33 34

2 Linear Time–Invariant Systems 2.1 Discrete–Time LTI Systems: The Convolution Sum

Classical subject of signals and systems theory  Use linearity and time invariance to obtain compact characterization
of discrete-time LTI system,
 Amenable to analysis, well-developed theory, rich set of tools i.e., to compactly express transformation x[n] −→ y[n]
 Model/approximate many physical processes  Observe: any signal can be composed as superposition of time–
shifted unit impulses — Sampling or sifting property of unit impulse
 Allow for relatively easy synthesis
∞
X
Outline x[n] = x[k]δ[n − k]
k=−∞
2.1 Discrete–Time LTI Systems: The Convolution Sum Example:
2.2 Continuous–Time LTI Systems: The Convolution Integral x[n]

2.3 Properties of Linear Time–Invariant Systems

−2 −1 0 1 2 3 n
2.4 Differential and Difference Equation Representation
x[−1]δ[n + 1]

−2 −1 0 1 2 3 n

x[0]δ[n]

−2 −1 0 1 2 3 n

x[1]δ[n − 1]

−2 −1 0 1 2 3 n

Obviously: x[n] = x[−1]δ[n + 1] + x[0]δ[n] + x[1]δ[n − 1]

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 35 36

Example: Example:
Unit step function – Input: x[n] = δ[n] + 2δ[n − 1] + 3δ[n − 2]
n
X ∞
X Impulse response: h[n] = 2δ[n] + δ[n − 1]
u[n] = δ[k] = δ[n − k] Output: y[n] = ?
k=−∞ k=0
– Convolution sum
∞
X
 LTI system characterization y[n] = x[k]h[n − k]
k=−∞
– Define
δ[n − k] −→ hk [n] Simplify (x[k] = 0 for k < 0 and k > 2)
2
– Linearity y[n] =
X
x[k]h[n − k]
∞ ∞
X X k=0
x[n] = x[k]δ[n − k] −→ y[n] = x[k]hk [n]
k=−∞ k=−∞
Causal system (h[n] = 0 for n < 0)
– Time invariance y[n] = 0 for n < 0

δ[n − k] −→ hk [n] = h0[n − k] n ≥ 0:

2
– Impulse response
X
y[0] = x[k]h[−k] = x[0]h[0] = 2
h[n] ≡ h0[n] k=0
2
⇒ Convolution sum X
∞
y[1] = x[k]h[1 − k] = x[0]h[1] + x[1]h[0] = 5
k=0
X
y[n] = x[k]h[n − k] 2
X
k=−∞
y[2] = x[k]h[2 − k] = x[0]h[2] + x[1]h[1] + x[2]h[0] = 8
k=0
 Symbolical representation of convolution 2
X
y[n] = x[n] ∗ h[n] y[3] = x[k]h[3 − k] = x[2]h[1] = 3
k=0
n > 3: h[n − k] = 0 ⇒ y[n] = 0
LTI systems are fully described by their impulse response.

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 37 38

 Graphical interpretation 2.2 Continuous–Time LTI Systems: The Convolution Integral

– Time-reverse h[k] ⇒ h[−k]
– Shift h[−k] by n ⇒ h[n − k]  Similar to the discrete–time case:
– Sum products of overlapping components system characterization x(t) −→ y(t) with impulse response h(t)

For previous example:  Observation: any continuous-time signal can be expressed as su-
perposition of infinitesimally time–shifted continuous-time unit im-
x[k] h[k]
pulses — Sampling or sifting property of unit impulse
−2 −1 0 1 2 3 k −2 −1 0 1 2 3 k Z∞
x(t) = x(τ )δ(t − τ ) dτ
−∞
∗
Example:

h[−k] Unit step function

y[0] : Zt Z∞
−2 −1 0 1 2 3 k
u(t) = δ(τ ) dτ = δ(t − τ ) dτ
h[1 − k] −∞ 0

y[1] :
−2 −1 0 1 2 3 k

h[2 − k]
y[2] :
−2 −1 0 1 2 3 k
h[3 − k]

y[3] :
−2 −1 0 1 2 3 k

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 39 40

 LTI system characterization Example:

– Define – Input: x(t) = e−at u(t) , Re{a} > 0
δ(t − τ ) −→ hτ (t) Impulse response: h(t) = u(t)
Output: y(t) = ?
– Linearity
– Convolution integral
Z∞ Z∞
Z∞
x(τ )δ(t − τ ) dτ = x(t) −→ y(t) = x(τ )hτ (t) dτ
y(t) = x(τ )h(t − τ ) dτ
−∞ −∞
−∞
– Time invariance Simplify (x(t) = 0 for t < 0 )
δ(t − τ ) −→ hτ (t) = h0(t − τ ) Z∞
y(t) = x(τ )h(t − τ ) dτ
– Impulse response 0
h(t) ≡ h0(t)
Causal system (h(t − τ ) = 0 for t − τ < 0)
⇒ Convolution integral Zt
y(t) = x(τ )h(t − τ ) dτ
Z∞
0
y(t) = x(τ )h(t − τ ) dτ
−∞ t≥0:
Zt t
1  1
Symbolically representation of convolution y(t) = e−aτ dτ = − e−aτ  = (1 − e−at )


a  a
0 0
y(t) = x(t) ∗ h(t)
⇒ For all t
1
y(t) = (1 − e−at )u(t)
a

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 41 42

 Graphical interpretation ∗ T ≤ t < 2T

– Time-reverse h(τ ) ⇒ h(−τ )
– Shift h(−τ ) by t ⇒ h(t − τ )
– Integrate products of overlapping components
T t 2T τ

Example: Zt−T Zt
 y(t) = 3 dt + 2 dt = 3(t − T ) + 2T
1, 0 ≤ t ≤ 2T
∗ Input: x(t) = 0 t−T
0, otherwise
∗ 2T ≤ t < 3T

 2, 0≤t<T
Impulse response: h(t) = 3, T ≤ t ≤ 2T
0 otherwise


h(t − τ )
∗t<0
T 2T t τ
x(τ ) Z2T Zt−T
y(t) = 2 dt + 3 dt = 2(3T − t) + 3T
t−T t−2T
t − 2T t T 2T τ

No overlap: y(t) = 0 ∗ 3T ≤ t < 4T

∗0≤t<T

T 2T t τ
Z2T
t T 2T τ y(t) = 3 dt = 3(4T − t)
Zt
y(t) = 1 · 2 dt = 2t t−2T

0
Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 43 44

∗ t ≥ 4T 2.3 Properties of Linear Time–Invariant Systems

 Linearity and time invariance ⇒ complete characterization by im-

pulse response
∞
X
T 2T t τ y[k] = x[k]h[n − k] = x[k] ∗ h[k]
k=−∞
No overlap: y(t) = 0
Z∞
y(t) = x(τ )h(t − τ ) dτ = x(t) ∗ h(t)
∗ In summary −∞


 0, t<0  Further properties based on and in terms of impulse response repre-




 2t, 0≤t<T sentation
2T + 3(t − T ), T ≤ t < 2T

y(t) =
 2(3T − t) + 3T, 2T ≤ t < 3T
 2.3.1 Commutative, Distributive, and Associative Property
3(4T − t), 3T ≤ t < 4T



of Convolution — Interconnections of LTI Systems


 0, t ≥ 4T
y(t)/T  Commutative property: order of the signals to be convolved can be
5 changed
4 – Continuous–time case
3 Z∞
2 x(t) ∗ h(t) = h(t) ∗ x(t) = h(τ )x(t − τ ) dτ
−∞
1
– Discrete–time case
1 2 3 4 ∞
t/T X
x[n] ∗ h[n] = h[n] ∗ x[n] = h[k]x[n − k]
k=−∞

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 45 46

– Implications:  Associative Property

∗ output of system with impulse response h(t) to input x(t) – Continuous–time case
identical to output of system with impulse response x(t) to
x(t) ∗ (h1(t) ∗ h2(t)) = (x(t) ∗ h1(t)) ∗ h2(t)
input h(t)
∗ Irrelevant whether h(t) or x(t) is reflected and shifted for – Discrete–time case
convolution x[n] ∗ (h1[n] ∗ h2[n]) = (x[n] ∗ h1[n]) ∗ h2[n]
– Consequently
 Distributive property
y[n] = x[n] ∗ h1[n] ∗ h2[n]
– Continuous–time case – Implications
x(t) ∗ (h1(t) + h2(t)) = x(t) ∗ h1(t) + x(t) ∗ h2(t) ∗ Chose convenient order of convolution
∗ Cascade connection of two LTI system represented by single
– Discrete–time case LTI system
x[n] ∗ (h1[n] + h2[n]) = x[n] ∗ h1[n] + x[n] ∗ h2[n] x(t) h1 (t) h2 (t) y(t)

– Implication ≡
Parallel connection of two LTI systems represented by single
x(t) h1 (t) ∗ h2 (t) y(t)
equivalent LTI system

h1 (t)  Associative + commutative property ⇒ order in a cascade of LTI

x(t) y(t) systems irrelevant
h2 (t) y(t) = (x(t) ∗ h1(t)) ∗ h2(t) = (x(t) ∗ h2(t)) ∗ h1(t)

≡ x(t) h1 (t) h2 (t) y(t)

x(t) h1 (t) + h2 (t) y(t) ≡

x(t) h2 (t) h1 (t) y(t)

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 47 48

Proofs:  Associative property

(for continuous–time case only) x(t) ∗ (h1(t) ∗ h2(t)) = x(t) ∗ (h2(t) ∗ h1(t))
Z∞
 Commutative property = x(t) ∗ h2(τ )h1(t − τ ) dτ
Z∞ Z−∞
τ 0 =t−τ −∞
x(t) ∗ h(t) = x(τ )h(t − τ ) dτ = − x(t − τ 0 )h(τ 0) dτ 0 Z∞ Z∞
−∞ ∞ = x(τ 0 ) h2(τ )h1(t − τ 0 − τ ) dτ dτ 0
Z∞
−∞ −∞
= x(t − τ 0)h(τ 0) dτ 0
Z∞ Z∞
−∞
= x(τ 0 )h1(t − τ − τ 0 ) dτ 0h2(τ ) dτ
= h(t) ∗ x(t)
 −∞ −∞
Z∞
 Distributive property = x(τ 0 )h1(t − τ 0 ) dτ 0 ∗ h2(t)
Z∞ −∞

x(t) ∗ (h1(t) + h2(t)) = x(τ )(h1(t − τ ) + h2(t − τ )) dτ = (x(t) ∗ h1(t)) ∗ h2(t)

−∞ 
Z∞ Z∞
= x(τ )h1(t − τ ) dτ + x(τ )h2(t − τ ) dτ
−∞ −∞
= x(t) ∗ h1(t) + x(t) ∗ h2(t)


Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 49 50

2.3.2 Relation between LTI System Properties and the Im- – Continuous-time LTI system
pulse Response
h(t) ∗ hinv (t) = δ(t)
 LTI systems with and without memory
Example:
– Recall: memoryless system if output signal depends only on
present value of input signal Accumulator: h[n] = u[n]
∞ n
– Discrete-time LTI system X X
∞
y[n] = x[k]u[n − k] = x[k]
! k=−∞ k=−∞
X
y[n] = x[k]h[n − k] = Kx[n] (K constant)
k=−∞ Note that u[n] − u[n − 1] = h[n] − h[n − 1] = δ[n]
⇒ h[n] = Kδ[n] = h[0]δ[n] ⇒ hinv [n] = δ[n] − δ[n − 1]
= Differentiator (see Example in 1.4.4)
– Continuous–time LTI system
Z ∞
!
y(t) = x(τ )h(t − τ ) dτ = Kx(t)
−∞
 Causality of LTI systems
⇒ h(t) = Kδ(t) = h(0)δ(t)
– Recall: causal system if output at any time depends only on
– Identity system: K = 1 past and present values of the input

x[n] = x[n] ∗ δ[n] – Discrete–time LTI system

and ∞ ∞ ∞
!
X X X
x(t) = x(t) ∗ δ(t) y[n] = x[k]h[n−k] = h[k]x[n−k] = h[k]x[n−k]
k=−∞ k=−∞ k=0

 Invertibility of LTI systems

⇒ h[n] = 0, for n < 0
– Invertible LTI system has LTI inverse
– Discrete–time LTI system – Continuous–time LTI system
h[n] ∗ hinv [n] = δ[n] h(t) = 0, for t < 0

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 51 52

 Stability of LTI systems – Continuous–time LTI system

– Recall: stable system if bounded output for any bounded input Sufficient and necessary condition for BIBO stability
Z∞
– Discrete–time LTI system |h(τ )| dτ < ∞
∗ Bounded input |x[n]| ≤ Bx ∀n −∞
∗ Output: h(t) is absolutely integrable
 ∞ 
X 
|y[n]| = 

h[k]x[n − k]
 Example:
 
k=−∞ Integrator: h(t) = u(t)
∞
Z∞ Z∞
X
≤ |h[k]||x[n − k]|
k=−∞ |u(τ )| dτ = 1 dτ = ∞
∞
X ! −∞ 0
≤ Bx |h[k]| ≤ By
k=−∞
This system is not stable (see Example in 1.4.6).
⇒ Sufficient condition for BIBO stability
∞
X  Summary
|h[n]| < ∞
n=−∞
Property Discrete-time LTI system Continuous-time LTI system
h[n] is absolutely summable
∗ Condition also necessary Memoryless h[n] = δ[n] h(t) = δ(t)
x[n] = h∗[−n]/|h[−n]|
∞ ∞ Invertibility h[n] ∗ hinv [n] = δ[n] h(t) ∗ hinv (t) = δ(t)
X X
⇒ y[0] = h[k]x[−k] = h[k]h∗[k]/|h[k]|
k=−∞ k=−∞ Causal h[n] = 0, for n < 0 h(t) = 0, for t < 0
∞
∞ R∞
X
= |h[k]| Stability
P
|h[n]| < ∞ |h(τ )| dτ < ∞
k=−∞ n=−∞ −∞

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 53 54

2.3.3 The Unit Step Response of an LTI System 2.4 Differential and Difference Equation Representation

 Alternative to impulse response:  Another powerful tool:

Unit step response s(t) or s[n] – Continuous-time system: description by

= system response to input u(t) or u[n] linear constant–coefficient differential equations
= unique representation for LTI system – Discrete-time system: description by
linear constant–coefficient difference equations
 Relationship:
Zt  In general: provides for initial conditions (initial system energy)
s(t) = u(t) ∗ h(t) = h(t) ∗ u(t) = h(τ ) dτ
 Here: condition of initial rest ⇒ causal LTI systems
−∞
Xn
s[n] = u[n] ∗ h[n] = h[n] ∗ u[n] = h[k] 2.4.1 Linear Constant–Coefficient Differential Equations
k=−∞

and  General N th order linear constant–coefficient differential equation

h(t) = ṡ(t) N M
h[n] = s[n] − s[n − 1]
X dk y(t) X dk x(t)
ak = bk
dtk dtk
k=0 k=0
Example:
RC circuit  Solution:
1 −t/(RC)
y(t) = yh (t) + yp (t)
– Impulse response: h(t) = RC e u(t)
Rt −τ /(RC) – Homogeneous solution
1
– Step response: s(t) = RC
e dτ N
0 X dk yh (t)
t ak =0
1

−τ /(RC)  dtk
= RC (−RC)e  k=0

0 – Particular solution
= (1 − e−t/(RC) ) u(t)
any solution yp(t) for given x(t)

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 55 56

– Condition of initial rest and x(t) = 0 for t < t0 ∗ Combination

dy(t)  dN −1y(t)  K 3t
y(t) = Ae−2t + e for t > 0

y(t) t− =
  = ... =  =0 5
0 dt t−0 dtN −1 t−0
(corresponds to causal LTI system) ∗ Condition of initial rest
y(t) = 0 for t < 0 since x(t) = 0 for t < 0
 Differential equation + initial condition specify system output com-
No impulse introduced at t = 0 ⇒ y(0) = 0
pletely
K
⇒ A+ =0
Example: 5
∗ Complete solution
– First–order differential equation
dy(t) y(t) = yh (t) + yp (t)
+ 2y(t) = x(t) K 3t
dt = (e − e−2t ) for t > 0
with input 5
x(t) = Ke3t for t > 0 , K ∈ IR
– Solution
∗ Particular solution: 2.4.2 Linear Constant–Coefficient Difference Equations
Hypothesis: yp (t) = Y e3t for t > 0
 General N th order linear constant–coefficient difference equation
⇒ 3Y e3t + 2Y e3t = Ke3t
N M
⇒ Y = K5 X X
ak y[n − k] = bk x[n − k]
⇒ yp (t) = K5 e3t for t > 0
k=0 k=0
∗ Homogeneous solution
dy(t)  Solution
+ 2y(t) = 0 y[n] = yh[n] + yp [n]
dt
Hypothesis: yh (t) = Aest – Homogeneous solution
⇒ Asest + 2Aest = 0 N
X
⇒ s = −2 ak y[n − k] = 0
⇒ yh (t) = Ae−2t for t > 0 k=0

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 57 58

– Particular solution or
1
any solution yp [n] for given x[n] y[n] = x[n] + y[n − 1]
2
– Condition of initial rest and x[n] = 0 for n < n0 Impulse response of underlying LTI system
(Condition of initial rest: y[n] = 0 for n < 0)
y[n] = 0 for n < n0
1
(corresponds to causal LTI system) y[0] = x[0] + y[−1] = 1
2
1 1
Alternatively: rewrite difference equation y[1] = x[1] + y[0] =
 2 2
M N
! 1 1
1 X X y[2] = x[2] + y[1] =
y[n] = bk x[n − k] − ak y[n − k] 2 4
a0 ···  n
k=0 k=1
1 1
and solve for successive values of n > n0 y[n] = x[n] + y[n − 1] =
2 2
⇒ impulse response h[n] = (1/2)nu[n]
Example:
Since h[n] has an infinite duration, the underlying system is
1. Assume nonrecursive difference equation
referred to as infinite impulse response (IIR) system.
y[n] = x[n] − x[n − 1]

Impulse response of underlying LTI system obtained for

x[n] = δ[n]
⇒ y[n] = h[n] = δ[n] − δ[n − 1]
Since h[n] = 0 for n > 2 the underlying system is referred to as
finite impulse response (FIR) system.

2. Consider recursive difference equation

1
y[n] − y[n − 1] = x[n]
2

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 59 60

2.4.3 Block Diagrams – N th order difference equation

 Block diagram: Illustration of systems useful for analysis, design, Direct Form I
and implementation
b0 1/a0
 Systems described by difference equations x[n] y[n]

– Basic modules
x2[n] D D
b1 −a1
x1[n] x1[n] + x2 [n]

a
x[n] ax[n] D D
b2 −a2

x[n] D x[n − 1]
.. .. .. ..
– First order difference equation
bM −1 −aN −1
y[n] + ay[n − 1] = bx[n]
or
y[n] = −ay[n − 1] + bx[n]
D D
Illustration: bM −aN
b
x[n] y[n]

2N delay elements necessary ?

D

−a y[n − 1]

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 61 62

– N th order difference equation – N th order difference equation

Interchange order of cascaded LTI subsystems Direct Form II

1/a0 b0 1/a0 b0
x[n] y[n] x[n] y[n]

D D D

−a1 b1 −a1 b1

D D D

−a2 b2 −a2 b2

.. .. .. .. .. .. .. ..
−aN −1 bM −1 −aN −1 bM −1

D D D

−aN bM −aN bM

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications
 63 64

 Systems described by differential equations – N th order differential equation

– Basic modules Direct Form II
x2(t)
1/aN bM
x(t) y(t)
x1 (t) x1(t) + x2 (t)
R
a
x(t) ax(t) −aN −1 bM −1

R Rt
x(t) x(τ ) dτ
−∞
R

– First order differential equation −aN −2 bM −2

dy(t)
+ ay(t) = bx(t)
dt .. .. .. ..
or
Zt
−a1 b1
y(t) = y(t0) + (bx(τ ) − ay(τ )) dτ
t0

Illustration: R
b R
x(t) y(t) −a0 b0

−a

– Convenient: implementation based on integrators (operational

amplifiers) instead of differentiators, which are difficult to im-
plement and extremely sensitive to errors and noise

Lampe, Schober: Signals and Communications Lampe, Schober: Signals and Communications