Frequency Transformation with Pascal Matrix

Equations
Phuoc Si Nguyen

The transfer function of a given and desired analogue filter

Abstract Frequency transformation with Pascal matrix can be written as:
equations is a method for transforming an electronic filter (analogue or 1
digital) into another filter. The technique is based on frequency s
Ag Ag ... Ag s 2
transformation in the s-domain, bilinear z-transform with pre-warping 0 1 n
Ag i s i
n

frequency, inverse bilinear transformation and a very useful n

s Ag S
application of the Pascals triangle that simplifies computing and H g ( s ) i n0 (1)
1 Bg S
enables calculation by hand when transforming from one filter to Bg i s i
s
Bg Bg ... Bg s 2
i 0
another. This paper will introduce two methods to transform a filter n
0 1
into a digital filter: frequency transformation from the s-domain into n
the z-domain; and frequency transformation in the z-domain. Further, s
1
two Pascal matrix equations are derived: an analogue to digital filter s
Pascal matrix equation and a digital to digital filter Pascal matrix Ad 0 Ad 1 ... Ad N s 2
N
equation. These are used to design a desired digital filter from a given
Ad i s i N
filter. H d ( s ) i N 0 s Ad S (2)
1 Bd S
Keywords Frequency transformation, Bilinear z-transformation, Bd i s i
s
Pre-warping frequency, Digital filters, Analog filters, Pascals triangle. i 0
Bd 0 Bd 1 ... Bd N s 2

N
I. INTRODUCTION s
The transfer function of a given and desired digital filter can
C URRENTLY in filter design, a low pass prototype filter (i.e.
a low pass, high pass, band pass or band stop filter) is used
to convert between types of analogue or digital filters using
be written as:
1
z 1
frequency transformation and bilinear z-transformation with ag ag ... ag z 2
0 1 n
a
n
i
pre-warping frequency techniques [2]. With the support of the g z n
H g (z) i 0
i
z ag Z
Pascals triangle and two techniques mentioned, this paper will 1 bg Z
(3)
b
n
zi z 1
introduce a new method to transform between an analogue low g i
bg bg ... bg z 2
i 0

pass prototype or a digital low pass, and a digital filter in a 0 1 n

n
general mathematical way described as Pascal matrix z
equations. There are two kinds of Pascal matrix equations: the 1
z 1
analogue low pass prototype to digital filter matrix equation, ad 0 ad 1 ... ad N z 2
N
which is used for frequency transformation from the s-domain a zi N
d
to the z-domain; and the digital low pass to digital filter matrix H d (z) i 0
i
z ad Z (4)
N
1 bd Z
equation, which is used for frequency transformation in the z- b d i
zi z 1
bd 0 bd 1 ... bd N z 2
i 0
domain. The main purpose of this new method is to provide a

new way to design a digital filter with Pascal matrix equations, N
z
which facilitates easy hand calculation and programming.
In the next sections, the relationship between the coefficients
matrix A and B of a given analogue low pass prototype and
II. TRANSFER FUNCTION
coefficients matrix a and b of a desired filter is demonstrated
A transfer function is a mathematical way of describing the using frequency transformation and bilinear z-transformation
ratio between output and input of a system, and can be written with pre-warping frequency.
in matrix form for analogue and digital filters, as shown in
(1),(2),(3) and (4) below. Let A and B represent the coefficient III. BILINEAR Z-TRANSFORMATION WITH PRE-WARPING
matrices in the numerator and denominator of an analogue FREQUENCY
filter: a and b are used similarly for a digital filter. The subscript
Bilinear z-transformation is a popular method to transform
g denotes given and d denotes desired.
an analogue filter to a digital filter [1] which preserves the

P.S Nguyen was a staff at the Faculty of Electrical and Electronic Victoria University of Technology, Melbourne, Australia (Email:
Engineering, Department of School of Communications and Informatics, nguyensiphuoc@frequencytransform.com).
frequency characteristics and it is defined as shown in (5). This TABLE I
technique involves one-to-one mapping from the s-plane onto BILINEAR Z-TRANSFORMATION WITH PRE-WARPING FREQUENCY
z-plane, such as the imaginary axis j (s=j) is mapped into the Converting Analogue low pass
s = f(z)
types prototype H(s)
unit circle |z|=1, the left half-plane s (s = +j) is mapped onto
Low pass to H(f(z)) 1 z 1
the interior of the unit circle of the z-plane as shown in Fig.1 c
low pass 1 z 1
below Low pass to H(f(z)) 1 z 1
2 1 z 1 high pass
t
1 z 1
s (5)
T 1 z 1 Low pass to H(f(z)) 1 z 1 1 z 1
U L
band pass 1 z 1 1 z 1

j j 1 Low pass to H(f(z)) 1

band stop 1 z 1 1 z 1
U L
1 z 1 1 z 1

s=+j
-1 1 fU cU
c cot U
Re f f U fs 1 cU t L
Re c cot c , t tan c ,
fs t tan fL tL
fs L
1 cU t L

L
fs

-1 The transformation s=f(z) from an analogue low pass

s-plane z-plane prototype H(s) must satisfy the requirement the digital transfer
Fig. 1 Mapping from s-plane onto z-plane function H(f(z)) be stable. From Table. I, the inverse bilinear z-
transformation can be founded as below
The main advantage of this method is its ability to transform cs
z (9)
a stable designed analogue filter to a stable digital filter. c s
However, the mapping of the frequency in s-plane into
frequency in z-plane is a non-linear relationship, called Sub s = + j into (9),
warping frequency as show in (6). c j c 2
2

T z z
2
tan (6) c j c 2
2

T 2
Let 0 is a cut-off frequency of a desired digital filter, from If = 0, |z| = 1 and < 0, |z| < 1, the left haft plane in s-
(6) this frequency is tangentially warped compared with the cut- domain maps into the inside of the unit circle. Hence the
off frequency 0 of a designed analogue filter and this is an bilinear z-transformation with pre-warping frequency is a stable
undesirable effect when transforming from s-domain into z- transformation.
domain. To overcome this effect, one method called pre-
warping frequency as expressed in (7) From (1), (2), (3) and (4) and Table I, the relationship
0 1 z 1 between the coefficients of a given and a desired filter can be
s (7) described as a matrix equation, which is a multiplication of
T 1 z 1
tan 0
2 some specified matrices introduced in the next section.
In the next section will introduce one new method to
transform an analogue filter into a digital filter called bilinear V. DEFINITION OF SOME SPECIFIED MATRICES
z-transformation with pre-warping frequency. A. Matrix T and Dc
Two kinds of matrix T are matrix TLP and TUL. Matrix TLP
IV. BILINEAR Z- TRANSFORM WITH PRE-WARPING FREQUENCY
and Dc are a diagonal matrix with a size of (n+1;n+1), where n
Bilinear z-transformation with pre-warping frequency is a is the given nth-ordered analogue low pass prototype as
method to convert an analogue low pass prototype with a corner illustrated below.
angular frequency at 1 rad/s to a digital filter [5], as shown in
Table I: fs is the sampling frequency, c is the corner angular 1 0 0 0 0 cn 0 0 0 0
0 c 0 0 0 0 c n 1 0 0 0
frequency of the desired low pass (Lp) and high pass (Hp), and TLP 0 0 c2 0 0 Dc 0 0 c n 2
0 0
U and L are the upper and lower angular frequency of the
0 0 0 0 0 0 0 0
desired band pass (Bp) and band stop filter (Bs). In the case of 0 0 0 0 cn 0 0 0 0 1
a narrow band filter, if f0 is a centre frequency and Q is a quality A neat application of the Pascals triangle is in the
factor, the upper frequency (fU) and the lower frequency (fL) of expansion of a binomial expression (U+L)n. Inserting zeros into
the narrow band filter can be found by (5), and the coefficients the Pascals triangle makes a matrix TUL with a size of (n+1;
c, t, U and L can be calculated as follows: 2n+1) as shown below
1 1 0 0 0 0 0 1 0 0 0 0 0
f L f0
1
4Q 2 2Q

0 0 0 0 U 0 L 0 0 0 0
(8) U2 L2
f f 1 TUL 0 0 0 0 2UL 0 0 0 0
0 1
1



U
4Q 2
2Q
0 0 U3 0 3U 2 L 0 3UL2 0 L3 0 0
0 0 0 0 0 0
Un 0 0 0 0 0 Ln
 B. Matrix P and Ptr Form Table II and the inverse bilinear z-transformation, a
The matrix P contains the positive and negative binomial digital low pass can be transformed into another digital, as
coefficients of the Pascals triangle in the first, last row and the f
first, last column corresponding to the edge size and the nth row shown in Table 2, where c N cot CN , and fCN is the new
fs
of the Pascals triangle; another element in the matrix P can be
calculated from its left, diagonal and above element. There are corner frequency of a digital low pass filter.
two different matrices for matrix P: PLP for a low pass filter with TABLE II
size (n+1; n+1), and PHBS for high pass with size (n+1; n+1); the FREQUENCY TRANSFORMATION IN DIGITAL DOMAIN
band pass and band stop has a size of (2n+2; 2n+1). The Digital
Converting low pass
matrices can be defined as: types prototype
Z(z-1)

H(z)
PLP i1; j1n1 1 PHBS i1; j1N 1 1 Low pass to H(Z(z-1)) c cN (c cN ) z 1

PLP in1; j1n1 1 PHBS i N 1; j1N 1 1
j 1 j 1 low pass c cN (c cN ) z 1

Low pass to H(Z(z-1)) c t ( c t ) z 1
PLP i1n1; j1 i 1 PHBS i1N 1; j1 1 i 1
n i 1 N
PLP PHBS high pass c t ( c t ) z 1



PLP i1n1; jn1 1 i 1
i 1 n





PHBS i1N 1; j N 1 i 1
N

Low pass to
band pass
H(Z(z-1)) c U L 2(U L) z 1 (U L c ) z 2
c U L 2(U L) z 1 (U L c ) z 2
P P P
LP i 1; j 1 PLP i 1; j
P P P
HBS i 1; j 1 PHBS i 1; j
Low pass to H(Z(z-1)) cU cL 1 2c (U L) z 1 ( cU cL 1) z 2

LP i ; j LP i ; j 1 HBS i ; j HBS i ; j 1

band stop cU cL 1 2c (U L) z 1 cU cL 1)z 2

Matrix Ptr is a transpose matrix of the matrix P and it is used

in the case of transforming a digital filter to another digital
filter. From Table II, transforming a digital low pass with the
transfer function H(z) to a digital low pass, high pass, band pass
VI. FREQUENCY TRANSFORMATION WITH PASCAL MATRIX or band stop requires just the replacement of the variable z-1 by
EQUATIONS a function (z-1). With the supporting of the Pascals triangle in
The frequency transformation with Pascal matrix equations binomial expansion, the relationship between coefficients ag, bg
may be used in the analogue or digital domain by of the given digital low pass and coefficients ad, bd of the
mathematically mapping a designed filter onto another filter desired digital filter can expressed as
with the desired specifications. This section will introduce two
ad ag PLP DcTP
tr tr
algorithms for transformation from a given filter to a digital b b P tr D TP tr (11)
filter: frequency transformation from the s-domain into the z- d g LP c

domain and frequency transformation in the z-domain.

Depending on the desired digital filter, Ptr and T can be PLP
A. Frequency transformation from s-domain into z-domain and TLP for low pass, PHBS for high pass, band pass, band stop
Frequency transformation from the s-domain into the z- and TUL for band pass, band stop and narrow band filter.
domain is a method to convert an analogue low pass prototype
to a digital filter [3], [4]. From Table I and (1) and (3), an VII. DESIGN OF A DIGITAL FILER USING PASCAL MATRIX
analogue low pass prototype is converted to another analogue EQUATIONS
filter by using frequency transformation in the s-domain, then As discussed in Section VI, a digital filter can be designed
applying bilinear z-transformation with pre-warping frequency from a given analogue low pass prototype or a digital low pass
to convert it to the desired digital filter. The relationship of the filter. A block diagram below illustrates a new algorithm to
coefficients of a given analogue low pass filter and a desired design a digital filter using Pascal matrix equations, which can
digital can be described as in (10), which is termed the analogue be implemented by programming in digital signal processors.
low pass prototype to digital filter Pascal matrix equation:
ad Pd ( AgTd )
b P ( B T ) (10) Analog low pass
d d g d Analog low prototype to digital Digital
pass prototype filter Pascal matrix filter (low
Equation (10) is a general formula to transform a transfer
equation pass, high
function H(s) of an analogue low pass prototype in the s-domain pass,
into a transfer function H(z) of a digital filter in the z-domain. band
pass,
B. Frequency transformation in digital domain Digital Digital low pass to band stop,
Low pass digital filter Pascal narrow
Frequency transformation in the z-domain is a method of pass band)
matrix equation
transforming one digital filter into another. This section will
first introduce a new way to transform a low pass digital to a
low pass, high pass, band pass and band stop digital filter, from Block diagram illustrates design a digital filter using Pascal matrix
which a general equation will be found for applying this to equations
transform a low pass digital filter to another digital filter as
Example 1: Transform a 2nd-order analogue Butterworth low
called Low pass digital to digital filter Pascal matrix equation.
pass prototype to a digital band pass with upper frequency of
200 Hz and lower frequency of 100 Hz, at the sampling
frequency at 1 KHz.
1 Ag 1 0 0
H ( s)
s 2 1.4141s 1 Bg 1 1.4141 1

200 c
c cot 1000 1.3764 U 1 ct 2.4899


t tan 100 0.3249
t
L 0.5878
1000 1 ct

Apply (10)
ad PHBS AgTUL Fig.3 Transforming a digital low pass prototype to digital band

bd PHBS BgTUL
pass using Pascal matrix equations.
1 1 1 1 1
4 2 0 2 4
ad 6 0 2 0 6 VIII. CONCLUSION
4 2 0 2 4
1 1 1 1 1
0 0 1 0 0 Two methods for designing a digital filter based on a given
1 0 0 0 2.4899 0 0.5878 0


6.1996 0 2.9271 0 0.3455 analogue low pass filter or a given digital low pass filter were
1 0 2 0 1 examined. With the support of Pascal matrix equations, the
1 1 1 1 1
analogue low pass prototype to digital filter Pascal matrix
4 2 0 2 4 equation and the digital low pass to digital filter Pascal matrix
bd 6 0 2 0 6
4 2 0 2 4 equation can be implemented by programming using
1 1 1 1 1
0
1 1.4141 0 0
0 1 0 0 MATLAB, C, C++ or assembly program languages for digital
2.4899 0 0.5878 0

6.1996 0 2.9271 0 0.3455 signal processors. Its inherent simplicity could make the
14.8246 -28.7964 31.4164 -18.0364 6.1196 algorithm attractive for many applications demanding
1 2 z z 4 2 minimization of computational requirements and workloads.
H (z)
14.8246 28.7964z 1 31.4164z 2 18.0364z 3 6.1196z 4

ACKNOWLEDGMENT

This is my own research and I would like to share my idea how

to apply Pascals triangle in filters design to whom have the
same interesting. I would greatly appreciate all feedbacks from
readers to improve this research!

REFERENCES
[1] [1] Emmanuel, C., Ifeacor, B. & Jervis, W.: Digital signal processing
Fig. 2 Transforming an analog low pass prototype to digital band (Addison-Wesley, 1993)
pass using Pascal matrix equations. [2] Konopacki J.: The frequency transformation by matrix operation and its
application in IIR filters design, IEEE Signal Processing Letters, 2005,
12, (1), pp. 58
Example 2: Transform the digital low pass H(z) at corner [3] Garcia-Ugalde, F. J.: Z-transform by Pascal matrix and its applications
frequency fc= 50Hz to a digital band pass with upper frequency in the design of IIR Filters, IEEE, 2011, 9, (3), pp. 355366
of 200 Hz and lower frequency of 100 Hz at the sampling [4] Chivapreecha, S.: Bilinear s-z frequency transformation using matrix
Pascal operation, IEEE Communications and Information Technology,
frequency 1KHz. 2005, 1, pp. 764767
[5] Phuoc S. N.: Bilinear z-transformation using Pascals triangle,
1 2 z 1 z 2 ag 1 2 1 International Journal of Scientific and Engineering Research, 2013, 4, pp.
H (z)
49.7925 77.7269z 2 31.9345 bg 49.7925 77.7269 31.9345 19741978.
50
c cot 6.3138
1000
Phuoc Si Nguyen: At first, Phuoc Si Nguyen received a
Apply (11) Bachelor's degree in Mathematics and Physics from the
ad ag PLP DcTUL PHBS
tr tr

b b P tr D T P tr University of Can Tho, Vietnam and experienced teaching

d g LP c UL HBS
Mathematics and Physics for high school students in his
1 2 1 39.8635 0 0 0 0 1 0 0 hometown. A few years later, he received a bachelor's degree in Electrical and
ad 1 2 1 1 0 1 0 6.3138 0 0 2.4899 0 0.5878 0
1 2 1 0 0 1 6.1996 0 2.9271 0 0.3455 Electronic Engineering from Victoria University of Technology in Melbourne,
1 4 6 4 1 Australia. Because of the outstanding graduation , he joined the research group
1 2 0 2 1
1 0 2 0 1 159.4538 1 0 2 0 1 in High Efficiency Power Amplifier for Mobile Communication System and he
1 2 0 2 1 was also a lab assistant and tutor at Department of Communication and
1 4 6 4 1
Informatics, Victoria University of Technology. His interests are in the field of
1 2 1 39.8635 0 0 0 0 1 0 0 mathematics and signal processing techniques. Currently, while Maths tutoring
bd 49.7925 77.7269 31.9345 1 0 1 0 6.3138 0 0 2.4899 0 0.5878 0
for high school students, he enjoys doing self- study on How Pascals triangle
1 2 1 0 0 1 6.1996 0 2.9271 0 0.3455
1 4 6 4 1 can be applied in design of digital filters. This idea shows a lot of promise
1 2 0 2 1 applications which he absorbed in them and hope to share to whom have the
1 0 2 0 1 159.4538 14.8246 -28.7964 31.4164 -18.0364 6.1196
1 2 0 2 1 same interests.
1 4 6 4 1

1 2 z 2 z 4
H (z)
14.8246 28.7964z 1 31.4164z2 18.0364z3 6.1196z 4