Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
This paper is submitted for review on 10 June 2016. This research was of School of Communications and Informatics, Victoria University of
independently undertaken by P. S. Nguyen, who was previously a staff Technology, Melbourne, Australia (Email:
member at the Faculty of Electrical and Electronic Engineering, Department nguyensiphuoc@frequencytransform.com).
The polezero transfer function of the nth-order analog it can be seen that the numerator Nk(s) and denominator Dk(s)
Butterworth low pass prototype stable filter is found as indicate all the stable and unstable poles sk move along the
follows: real axis by a value of c. This means that using bilinear z-
- if n is even number, transformation with pre-warping frequency to transform an
(2) analog Butterworth low pass prototype to a digital low pass
1
H s n/ 2
; simply moves all of the poles on the unit circle from the s-
(s 2
2 s cos k 1) domain into the z-domain by pre-warping parameter c along
k 0
the real axis, as shown in Fig.2.
- if n is odd number,
(3)
1
H s ( n 1)/ 2
.
s 1 ( s 2 2 s cos k 1)
k 0
From (2) and (3), multiplying out all the factorized terms
in the denominator will give the coefficient transfer function
as follows:
1 1 (4)
H s
n
B0 B1 s Bn s n
B
k 0
k sk
. g
2 2 2 n
1
DB BO DO
c cos
2
sin 2 k
2
k 0
k
- If n is an odd number:
1
g n 1
1
c 1 c cos k
2
sin 2 k
2
k 0
The analog Butterworth low pass prototype filter is known
as an all-pole filter as described in (3) and (4). The polezero
transfer function of the digital Butterworth low pass filter as
described in (10) has a term (z+1)n, which arises from the term
(z+1) in the denominator of (5). The (z+1)n terms are known
as the zeros of the digital Butterworth low pass filter, and they
are located at z=1 in the z-domain.
(unstable pole), k 0
z 2cos k zk z zk
2
k 0
2
(10) left, diagonal and above element:
k 0
H ( z ) go n(1z 1) ( z 1)n
n
g P P P P .
k 1 ; k 1 k 1 1; k 1 k 1 ; k k 1 1; k
n1
if n is odd
th th th th
z zk
2
z z 2 2cos k zk z zk
2
k 0
k 0 Equation (11) is used to generate the coefficients of the nth-
where g is a system gain. order digital Butterworth low pass filter; it is called the
- If n is an even number: analog Butterworth low pass to digital Butterworth low pass
Pascal matrix equation.
n n!
ak g k g k ! n k !
(14)
D. Special case when fc is one-quarter of fs
This section will analyze the effect of the cut-off frequency bk g Bk P
fc on the pole locations of the digital Butterworth low pass
filter. The coordinator form of zk in (8) shows that depending
on angle c, the poles may be located on the right or left side IV. DESIGN OF A DIGITAL BUTTERWORTH LOW PASS FILTER
of the unit circle in the z-domain.
As discussed in Section III, three parameters are needed to
f cos c 0 design a transfer function for a digital Butterworth low pass
0 fc s 0 c right hand side
4 2 sin 0 filter: the cut-off frequency fc, the sampling frequency fs, and
c
the nth-order of the filter. There are two kinds of transfer
cos c 0
fs f function: the polezero transfer function is used to determine
fc s c left hand side .
sin c 0
4 2 2 stability of the filter, and the coefficient transfer function is
In this case the cut-off frequency fc the nth-order digital used to derive the finite difference equation for the realization
Butterworth low pass filter is equal to one-quarter of the structure of the designed filter (e.g. direct form IIR digital
sampling frequency fs: filter structure). A block diagram such as that in Fig. 5
outlines the algorithm for design of an nth-order digital
Butterworth low pass filter from an analog Butterworth low
c cot 1
4
f pass prototype filter.
fc s c R tan c .
4 2
1
F ,0 F ,0
cos
c
The poles zk in (7) are described as follows:
z k i cot k . (12)
2
z cot 2
2
2 k 0
H z k 0
(13) Apply (9) and (10) to find the system gain g and poles zk:
z 1
n
1
if n is odd 1
n1 1 n 1
1 g 0.5276
2 k 2 k
n 1
n 2 2 2 1
2 sin 2 z z cot 2
c 1 c cos k
2
sin k
2 2
k 0 k 0
k 0
- The coefficient transfer function:
z0 0.73872.58 ; z1 0.50953.1416 ; z0 0.7387 2.58 .
|zk|<1 => all poles inside the unit circle in the z-domain
(stable), as shown in Fig. 6.
determine where the poles are distributed in the z-domain and
it also indicates the degree of stability of the filter. The
coefficient transfer function is used to generate the
coefficients of the filter and is described by a general
equation, called the analog Butterworth low pass to digital
Butterworth low pass Pascal matrix equation. Both the pole
zero and coefficient transfer functions are easy to use in hand
calculations and can be implemented by programing using
Matlab, C, C++, or assembly program languages for digital
signal processors. With the involvement of the Pascals
triangle, matrices [P] and [T] are easy to find. Given an order
n of the low pass filter, if the pre-warping parameter c is
changed, only matrix [T] is changed, which indicates that the
analog low pass to digital low pass Pascal matrix equation
Fig. 6. The pole locations of the 3rd-order digital Butterworth low pass might be used to design a tunable digital low pass filter.
filter.
Determine [Bk], [P] and [T], from H(s) => [Bk] = [1 2 2 1]: ACKNOWLEDGMENT
1 0 0 0 1 3 3 1
0 ; P 1 1 1 1
T 00 0.3249
0
0
0.1056 0 1 1 1 1
This paper describes independent research by the author,
0 0 0 0.0343 1 3 3 1 who would like to share his ideas on how to apply Pascals
triangle in filter design to others with the same interests. He
Applying (11), we find the coefficients ak and bk: would greatly appreciate all feedback from readers to
ak 0.5276 1 3 3 1 improve this research!
b g B T P 1 1.7600 1.1829 0.2781
k k
The coefficient transfer function H(z) of the 3rd-order digital
Butterworth low pass filter is: REFERENCES
0.5276 1.5829z 1 1.5829z 2 0.5276z 3 N. S. Phuoc, Bilinear z-transformation using Pascals triangle,
H z
[1]
1 1.7600z 1 1.1829z 2 + 0.2781z 3 International Journal of Scientific and Engineering Research, vol. 4,
pp. 19741978, 2013.
The amplitude frequency response of the 3rd-order digital [2] S. Chivapreecha, Bilinear sz frequency transformation using matrix
Butterworth low pass filter is described in Fig.7. Pascal operation, IEEE Communications and Information
Technology, vol. 1, pp. 764767, 2005.
[3] C. Emmanuel, B. Ifeacor, and W. Jervis, Digital Signal Processing,
Addison-Wesley, Workingham, England, 1993, ch. 7, pp. 392394.
[4] N. S. Phuoc, Frequency transformation in digital domain, presented
at the International Journal of Signal Processing System, Taiwan,
August 2324, 2015.
[5] N. S. Phuoc, Frequency transformation with Pascal matrix equation,
presented at the 18th International Conference on Communications and
Information Technology, Dubai, January 2829, 2016.
[6] C. Subhadeep, K. J. Krishana, and P. Abhiup, Design of IIR digital
high pass Butterworth using analog to digital mapping technique,
International Journal of Computer Applications, vol. 52, no. 7, pp.06-
10, August 2012.
[7] B. A. Shenoi, Introduction to Digital Signal Processing and Filter
Fig. 7. The amplitude frequency response of the 3rd-order digital Design, John Wiley & Sons, Hoboken, New Jersey, USA, 2005, ch. 4,
Butterworth low pass filter with fc=400Hz and fs=1KHz. pp. 212218.