Signals & Systems

2021-1 Week 5

School of Electronics Engineering

Hyunyeol Lee, Ph.D.
1.6 Basic System Properties

• Memoryless:
• The output at a given time is dependent only on the input at that same time

• Examples of memoryless (static) system:

• Examples of memory (dynamic) system:

• Must store the preceding values of the input

• Differential operation : memory (dynamic) system

• Invertibility:
• Invertible if distinct inputs lead to distinct outputs (bijective function: 전단
사)
• If a system is invertible, then an inverse system exists that yields an output
equal to the input to the first system

𝑦 ( 𝑡 ) = 𝑥 (𝑡 ) 𝑤 =√ 𝑦 =|𝑥|
2

not invertible

𝑡
𝑑𝑦
𝑦 ( 𝑡 )= ∫ 𝑥 ( 𝜏 ) 𝑑 𝜏 𝑑𝑡 =𝑤=𝑥
−∞
• Causality:
• Causal if the output at any time depends only on past and present values of the input,
or if the system cannot anticipate the future input signal
• Examples of non-causal system:

• Example 1.12 (causal)

(non-causal)

• y[n]={x[n-1]+x[n]+x[n+1]}/3 : non-causal

• y[n]={x[n-2]+x[n-1]+x[n]}/3 : causal

• Causality matters only when the independent variable represents time.

moving
average
system
• Stability:
• Stable if a bounded input leads to a bounded output (not diverge)

B D
stable system
|x(t)|< B < < C < y(t) < D <

-B C

y=tB: growing with time regardless of the

• Example 1.13:
bounded input |x(t)|<B  unstable system

stable system
• Example: 1st order linear ODE y’(t) + ay(t) = bx(t) (a,b: constants)
• Solution: y(t) = e−at∫eat bx(t)dt + c e−at
yp:particular solution yh: homogeneous solution

• For bounded x(t), i.e., |x(t)|<B,

• Case 1: a ≠ 0, |yp| = |e−at∫eat bx(t)dt|< |e−at∫eat bBdt|= |bB/a|

• Case 1-1: a > 0, yh is decaying with increasing t  stable

• Case 1-2: a < 0, yh is increasing with increasing t  unstable

• Case 2: a = 0  |yp| = |∫ bx(t)dt|< |∫ bBdt|= |bBt|  unstable

• y[n]+ay[n-1]=bx[n]: |a|<1  stable, |a|≥1  unstable

• Time-invariance:
• Given any pair of input/output:

• A time-invariant system satisfies:

• A convenient way to test system’s time-invariance:

• Given , test whether

• Ex: Examine the time-invariance of the system:

•

• Let , then the system’s output , which is not equal to

• Thus, the system is not time-invariant.

• Linearity:
• Given any two pairs of input/output: ,

• A linear system satisfies the superposition principle:

• Additivity:

• Homogeneity (scaling property):

• Via the two properties above,

• y(t)=ax : linear

• y[n]=ax[n]+b : incrementally linear (nonlinear).

• y1+y2 = ax1+b + ax2+b = a(x1+x2)+2b ≠ a(x1+x2)+b

• ∆y[n]=y[n]-y[n-1], ∆x[n]=x[n]-x[n-1]  ∆y[n]=a∆x[n]

Properties of Ordinary Differential Equations

• y’(t) + a(t)y(t) = b(t)x(t)

• The system is time-invariant if a(t)= a(t-t0) and b(t)= b(t-t0) t0.

(i.e., both a and b are constants)

• Verify its linearity using the superposition principle

• Generalization to n-th order ODE

• Details in the classroom

Chapter 2. LTI Systems

• In linear + time-invariant (LTI) systems,

• Signals can be represented as linear combinations of shifted impulses.

• Impulse response: complete characterization of a system

• An output of a LTI system can be represented by convolution of an input and

an impulse response of the system.
2.1.1 Representation of DT Signals in Terms
of Impulses
• x[n] as summation of shifted impulses:

superposition of scaled versions of shifted unit impulses

(sifting property: “sift” x[k] only when n = k)

• In the case of x[n] = u[n] (unit step: u[n]=1 for n>=0 and =0 for n<0)
2.1.2 The D-T Unit Impulse Response and the
Convolution Sum Representation of LTI Systems

Let hk[n] denote the response of the linear

δ[n-k] hk[n]
system to the shifted unit impulse
Linear
system
By the superposition property
(homogeneity, additivity) of
the linear system
= x[n] = y[n]

• If we know hk[n] for each k, we can construct the output y[n] in

response to an arbitrary input.
• If the linear system is also time-invariant (TI),

δ[n] h0[n]
TI system
δ[n-k] h0[n-k] = hk[n]

• we just need to know h0[n] =: h[n] (the unit impulse response: the
output of the LTI system in response to δ[n]).
 LTI system

*: Convolution operator

• An LTI system is completely characterized by its response to a single

signal, namely, the impulse response h[n].
Impulse Response and Convolution
𝛿[n] h[n] (impulse response)
𝛿[n-k] h[n-k] (time-invariance)

x[k]𝛿[n-k] LTI system x[k]h[n-k] (homogeneity)

Σkx[k]𝛿[n-k] Σkx[k]h[n-k] (additivity)

𝛿h

x[n] LTI system y[n]  x[n] * h[n] (convolution)

h[n]
Calculating Convolution: Method 1

x[k]

*
h[k]

k
 Calculating Convolution: Method 2


• Four steps for each n of y[n]   x[k ] h[n  k ]

k  

=h[-(k-n)]

x[k] h[k]

k k
• Example 2.3 Given an input x[n] and a unit impulse response h[n]:
0<α<1 , calculate the output y[n].

• Strategy to calculate convolution:

• Perform the blue box operations for each n too tedious =h[-(k-n)]

• Set intervals for n such that overlap between x and h

in each interval has the same pattern w.r.t n
• From the figures on the right, we can figure that
the given convolution takes
two intervals: n ≥ 0 and n < 0.
n=0
• 1) n < 0: no overlap btw n and h  y[n] = 0

• 2) n ≥ 0: overlap appears from 0 to n

n = -1

n=1
• Given the results in 1) and 2), y[n] can be expressed as:
• Example 2.4
Interval 1: no overlap  y[n] = 0

Interval 2: overlap for 0 ≤ k ≤ n

Interval 3: overlap for 0 ≤ k ≤ 4

Interval 4: overlap for n-6 ≤ k ≤ 4

 Interval 5: no overlap  y[n] = 0

Summary
2.2 C-T LTI Systems: The Convolution
Integral
• In analogy with the results in D-T systems, here we will obtain a complete
characterization of C-T LTI system in terms of its unit impulse response.

• C-T unit impulse: regarded as idealization of a very short pulse

 Enables a representation of arbitrary C-T signals in terms of impulses

 Develop the convolution integral representation of C-T LTI systems.

LTI
 The C-T Unit Impulse

lim 𝛿∆ (𝑡 )
lim 𝑢 ∆ (𝑡 ) ∆→0
∆→0

1 indicates its area is 1,

not a value at t=0
Some more on the C-T Unit Impulse

• Strictly speaking, because with , and neither for

• Nevertheless, can be defined, which is 1. Thus, for convenience, we

write
Sampling Property of C-T Unit Impulse
∞ ∞
lim
∆→0
∫ 𝑥 (𝑡 ) 𝛿 ∆ (𝑡 ) 𝑑𝑡 ≈ lim ∫ 𝑥 ( 0 ) 𝛿∆ ( 𝑡 ) 𝑑𝑡=𝑥 (0)
∆ →0
−∞ −∞

lim ❑
∆→0

∫ 𝑥 ( 𝑡 ) 𝛿 ( 𝑡 ) 𝑑𝑡 =𝑥 (0)
−∞

∫ 𝑥 ( 𝑡 ) 𝛿 ( 𝑡 −𝑡0 ) 𝑑𝑡=𝑥 (𝑡0 )

−∞
2.2.1 The Representation of C-T Signals in
Terms of Impulses
• Derivation 1: Piecewise linear (staircase) approximation of x(t)

=1/∆ for k∆ <t<(k+1)∆

∞
𝑥 ( 𝑡 )=lim
∆→0
∫ 𝑥 (𝜏 ) 𝛿∆ ( 𝑡 −𝜏 ) 𝑑 𝜏
−∞
∞
𝑥 ( 𝑡 )= ∫ 𝑥 ( 𝜏 ) 𝛿 ( 𝑡 −𝜏 ) 𝑑 𝜏
−∞
• In the case x(t) = u(t),

• Derivation 2.
∞ ∞ ∞

∫ 𝑥 ( 𝜏 ) 𝛿 ( 𝑡 − 𝜏 ) 𝑑 𝜏 = ∫ 𝑥 ( 𝑡 ) 𝛿 ( 𝑡 − 𝜏 ) 𝑑 𝜏 = 𝑥 (𝑡 ) ∫ 𝛿 ( 𝑡 − 𝜏 ) 𝑑 𝜏 = 𝑥 (𝑡 )
−∞ −∞ −∞

Using the sampling property

 2.2.2 The C-T Unit Impulse Response and the
Convolution Integral Representation of LTI Systems

δ∆(t-k∆) h^ k ∆(t) Let denote the response

of the LTI system to the shifted
x(k∆)∆δ∆(t-k∆) x(k)∆
unit impulse
∑kx(k∆)∆δ∆(t-k∆) Linear ∑kx(k)∆
system
lim∑kx(k∆)∆δ∆(t-k∆) lim∑kx(k)∆
∞
𝑦 ( 𝑡 )= ∫ 𝑥 ( 𝜏 ) h𝜏 (𝑡 ) 𝑑 𝜏
−∞
If the system is TI
 LTI system

x(t) is a weighted sum of shifted impulses y(t) is a weighted sum of shifted impulse responses

• The procedure for evaluating the convolution integral:

• consider x(τ) and h(τ) and reflect h(τ) to h(-τ)

• Shift h(-τ) by an amount of t (shift right if t>0, shift left if t<0)

• multiply h(t-τ) with x(τ)

• Integrating x(τ)h(t-τ) w.r.t τ from negative infinity to positive infinity

Example 2.6
y(t)=0
Example 2.7