4. Discrete signals.

Z transform

Definiții. Semnale cu proprietăți remarcabile.

• Discrete signals x: Z →C take values at well known time instants

t1 , t2 , t3 ....  x(t1 ), x(t2 ), x(t3 ) .... and in rest is null or does not exist.

• usually the samples are equally spaced tn = t0 + nT  x(nT ) = xn  C , where n  Z

• A discrete signal is a succession of numbers x  n  C , unde n  Z . If

o x  n  = 0, n  0  causal signal,

o x  n  = 0, n  0 anticausal signal
• We define
1, n = 0
- discrete unit impulse:  n = 
0, n  0
1, n  0
- discrete causal step function:  n = 
0, n  0
0, n  0
- discrete anti-causal step function  − n − 1 = 
1, n  0
• Properties of [n]
o Sonding  n − n0 xn =  n − n0 xn0 
 nxn =  nx0
(x *  )n = ( * x )xn = xn,
 
o Convolution
unde ( x1 * x2 )n =  x1mx2 n − m =  x n − mx m = (x
1 2 2 * x1 )n
m=− m=−
Analogue signals sampling
• Let the band limited signals xa (t )  X a ( ), X a ( ) = 0 pt   M

 xa (t ) t = nT xe (nT ) = xa (nT )
• The sampled signal is, then e
x (t ) =  
xe (nT ) = xn
not
0 t  nT

x(t ), xe (t ) = xn

t
0 T 2T 3T

 
 xn =  T (t )  xa (t ) =   (t − nT )  xa (t ) =  xa (nT ) (t − nT )
k  k

2 2
( ) k ( ) 1
    ( ) ( ) ( ) 1

jn t jn t

But T t =  t − nT = e T
 x n = x a nT  t − nT = x a t e T
T k k T n
• Computing the Fourier transform of the relation above
 2 
F xn = X d ( ) =  xa (nT )F  (t − nT ) = xa (nT )e − jnT = X a ( )
1
k k
   − n 
T k  T 
2
Using the notation  = for the angular sampling frequency, it results
T

F xn = X d ( ) =  xa (nT )e − jnT =  X a (n) ( − n) (*)

1
k T k
o If   2M - the spectra repeats  the original signal can be obtained by filtering
o The maximum angular sampling frequency that can be used without spectral
aliasing  Nyquist frequency   N = 2M  minimum sampling period
1 1
Tmin = =
FN 2 f M
 X a ( )

− M M 

X e ( )
X a ( )
1
X a ( +  e )
1
X a ( −  e )
1
T T T

- e − M -  e − e + M − M M e − M  e e + M 

X ( )
X a ( ) X a ( +  e )
1 1
X a ( −  ee )
1
T T T

- e − M - e − M  M e e + M 
−  e +  M e − M
aliere spectrala
Signal recovery
• In order to recover the signal  an ideal low pass filter has to be used F
  
1,   = ,
H TJI ( ) =  2 T
0 rest

 the recovered signal is, thus
  
T T T
  jt
xa (t ) =  X ( )e TX ( )e    x (nT )e
1 jt 1 jt T − jnT
d = d = e d =
2 X a ( )=TX d ( ) 2 2
a d a
− − −
n 
T T T

 
=  xa (nT )sinc (t − nT )
n T 
The Fourier Transform of Discrete Signals (FTDS)

• From (*)
F xn = X d ( ) =  xne − jnT
k

 2  2
OBS! X d ( ) = X d   + k  is periodic in frequency domain with the period  the
 T  T
above relation may be envisioned as an Exponential Fourier Series decomposition in time
domain  this

xn =  X ( )e
T  j nT
d
2
d
2
T

Justificare!!
• Using the normalized frequency  = T the above relations become
F xn = X d ( ) =  xne − jn
k

X d ( ) = X d ( + k 2 )

xn =  X ( )e
1 jn
d
2
d
2
T

Proprietățile FTSD
 x1 n  X d 1 ( ) 
1. Liniarity    a1 x1 n + a2 x2 n  a1 X d 1 ( ) + a2 X d 2 ( )
 x2 n  X d 2 ( )
Demo - Hw
2. Discrete Time shifting xn − n0   X d ( )e − jn0 
Demo – Hw

3. Frequency shifting xne jn 0  X d ( −  0 )

Demo - Hw

4. Conjugation and symmetry x * n  X d* (− )

 x− n  X d (− )

From those results also x  −n   X d* (  ) Conjugation of the FTSD transform

 
Re X d* (  ) = even simmetry

Consequences x ( t )  R 
 
Im X d (  ) = odd simmetry
*

 X d* (  ) = even function


 
Demo ?
arg X d (  ) = odd function
*

Demo ? we use X d ( ) = X d ( − ) and

*

X d ( ) = Re  X d ( ) + j Im  X d ( ) = X d ( ) e
j arg X d ( )

X d* ( − ) = Re  X d ( − ) − j Im  X d ( − ) = X d ( − ) e
− j arg− X d ( )

5. Convolution in discrete time

x1 n  X d 1 ( ) , x2 n  X d 2 ( )
 
(x1 * x2 )n =  x1 mx2 n − m =  x1 n − mx2 m  X d 1 ()X d 2 ()
m = − m = −

Demo ? Hw
Consequences


- if n=0, ( x1 * x2 )  n  =  x  m x  −m  X (  ) X (  ) , then change

1 2 d1 d2 -m=m
m =−

x2  −m = x2  m  deci X d 2 ( ) = X d 2 ( − )


 x mx m = 2  X ()X (− )d 

1
(1) 1 2 d1 d2
m = −  2

 x m = 2  X ()X (− )d 

21
Parceval d d
m = −  2

- dacă n=0, ( x1 * x2 )  n  =  x1  m  x2  −m   X d 1 (  ) X d 2 (  ) , then change

m =−

x2  −m = x2*  m  deci X d 2 ( ) = X d* 2 ( )

 

 x mx m = 2  X ()X ()d 

1 * *
(2) 1 2 d1 d2
m = −  2

Rayleigh  xm = X d () d 

1

2 2

m = − 2 2

Demo- dacă n=0, ( x1 * x2 )  n  =  x1  m  x2  −m   X d 1 (  ) X d 2 (  ) , apoi schimbăm

m =−

x2  −m = x2  m  deci X d 2 ( ) = X d* 2 ( )

The Z transform


xn  X z ( z ) =  xnz −n
, zD  C
n = −

D = is the convergence domain of the z transform

F xn = X d ( ) =  xne − jn
k

X d ( ) = X d ( + k 2 )
j
Remark dacă e = 1 D X d (  ) = X z ( z ) z =e j 
The convergence domain of Tz
• in general R1  z  R2
o Finite length sequences R1 = 0, R2 = 
o Causal sequences R1 = finit, R2 = 
o Anti-causal sequences
Justificare

Inverse Tz

1. 1. Series expansion
2. can be used when X ( z ) is a rational function with simple poles or a polynomial

function. Then, if X ( z ) can be decomposed in series

X ( z ) = a0 + a1 z −1 + a2 z −2 + ... + ak z − k + ..
 and, taking into account
a n − k   az − k
we have
x ( n ) = a0  n  + a1  n − 1 + a2  n − 2 + ... + ak   n − k  + ..

3. Metoda recunoaşterii
X (z)
• We decompose the ratio , in simple fractions around each of its poles; then we
z
can use the recognition formulas
- causal sequences:

C nk a n−k  n 
z
z  a
(z − a )k +1 ,
- anti-causal sequenes

− C nk b n−k  − n − 1 
z
z b
(z − b )k +1 ,
where
n(n − 1)...(n − k + 1)  n  = 1
Cnk = ,
k! 0
and a, bC are constants.
• If the convergence domain is R1  z  R2 and X z ( z ) has the poles

o inside the domain p , ordin n , p

int
i
int
i
int
i  R1

outside the domain p , ordin n , p  R2

ext ext ext
o i i i

X z (z ) ni −1 ni −1
int ext
Aik Bik
=  +  
z i

k =1 z − pi

int k +1
(i
 
k =1 z − pi )
ext k +1

( )
z  p iint z  p iext
cauzal anti − cauzal

niint −1 niext −1
xn =   Aik C p k
n ( ) int n−k
i  n −   Bik Cnk ( piext )  − n − 1
n−k

i k =1 i k =1

3. Inverse integral computation

 We can determine the discrete signal using

xn =
1

2j C
X z ( z ) z n−1dz

where C is a circle inside the convergence domain of the z transform, with the center
inside the origin.
xn =  rez X ( z) z n −1
, n0
polii int eriori C

xn = − rez X ( z ) z n−1 −  rez X ( z) z n −1 ,

, n  −1
z =
polii exteriori C

For the residue at infinity we can use

1 1
rez G ( z ) = − lim G 
z = z →0 z2  z 

Properties

 x1 n  X z1 ( z ) 
1. Linearity    a1 x1 n + a2 x2 n  a1 X z1 ( z ) + a2 X z 2 (z )
 x2 n  X z 2 ( z )

2. Discrete time shifting xn − n0   X z (z )z − n0

Demo ? Hw

3. Multiplying the discrete signal with an exponential (scaling in the z plane) =

(
xn  X z (z ), R1  z  R2  a n xn  X z a −1 z , a R1  z  a R2 )
Demo ? Hw

Remark - the multiplication of a sequence with an exponential may lead to change in the
poles / zeros position
- if a>1 farther from the origin
- if a<1 closer to the origin
- of a = 1, a  C  rotation in the z plane
 dX z (z )
4. derivation of the z transform nxn  − z R1  z  R2
dz
Demo ? Hw

Consequence na  n 
n az
(z − a )2

5. Complex conjugation of the signal

xn  X z (z ), R1  z  R2  x* n  X z* (z * ), R1  z  R2
Demo ? Hw

1 1
xn  X z (z ), R1  z  R2  x− n  X z  ,
1
6. Symmetry  z
 z  R2 R1
Demo ? Hw

7. Convoluția
 x1 n  X z1 (z ), R1  z  R1
 (x1 * x2 )n  X z1 (z )X z 2 (z ), max R1 , R 2   z  min R1 R 2 
 1 2

 x2 n  X z 2 (z ), R  z  R
2 2 1 1 2 2

1 2

Unilateral z Transform
For causal signals x[n] = x[n] [n]  0 pt n  0

X z ( z ) =  x  n  z − n z  r1
n =0

1. The properties are preserved, excepting the discrete time shifting property  the initial
conditions must be taken into account
Z xn + k   z k X z (z ) − z k x0 − z k −1 x1 − .... − zxk − 1
Z xn − k   z − k X z (z ) + z − k +1 x− 1 + z − k +2 x− 2 − .... + x− k 
Transformata Fourier a semnalelor discrete periodice. Seria Fouriei Discreta (SFD)
Fie xn = xn + kN  ()n  N, k  N*  poate fi dezvoltata in serie  folosind setul
 jk 2N  2
j (k + N )
2
, k = 1, (N − 1),
jk
e e N
=e N

 
Obs - deosebirea fundamentala intre setul de mai sus si cel folosit in SFE este ca in acest
caz avem o suma finita  se preteaza la calcule
N −1 2 N −1
xn =  ck e =  ck wNkn
jkn
N

k =0 k =0

N −1 2 N −1
ck  =  xne  xnw
1 − jkn 1 − kn
N
= N
N n =0 N n=0

2
wNkn = wNk (n + N )
jkn
Unde w = e
kn N
N ,
Spunem ca xn si ck  sunt perechi in sensul seriei Fourier Discrete cu perioada N
xn ck 
N

Proprietățile SFD
1. Liniaritate

 x1 n c1 k  
 
  a1 x1 n + a2 x2 n a1c1 k  + a2 c2 k , N = cmmmc( N1 , N 2 )
N1

x n  
 2 N 2 2 c k  N

Demo ?

2. Deplasarea în timp normat

x(n − n0 )N  ck wN− n0 k unde (n − n0 )N = (n − n0 ) modulo N

not

Demo ?

3. Deplasarea in domeniul SFD

xnw  c(k − k 0 )N  unde (k − k 0 )N = (k − k 0 ) modulo N

not
k0 n
N
N

Demo ?
4. Conjugarea semnalului x * n c * (− k )N 
N

Demo ?

5. Simetria x− n c(− k )N 

N

Demo ?

Consecințe
1. xn R  xn = x n  ck  = c (− k )N 
* *

Reck  = functie para, Imck  = functie impara,

ck  = functie para, argck  = simetrie impara

2. Combinand simetria si conjugarea  x (− n )N  c * k 

*
N

6. Convoluția circulara
 x1 n c1 k  
  (x1 * x2 )N n c1 k c2 k 
N

 2
x n  
N
c 2 k  N

N −1 N −1
(x1 * x2 )N n = 1  x1 mx2 n − m = 1  x n − mx m
1 2
N m=0 N m=0
Demo ?

Consecințe
N −1 N −1

 x n =  ck c(− k ) 

1 2
(1) N Parseval
N n =0 n =0

N −1 N −1
xn =  ck 
1

2 2
(2) Rayleigh
N n =0 n =0

Demo ?