005)
[6] Introduction to MATLAB 7 for engineering, William J. Palm III,
Gao Li Book Company, ISBN:986-157-204-X,(2005)

30