Time Discrete Filters

© PFH Private Hochschule Göttingen 1

The z-Transformation
▪ Sampled-data systems can be described in the frequency domain by using the bilateral
z-transform

X (z) =  x (k )  z
n =−
−n

The Fourier-transform is a special case of the z-transform. With the definition

z = e jTs
it is possible to describe the sampled-data system in the frequency domain

X ( j) =  x (k )  e
n =−
− jTs n

Important is the result, if only one time-delay exists



 x ( k − 1)  z
n =−
−n
= z −1  X ( z )
© PFH Private Hochschule Göttingen 2
The transfer function of sampled-data systems
The z-Transformation

▪ The z-transform H(z) of the impulse response h(k) is the transfer function

h(k ) − • H ( z)
▪ Based on the convolution theorem:

Y (z)
H (z) =  h ( k )  z −n
n =−
H ( z) =
X (z)
▪ This is the fundamental relationship between the input- and output signal of a linear,
time discrete system.

▪ The mathematical description of a linear, time discrete system can be given by the
impulse response h(k) or the transfer function H(z) . Both are equivalent!

▪ Note: For many application we can define y(n) = x(n) = 0 for n < 0

© PFH Private Hochschule Göttingen 3

The transfer function of sampled-data systems
Finite difference equation
▪ The most important description form of linear, time discrete systems
is the finite difference equation:
M N
y ( k ) = − ai  y (k − i ) +  bi  x(k − i )
i =1 i =0

▪ The output value y(k) can be calculated by a linear combination of

the last in- and output values and the actual input value weighted by b0
▪ The time-invariant coefficients ai and bi characterize the system
▪ This is a recursive description form
▪ If all ai = 0 (i = 1…M) , we have a non-recursive description form, then
▪ The system order is calculated with: max{M,N}

© PFH Private Hochschule Göttingen 4

The transfer function of sampled-data systems
Finite difference equation

M N
y ( k ) = − ai  y (k − i ) +  bi  x(k − i )
i =1 i =0

▪ With the z- transformation you get:

M N
Y (z ) = − ai  z  Y (z ) +  bi  z −i  X (z )
−i

i =1 i =0

▪ From which the transfer function will be calculated

Y (z ) b0 + b1  z −1 +  + bN  z − N
H (z ) = =
X (z ) 1 + a1  z −1 +  + aM  z − M

© PFH Private Hochschule Göttingen 5

Time discrecte Filters
Basic realization methods for digital filters

Digital filters described by the transfer function H(z) may be generally realized in
the following forms

▪ Direct form I
▪ Direct form II
▪ Cascade
▪ Parallel

© PFH Private Hochschule Göttingen 6

Direct form I realization of a recursive system
Recursive systems are also called IIR-systems (IIR: Infinite Impulse Response)

X ( z) b0 Y ( z)
z −1 z −1
b1 − a1
z −1 z −1 M N
Y (z ) = − ai  z −i  Y (z ) +  bi  z −i  X (z )
− a2 i =1 i =0
b2

Yz ( z ) b0 + b1  z −1 + + bM  z − M
H ( z) = =
− aN −1 X z ( z ) 1 + a1  z −1 + + aN  z − N
bM −1
−
z −1 z −1
bM aN
© PFH Private Hochschule Göttingen 7
Direct form II realisation of a recursive system
Recursive systems are also called IIR-systems (IIR: Infinite Impulse Response)
X ( z) Y ( z)
b0
z −1 z −1
b1 − a1 M N
Y (z ) = − ai  z −i  Y (z ) +  bi  z −i  X (z )
z −1 z −1 i =1 i =0

b2 − a2 Yz ( z ) b0 + b1  z −1 + + bM  z − M
H ( z) = =
X z ( z ) 1 + a1  z −1 + + aN  z − N

bM −1 − aN −1
−
z −1 z −1
bM aN
© PFH Private Hochschule Göttingen 8
Direct form II realization of a recursive system
Recursive systems are also called IIR-systems (IIR: Infinite Impulse Response)

b0
X ( z) Y ( z)
−1
z
b1 − a1
M N

z −1 Y (z ) = − ai  z −i  Y (z ) +  bi  z −i  X (z )
i =1 i =0

b2 − a2 Yz ( z ) b0 + b1  z −1 + + bM  z − M
H ( z) = =
X z ( z ) 1 + a1  z −1 + + aN  z − N

bM − aN −1
−
z −1

aN
© PFH Private Hochschule Göttingen 9
Direct form II realization of a recursive system
Example for a system 2nd Order

y ( k ) = −a1  y(k −1) − a2  y(k − 2) + b0  x(k ) + b1  x(k − 1) + b2  x(k − 2)

X ( z) Y ( z)
▪ The past values of x(k) and y(k) are stored
in the blocks z-1

▪ Some qualities of recursive systems

• The output signal is fed back

• As a consequence h(k) theoretically
is infinite
• Therefore, recursive systems are also
described as infinite impulse response
systems (IIR: Infinite Impulse Response)

© PFH Private Hochschule Göttingen 10

Cascade realization of a recursive system
The block diagram of the cascade, or series, realization

Y ( z ) = H1 ( z )  H 2 ( z )  H 3 ( z )  X ( z )

X ( z) Y ( z)
H1 ( z ) H2 ( z ) H3 ( z ) ▪ In practice, the
second-order filter
X ( z) Y ( z) module using the
direct-form I or
directform II is used
▪ The high-order filter
can be factored in the
cascade form with the
first- or second-order
sections

© PFH Private Hochschule Göttingen 11

Parallel realization of a recursive system
Parallel realization of first or second order transfer functions

Yz ( z ) =  H1 ( z ) + H 2 ( z ) + H 3 ( z )   X z ( z )

H1 ( z )

X ( z) Y ( z)
H2 ( z )

H3 ( z )

© PFH Private Hochschule Göttingen 12

Non-recursive Systems
Non-recursive systems are also described as FIR-filters or transversal filters

x (k ) y (k )
b0
Y ( z) z −1 z −1
H (z) = z = b0 + b1  z −1 + + bM  z −M
−
X z ( z) b1 a1
z −1 z −1
− a2
b2

▪ The denominator is eliminated

bM −1 − aM −1
▪ Only the numerator exist
−
▪ This is practicable only with digital filters z −1 z −1
bM aM
▪ FIR : finite impulse response
© PFH Private Hochschule Göttingen 13
Transfer function of non-recursive systems
▪ Finite difference equation
N
y ( k ) =  bi  x ( k − i )
i =0

▪ With the z-transformation

N
Y ( z ) =  bi  z − i  X ( z )
i =0

▪ Transfer function

Yz ( z )
H ( z) = = b0 + b1  z −1 + + bN  z − N
X z ( z)

© PFH Private Hochschule Göttingen 14

Properties of non-recursive systems (filters)
▪ The length of the impulse response of a FIR- filter N-th order is equal to
N+1

➢ The output signal of a FIR-filter can be calculated using the

convolution sum
 N N
y (n) = h(n)* x(n) =  h(i)  x(n − i) =  h(i)  x(n − i) =  b  x(n − i)
i =− i =0 i =0
i

▪ This is exactly the finite difference equation of the non-recursive system

▪ Convolution sum and finite difference equations of a FIR-system (FIR-filter)

are identical
© PFH Private Hochschule Göttingen 15
Structure of a FIR-filter (transversal-filter)
N
Y ( z ) =  bi  z − i  X ( z )
i =0

Xx ((kz) ) Yy((zk))
b0 x ( k − 1) x ( k − 2) x (k − N )
−1
x (k )
z z −1 z −1 z −1 z −1
b1
b0 b1 b2 bN
z −1
b2 y (k )


bN −1
z −1 ▪ This structure will be used by adaptive filter!
bN

© PFH Private Hochschule Göttingen 16

Convolution for discrete-time systems
The convolution sum can be described about the following matrix equation

 y ( 0)  h ( 0) 0 
   
 y ( 1)   h (1) h ( 0) 
 y ( 2)  h ( 2) h (1)   x (0) 
     
 = h ( 2) h ( 0 )   x (1) 

  h ( m ) h (1)   
     
   h (m) h ( 2 )   x ( N − 1) 
   
   
 y ( N + m − 1)   0 h ( m ) 

Length N and m
x (k ) = 0 für k  0, k  N  y ( k ) = 0 für k  0, k  N + m
© PFH Private Hochschule Göttingen 17
Convolution for discrete-time systems
The convolution sum of two signals with finite length will be shown clearly with
the following table
x1(k)={1,2,3} x2(k)={1,0,2,-1}
k -3 -2 -1 0 1 2 3 4 5 6 y(n) =x1(k)*x2(k)
x2(k) 0 0 0 1 0 2 -1 0 0 0 0
x1(-k) -1 3 2 1 0 0 0 0 0 0 0 0
0 0 3 2 1 0 0 0 0 0 0 1
1 0 0 3 2 1 0 0 0 0 0 2
2 0 0 0 3 2 1 0 0 0 0 5
3 0 0 0 0 3 2 1 0 0 0 3
4 0 0 0 0 0 3 2 1 0 0 4
n 5 0 0 0 0 0 0 3 2 1 0 -3 n
6 0 0 0 0 0 0 0 0 0 0 0

The length of the output signal is: L= N1 + N2 -1 = 6

© PFH Private Hochschule Göttingen 18

Impulse response of a sampled data system
Example of a 1st order system

Unit step function

Weighting function g(k)

Step response y(k)=h(k)

© PFH Private Hochschule Göttingen 19

Example for a non-recursive system
The finite difference equation for a 3rd order system

y[n] = b0  x[n] + b1  x[n − 1] + b2  x[n − 2] + b3  x[n − 3]

▪ Characteristics of non-recursive
systems X ( z)
− No feedback of the output signal
(all ai = 0)
− The impulse responses h[n] of
non-recursive systems have a
limited time duration Y ( z)
− The impulse responses h[n] of
non-recursive systems N-th order
h(k ) = h0 , h1 , hN  = b0 , b1 , bN 
are equal to the coefficients of the
system

© PFH Private Hochschule Göttingen 20