ELT2035 Signals & Systems

Lesson 4: Time domain representations of

discrete-time LTI systems

Hoang Gia Hung

Faculty of Electronics and Telecommunications
University of Engineering and Technology, VNU Hanoi
 Last lesson review
❑ Differential equation descriptions of CT LTI systems
➢ Calculus solution: homogeneous solution + particular solution.
➢ Engineering solution: zero-input response + zero-state response.

❑ Impulse response of CT LTI systems

➢ Impulse response = system’s output when input is impulse.
➢ Zero-state response = convolution(input, impulse response).

❑ Block diagram
➢ Not unique

❑ State-space
➢ Equations of state-variables
➢ Not unique, transformation between equivalent representations is
achieved by similarity transformation.

❑ Today’s lecture: Modelling of DT LTI systems in time domain

 Introduction
Input signal Output signal
(excitation) (response)
DT system
𝑥[𝑛] H[.] 𝑦[𝑛]

❑ Inputs and outputs are discrete-time signals.

➢ Used in situations which are inherently discrete-time, e.g. stock market
➢ Also used to process continuous-time signals via sampling. Why?
 The first example
❑ A person makes a deposit to a bank regularly at an interval of
𝑇 = 1 month. The bank pays certain interest on the account
during the period 𝑇, find the equation relating the account
balance to the deposit?
❑ Solution: denotes by
➢ 𝑥[𝑘]: the deposit made at the 𝑘𝑡ℎ discrete instant
𝑦[𝑘]: the account balance computed at the 𝑘𝑡ℎ instant after 𝑥[𝑘] is received
𝑟 : interest rate during the period 𝑇
➢ The balance 𝑦[𝑘] is the sum of (i) the previous balance 𝑦[𝑘 − 1], (ii) the
interest on 𝑦[𝑘 − 1], and (iii) the deposit 𝑥[𝑘]. Hence:
𝑦 𝑘 = 𝑦 𝑘 − 1 + 𝑟𝑦 𝑘 − 1 + 𝑥 𝑘 = 1 + 𝑟 𝑦 𝑘 − 1 + 𝑥[𝑘]
➢ Let 𝑎 = 1 + 𝑟, 𝑦 𝑘 is rewritten as 𝑦 𝑘 − 𝑎𝑦 𝑘 − 1 = 𝑥[𝑘] (1)
➢ Since the choice of 𝑘 is arbitrary, we can substitute 𝑘 + 1 for 𝑘 to obtain
𝑦 𝑘 + 1 − 𝑎𝑦 𝑘 = 𝑥[𝑘 + 1] (2)

❑ Eqs. (1), (2) are examples of first-order difference equations.

 Representation of DT LTI systems by
difference equations

Realization of the savings account system

❑ In general, DT LTI systems are characterized by linear,

constant-coefficient difference equations of either:
➢ Advance operator form
𝑦 𝑘 + 𝑛 + 𝑎𝑛−1 𝑦 𝑘 + 𝑛 − 1 + ⋯ + 𝑎1 𝑦 𝑘 + 1 + 𝑎0 𝑦 𝑘 = 𝑏𝑚 𝑥 𝑘 + 𝑚 +
𝑏𝑛−1 𝑥 𝑘 + 𝑚 − 1 + ⋯ 𝑏1 𝑥 𝑘 + 1 + 𝑏0 𝑥 𝑘 (3)
➢ Delay operator form
𝑦 𝑘 + 𝑎𝑛−1 𝑦 𝑘 − 1 + ⋯ + 𝑎1 𝑦 𝑘 − 𝑛 + 1 + 𝑎0 𝑦 𝑘 − 𝑛 = 𝑏𝑚 𝑥 𝑘 +
𝑏𝑛−1 𝑥 𝑘 − 1 + ⋯ 𝑏1 𝑥 𝑘 − 𝑚 + 1 + 𝑏0 𝑥 𝑘 − 𝑚 . (4)
❖ Causality condition: 𝑚 ≤ 𝑛.
Iterative solution of difference equation
❑ Eq. (4) can be rewritten as:
𝑦 𝑘 = −𝑎𝑛−1 𝑦 𝑘 − 1 − ⋯ − 𝑎1 𝑦 𝑘 − 𝑛 + 1 − 𝑎0 𝑦 𝑘 − 𝑛 +
𝑏𝑚 𝑥 𝑘 + 𝑏𝑛−1 𝑥 𝑘 − 1 + ⋯ 𝑏1 𝑥 𝑘 − 𝑚 + 1 + 𝑏0 𝑥 𝑘 − 𝑚
➢ If the system is causal, the output at the 𝑘𝑡ℎ instant, 𝑦 𝑘 , is computed from
the past 𝑛 values of the output and 𝑛 + 1 values of the input.
➢ Starting with 𝑦 0 , if the input is causal 𝑦 0 is computed from 𝑥 0 and 𝑛
initial conditions:
▪ 𝑦 0 = 𝑏0 𝑥 0 − 𝑎𝑛−1 𝑦 −1 − ⋯ − 𝑎1 𝑦 −𝑛 + 1 − 𝑎0 𝑦 −𝑛 .
➢ Repeat the procedure iteratively → all future outputs can be computed.
▪ 𝑦 1 = 𝑏1 𝑥 1 + 𝑏0 𝑥 0 − 𝑎𝑛−1 𝑦 0 − ⋯ − 𝑎0 𝑦 −𝑛 + 1
▪ …
❖ Iterative solution is NOT closed-form.
 Iterative solution example
𝑥𝑘

System
𝑘
𝑦[𝑘] = 𝑥[𝑘] + 𝑥[𝑘 − 1] + 𝑥[𝑘 − 2]

𝑦 −1 = 𝑥 −1 + 𝑥 −2 + 𝑥 −3

𝑦 1 = 𝑥 1 + 𝑥 0 + 𝑥 −1

𝑦 3 =𝑥 3 +𝑥 2 +𝑥 1
 Closed-form zero-input response
❑ Total response = zero-input response + zero-state response
➢ Rewrite (3) as 𝐸 𝑛 + 𝑎𝑛−1 𝐸 𝑛−1 + ⋯ + 𝑎0 𝑦[𝑘] = ሺ𝑏𝑚 𝐸 𝑚 + 𝑏𝑚−1 𝐸 𝑚−1 + ⋯ +
𝑏0 ሻ𝑥[𝑘], where 𝐸 the operation that advance a signal by 1 time unit.
➢ Eq. (3) becomes: 𝑄 𝐸 𝑦[𝑘] = 𝑃 𝐸 𝑥[𝑘].

❑ The zero-input response 𝑦0 𝑘 :

➢ Solution to 𝐸 𝑛 + 𝑎𝑛−1 𝐸 𝑛−1 + ⋯ + 𝑎0 𝑦 𝑘 = 0 → 𝑦0 [𝑘] must have the form
𝑐𝛾 𝑘 → 𝛾 𝑛 + 𝑎𝑛−1 𝛾 𝑛−1 + ⋯ + 𝑎0 = 0 or 𝑄 𝛾 = 0.
➢ Since 𝑄 𝛾 = 0 has 𝑛 solutions 𝛾1, … , 𝛾𝑛 → 𝑦0 𝑘 = 𝑐1 𝛾1𝑘 + ⋯ + 𝑐𝑛 𝛾𝑛𝑘 where
𝑐1 ,…, 𝑐𝑛 are arbitrary constants defined by 𝑛 initial conditions.
➢ 𝑄 𝛾 is referred to as the characteristic polynomial of the system while
𝑄 𝛾 = 0 is called the characteristic equation.
➢ 𝛾1, … , 𝛾𝑛 are called characteristic roots (a.k.a. eigenvalues, natural
frequencies).
➢ 𝛾𝑖𝑘 are called characteristic/natural modes of the system → 𝑦0 𝑘 is a
linear combination of characteristic modes of the system
 Relation with classical solution of
difference equations
❑ Engineering approach (two initial value problem):
Total response = Zero − input response + Zero − state response
𝑚𝑜𝑑𝑒𝑠 𝑚𝑜𝑑𝑒𝑠+𝑛𝑜𝑛𝑚𝑜𝑑𝑒𝑠

Classical approach (one initial value problem):

Total response = Natural response + Forced response
𝑚𝑜𝑑𝑒𝑠 𝑛𝑜𝑛𝑚𝑜𝑑𝑒𝑠

❑ Natural vs. zero−input responses:

➢ Same form; only differ in arbitrary constants: Natural response = zero-input
response + all characteristic mode terms from the zero-state response.
➢ Different initial conditions

❑ Forced vs. zero−state responses:

➢ Forced response = zero-state response - characteristic mode terms.
➢ Different initial conditions

❑ Both approaches yield the same general solution, but

➢ Classical solution cannot be separated into components arising from
physically meaningful conditions → useless in system theory.
 Special case: zero-state response of
DT LTI systems to unit impulse input
𝛿𝑘
System ℎ𝑘

𝑘
H[.]
𝑘

❑ 𝑦 𝑘 =? when 𝑥 𝑘 = 𝛿 𝑘 & zero initial conditions

➢ Solution of the equation 𝑄 𝐸 𝑦 𝑘 = 𝑃 𝐸 𝛿 𝑘 with zero initial conditions.
➢ Unlike the continuous time case, the DT solution can be computed
iteratively → not a closed form solution.

❑ Closed form solution:

ℎ[𝑘] = 𝐴0 𝛿[𝑘] + 𝑦𝑛 𝑘 𝑢[𝑘].
➢ 𝑦𝑛 [𝑘] is a linear combination of the characteristic modes of the system,
𝑢[𝑘] is the (DT) unit step signal.
𝑏0
➢ 𝐴0 = & 𝑛 unknown coefficients in 𝑦𝑛 𝑘 are determined from 𝑛 initial
𝑎0
values of ℎ[𝑘], which, in turn, are obtained iteratively.
➢ ℎ[𝑘] is called the impulse response of the DT system.
 Example
Find the unit impulse response (iteratively and closed form) of a DT
system described by 𝑦 𝑘 − 0.6𝑦 𝑘 − 1 − 0.16𝑦 𝑘 − 2 = 5𝑥[𝑘].
Solution
➢ Iterative method:
➢ ℎ 0 − 0.6ℎ −1 − 0.16ℎ −2 = 5𝛿 0 ⟹ ℎ 0 = 5,
➢ ℎ 1 − 0.6ℎ 0 − 0.16ℎ −1 = 5𝛿 1 ⟹ ℎ 1 = 3,
➢ ℎ 2 − 0.6ℎ 1 − 0.16ℎ 0 = 5𝛿 2 ⟹ ℎ 2 = 2.6,
➢ and so on

➢ Closed form: 𝑄 𝛾 = 𝛾 2 − 0.6𝛾 − 0.16 = 0 ⟹ 𝛾1 = −0.2, 𝛾2 = 0.8.

➢ 𝑦𝑛 𝑘 = 𝑐1 ሺ−0.2ሻ𝑘 +𝑐2 ሺ0.8ሻ𝑘
0 𝑘 𝑘
➢ 𝐴0 = ⟹ ℎ 𝑘 = 𝑐1 −0.2 + 𝑐2 0.8 𝑢[𝑘].
−0.16
𝑐1 + 𝑐2 = 5
➢ Since ℎ 0 = 5 and ℎ 1 = 3, we have ቊ ⟹ 𝑐1 = 1, 𝑐2 = 4.
−0.2𝑐1 + 0.8𝑐2 = 3
𝑘 𝑘
➢ ℎ𝑘 = −0.2 + 4 0.8 𝑢[𝑘]
 Zero-state response of CT LTI systems

𝑥 −1 ℎ 𝑘 + 1
𝑥[−1]𝛿 𝑘 + 1
Linear, time-invariant

𝑥[0]𝛿 𝑘 𝑥0ℎ𝑘

Linear, time-invariant

𝑥[1]𝛿 𝑘 − 1 𝑥 1 ℎ 𝑘−1

Linear, time-invariant

𝑥[2]𝛿 𝑘 − 2 𝑥 2 ℎ 𝑘−2

Linear, time-invariant

❑ The operator that performs the superposition of the scaled,

shifted impulse responses to obtain 𝑦[𝑘] is called convolution.
 Convolution sum derivation
𝑥𝑘 y𝑘

𝑥 𝑘 ∗ℎ 𝑘 =𝑦 𝑘

❑ Weighted & time shifted impulses

𝛿[𝑘] → ℎ[𝑘]
𝛿[𝑘 − 𝑚] → ℎ[𝑘 − 𝑚]
𝑥[𝑚]𝛿[𝑘 − 𝑚] → 𝑥[𝑚]ℎ[𝑘 − 𝑚]
𝑖𝑛𝑝𝑢𝑡 𝑠𝑎𝑚𝑝𝑙𝑒 #𝑚 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑡𝑜 𝑖𝑛𝑝𝑢𝑡 #𝑚

❑ Superposition:
∞ ∞

෍ 𝑥[𝑚]𝛿[𝑘 − 𝑚] → ෍ 𝑥 𝑚 ℎ[𝑘 − 𝑚]
𝑚=−∞ 𝑚=−∞
𝑥[𝑘] 𝑦[𝑘]

❑ Convolution: 𝑦[𝑘] = σ∞
𝑚=−∞ 𝑥 𝑚 ℎ[𝑘 − 𝑚].
❖ Note: do not be fooled by confusing but conventional notation
▪ 𝑦 𝑘 = 𝑥[𝑘] ∗ ℎ 𝑘 looks like an operation of samples; but it is not!
▪ Convolution operates on signals not samples → should be 𝑦 𝑘 = ሺ𝑥 ∗
ℎሻ[𝑘], i.e. convolving signal 𝑥 with signal ℎ generates a new signal 𝑥 ∗ ℎ.
 Convolution sum evaluation
➢ Step 1

➢ Step 2 – 4 for 𝑘 < 0

❑ Steps in convolution operation 𝑘<0

1. Fold/Reflect 𝑔[𝑘] → 𝑔[−𝑚].

2. Shift/Slide 𝑔[−𝑚] by 𝒌 (a constant
inside the sum) time units.
➢ Step 2 – 4 for 𝑘 ≥ 0
3. Find 𝑝𝑘 [𝑚] = 𝑓[𝑚]𝑔[𝑘 − 𝑚].
𝑘≥0
4. Sum 𝑝𝑘 [𝑚] over 𝒎.
5. Repeat 2 → 4 for all values of 𝑘.

❑ Results:
➢ Step 5
➢ No overlap for 𝑘 < 0 → 𝑐 𝑘 = 0.
➢ 𝑘 ≥ 0: 𝑐 𝑘 = σ𝑘𝑚=0 0.8 𝑚
0.3 𝑘−𝑚

𝑘+1 𝑘+1
➢ 𝑐 𝑘 = 2 0.8 − 0.3 𝑢[𝑘]
 Impulse response - a descriptor of DT
LTI systems
❑ Given a system’s impulse response, we can determine the
system’s output for any input by performing the convolution.
❑ Both impulse response and difference equation are used to
describe the input-output relationship of CT LTI systems, but:
➢ The impulse response description applies only to systems that are initially
at rest → can determine the system’s zero−state response only.
➢ Difference equation representation is more flexible: can apply to a broader
class of systems, which are either at rest or with non-zero initial conditions.

❑ Relationship between the responses from each description:

𝑛

Total response = ෍ 𝑐𝑗 𝛾𝑗𝑘 + ℎ[𝑘] ∗ 𝑥[𝑘]

𝑗=1 zero−state response
zero−input response

❑ Both descriptions are declarative (tell rules that the system

obeys) & do NOT provide insight into the system’s internal
mechanism to produce the output from the input.
 Properties of the convolution sum
(proofs: homework)
❑ Commutative:
𝑓1 [𝑘] ∗ 𝑓2 [𝑘] = 𝑓2 [𝑘] ∗ 𝑓1 [𝑘]
❑ Distributive:
𝑓1 𝑘 ∗ {𝑓2 [𝑘] + 𝑓3 [𝑘]} = 𝑓1 [𝑘] ∗ 𝑓2 [𝑘] + 𝑓1 [𝑘] ∗ 𝑓3 [𝑘]
❑ Associative:
𝑓1 𝑘 ∗ 𝑓2 𝑘 ∗ 𝑓3 𝑘 = {𝑓1 [𝑘] ∗ 𝑓2 [𝑘]} ∗ 𝑓3 [𝑘]
❑ The shift property: if 𝑓1 [𝑘] ∗ 𝑓2 [𝑘] = 𝑐[𝑘] then
𝑓1 [𝑘] ∗ 𝑓2 [𝑘 − 𝑚] = 𝑐[𝑘 − 𝑚], and
𝑓1 [𝑘 − 𝑚] ∗ 𝑓2 [𝑘] = 𝑐[𝑘 − 𝑚], and
𝑓1 [𝑘 − 𝑚] ∗ 𝑓2 [𝑘 − 𝑛] = 𝑐[𝑘 − 𝑚 − 𝑛].

❑ Convolution with an impulse

𝑓[𝑘] ∗ 𝛿[𝑘] = 𝑓[𝑘]
 Impulse response as an indicator of
system’s properties
❑ The output of a DT LTI system: 𝑦[𝑘] = σ∞
𝑚=−∞ 𝑓 𝑚 ℎ[𝑘 − 𝑚]
➢ Memoryless:
ℎ[𝑘] = 𝑐𝛿[𝑘], for c an arbitrary constant
➢ Causal:
ℎ[𝑘] = 0 for all 𝑘 < 0.
➢ Stable: ℎ[𝑘]is absolutely summable, i.e.
σ∞
𝑘=−∞ ℎ[𝑘] < ∞.

❑ Invertibility and deconvolution

➢ A system is invertible if the input can be recovered from the output except
for a constant scale factor ↔ there exists an inverse system (e.g. equalizer
in communication modems).
➢ Deconvolution: the process of recovering 𝑥 𝑘 from ℎ[𝑘] ∗ 𝑥[𝑘]→ an
inverse system performs deconvolution.
➢ The impulse response of the inverse system must satisfy
ℎ[𝑘] ∗ ℎinv [𝑘] = 𝛿[𝑘]

❑ Homework: prove all of the above properties.

 Block diagram representation
❑ The block diagram provides a more detailed representation of
a system.
➢ It describes how the system’s internal computations are ordered via
interconnections of elementary operations that act on the input signal.

➢ The block diagram description of a system is not unique.

❑ Example: find a block diagram representation for the system

described by the following differential equation
1 1
𝑦 𝑛 + 𝑦 𝑛 − 1 + 𝑦 𝑛 − 2 = 𝑥 𝑛 + 𝑥[𝑛 − 1]
4 8
 Example solution
1 1
❑ Direct form I: 𝑦 𝑛 = 𝑥 𝑛 + 𝑥 𝑛 − 1 + − 𝑦 𝑛 − 1 − 𝑦 𝑛 − 2
4 8

𝑆1 : 𝑤 𝑛 = 𝑣 𝑛 + 𝑣[𝑛 − 1]

1 1
𝑆2 : 𝑝 𝑛 = − 𝑝 𝑛 − 1 − 𝑝 𝑛 − 2 + 𝑞[𝑛]
4 8
❑ Direct form II: commutative of form I, i.e. 𝑆2 precedes 𝑆1 → save 1 shifter.
 State variable description of DT LTI
systems (1)

❑ The state of a system is a minimal set of signals that represents the

system’s entire memory of the past.
➢ Each signal in the selected set is called a state variables.
➢ The selection of the state of a system is not unique (simple choice: the
outputs of the shifters).

❑ The state variable description is obtained by writing down 1st-order

difference equations corresponding to the operations in the diagram.
➢ The state equations describe the dynamics of the state
➢ The output equation relates the output to the state and the input

❑ Example: find the state variable description for the above system
 State variable description of DT LTI
systems (2)

❑ Select state variables: 𝑞1 [𝑛], 𝑞2 [𝑛] ❑ State variable description is often

➢ The inputs to the shifters are 𝑞1 [𝑛 + written in matrix form as follows
1], 𝑞2 [𝑛 + 2] 𝒒[𝑛 + 1] = 𝑨𝒒[𝑛] + 𝒃𝑥[𝑛]
❑ Write down the state equations 𝑦[𝑛] = 𝒄𝒒[𝑛] + 𝐷𝑥[𝑛]
1 ❑ The state equation in matrix form
𝑞1 𝑛 + 1 = − 𝑞1 [𝑛] + 𝑥[𝑛] 1
2 − 0 𝑞 [𝑛]
1 𝑞1 𝑛 + 1 2 1 1
𝑞2 𝑛 + 1 = 𝑞1 𝑛 + 𝑞2 [𝑛] + 3𝑥[𝑛] = + 𝑥[𝑛]
3 𝑞2 𝑛 + 1 1 𝑞2 [𝑛] 3
1
❑ Write down the output equation 3
❑ The output equation in matrix form
𝑦[𝑛] = 𝑞2 [𝑛] + 2𝑥[𝑛] 𝑞1 𝑡
𝑦 𝑡 = 0 1 + [2]𝑥[𝑛]
𝑞2 𝑡
 Transformation of the state
❑ The transformation of basis in the state space, specified by a
matrix T, is called similarity transformation.
➢ The similarity transformation T changes 𝒒[𝑛], 𝑨, 𝒃, 𝒄, 𝐷 into 𝒒′[𝑛], 𝑨′, 𝒃′, 𝒄′,
𝐷′ but does not change the input-output characteristics of the system.
➢ With 𝒒′ [𝑛] = 𝑻𝒒[𝑛], it is straightforward to show that
𝑨′ = 𝑻𝑨𝑻−1 , 𝒃′ = 𝑻𝒃,
𝒄′ = 𝒄𝑻−1 𝐷′ = 𝐷,
then the new description of the system is
𝒒′[𝑛 + 1] = 𝑨′ 𝒒′ [𝑛] + 𝒃′𝑥[𝑛]
𝑦[𝑛] = 𝒄′ 𝒒′ [𝑛] + 𝐷′ 𝑥[𝑛].

❑ Example: a DT LTI system has the following state space

description:
−2 0 1
𝑨= , 𝒃= ,
1 −1 1
𝒄= 0 2, 𝐷 = 1.
Find the state space description 𝑨′, 𝒃′, 𝒄′, 𝐷′ corresponding to the new states
𝑞1′ [𝑛] = 2𝑞1 [𝑛] + 𝑞2 [𝑛], and 𝑞2′ 𝑛 = 𝑞1 𝑛 − 𝑞2 [𝑛].
 Example solution
2 1
❑ The similarity transformation is 𝑻 = , therefore
1 −1
𝟏 −4 −1
➢ 𝑨′ = 𝑻𝑨𝑻−1 =
𝟑 −2 −5
3
➢ 𝒃′ = 𝑻𝒃 =
0
𝟏
➢ 𝒄′ = 𝒄𝑻−1 = 2 −4
𝟑
➢ 𝐷′ = 𝐷 = 1