ELT2035 Signals & Systems

Lecture 10: Laplace transform

Hoang Gia Hung

Faculty of Electronics and Telecommunications
University of Engineering and Technology, VNU Hanoi
 Introduction
Time Domain 𝛿(𝑡) ℎ(𝑡)
H[.]
Frequency Domain 𝑒 𝑗𝜔𝑡 𝐻(𝑗𝜔)𝑒 𝑗𝜔𝑡

❑ Drawbacks of Fourier transform

➢ Does not exist for signals that are not absolutely integrable;
➢ Cannot be easily adapted for unstable or even marginally stable systems.

❑ Example: consider the growing exponential, e.g. 𝑓 𝑡 = 𝑒 2𝑡 𝑢(𝑡)

➢ This signal is not Fourier transformable → is there a generalization for FT?

➢ Yes, find of another type of eigensignal rather than 𝑒 𝑗𝜔𝑡 .
 Idea of the Laplace transform

?
𝒆−𝝈𝒕 𝒆𝝈𝒕

Fourier transformable

➢ Multiplying 𝑓 𝑡 with 𝐞−𝝈𝐭 with 𝜎 > 2 gives 𝜙 𝑡 = 𝑓(𝑡)𝑒 −𝜎𝑡 , whose Fourier
spectrum is continuous from -∞ to +∞, i.e. 𝜙 𝑡 = summing of (infinite) sinusoids
➢ 𝑓 𝑡 = 𝜙 𝑡 𝒆𝝈𝒕 → 𝑓 𝑡 = summing of 𝑒 𝜎𝑡 times each spectral component of 𝜙 𝑡 ,
thus a component of frequency 𝑗𝜔 in the spectrum of 𝜙 𝑡 is transformed into a
component of frequency 𝜎 + 𝑗𝜔 in the spectrum of 𝑓 𝑡 .
➢ 𝑠 = 𝜎 + 𝑗𝜔 is called the complex frequency, 𝑒𝑠𝑡 is an eigensignal of LTI system
 Bilateral Laplace transform
❑ Consider the Fourier transform of 𝑓(𝑡)𝑒 −𝜎𝑡 (𝜎 real)
∞ ∞
➢ ℱ 𝑓(𝑡)𝑒 −𝜎𝑡 = ‫׬‬−∞ 𝑓(𝑡)𝑒 −𝜎𝑡 𝑒 −𝑗𝜔𝑡 𝑑𝑡 = ‫׬‬−∞ 𝑓(𝑡) 𝑒 −(𝜎+𝑗𝜔)𝑡 𝑑𝑡 = 𝐹(𝜎 + 𝑗𝜔)
1 ∞
➢ The inverse Fourier transform is 𝑓 𝑡 𝑒 −𝜎𝑡 = ‫𝜎(𝐹 ׬‬ + 𝑗𝜔)𝑒 𝑗𝜔𝑡 𝑑𝜔.
2𝜋 −∞
1 ∞
Multiplying both sides with 𝑒 𝜎𝑡 yields 𝑓 𝑡 = ‫𝜎(𝐹 ׬‬ + 𝑗𝜔)𝑒 (𝜎+𝑗𝜔)𝑡 𝑑𝜔.
2𝜋 −∞
➢ Changing 𝜎 + 𝑗𝜔 to 𝑠 in the above integral → bilateral Laplace transform.

❑ Bilateral (two-sided) Laplace transform pair

∞
𝐹 𝑠 = ‫׬‬−∞ 𝑓(𝑡)𝑒 −𝑠𝑡 𝑑𝑡 (1)
1 𝜎+𝑗∞ 𝑠𝑡 𝑑𝑠
𝑓 𝑡 = ‫׬‬ 𝐹(𝑠)𝑒 (2)
2𝜋𝑗 𝜎−𝑗∞
ℒ
➢ Notations: 𝑓 𝑡 𝐹(𝑠) or ℒ 𝑓 𝑡 = 𝐹(𝑠), ℒ −1 𝐹(𝑠) = 𝑓(𝑡).
➢ In practice, we usually do not evaluate (2) directly; instead, we determine
ℒ −1 𝐹(𝑠) by compiling a table of Laplace transforms for a variety of signals.
➢ FT is a special case of LT when we limit 𝑠 to the imaginary axis, i.e. 𝑠 = 𝑗𝜔.
❖ If interested in the nonnegative side only → Unilateral LT (details later).
 Region of convergence (ROC)
❑ The ROC for 𝐹(𝑠) is a set of values of 𝑠 (a region in the complex
∞
plane) for which the integral ‫׬‬−∞ 𝑓(𝑡)𝑒 −𝑠𝑡 𝑑𝑡 in Eq. (1) converges.
➢ Consider 𝑓 𝑡 = 𝑒 2𝑡 𝑢(𝑡): there are infinitely choices for 𝜎 to make 𝜙 𝑡 =
𝑓(𝑡)𝑒 −𝜎𝑡 Fourier transformable; in particular, 𝜎 ≥ 2 → the ROC is the complex
plane on the right hand side of the line 𝜎 = 2.
𝑗𝜔
➢ The ROC is also required for evaluating the
inverse Laplace transform: the path of
integration, which is along 𝜎 + 𝑗𝜔 with 𝜔 varying
from -∞ to +∞, must lie in the ROC for 𝐹(𝑠).
𝜎
➢ It is possible that two different signals have the 2

same 𝐹(𝑠) but with different ROC, i.e. there is

no 1-to-1 correspondence between 𝐹(𝑠) and
𝑓(𝑡) unless the ROC is specified.

❑ Example:
➢ Find the Laplace transform and ROC for 𝑥 𝑡 =
𝑒 𝑎𝑡 𝑢(𝑡) & 𝑦 𝑡 = −𝑒 𝑎𝑡 𝑢(−𝑡), assuming 𝑎 is real.
 Example solution
∞
❑ 𝑋 𝑠 = ‫׬‬−∞ 𝑒 𝑎𝑡 𝑢(𝑡)𝑒 −𝑠𝑡 𝑑𝑡 = 𝑗𝜔

∞ −1 −(𝑠−𝑎)𝑡 ∞
‫׬‬0 𝑒 −(𝑠−𝑎)𝑡 𝑑𝑡 =
𝑠−𝑎
𝑒 ቚ
0
➢ Since 𝑠 is complex, the term 𝑒 −(𝑠−𝑎)𝑡 does
𝜎
not necessary vanish when 𝑡 → ∞.
➢ lim 𝑒 −(𝑠−𝑎)𝑡 = lim 𝑒 −(𝜎−𝑎)𝑡 𝑒 −𝑗𝜔𝑡 =
𝑡→∞ 𝑡→∞
0, 𝜎 > 𝑎 1
ቊ ⟹𝑋 𝑠 = , 𝑅𝑂𝐶 = Re(𝑠) > 𝑎.
∞, 𝜎 ≤ 𝑎 𝑠−𝑎
𝑗𝜔
∞ 1
❑ 𝑌 𝑠 = ‫׬‬−∞ −𝑒 𝑎𝑡 𝑢(−𝑡)𝑒 −𝑠𝑡 𝑑𝑡 =
𝑠−𝑎
,
𝑅𝑂𝐶 = Re(𝑠) < 𝑎.
𝜎
 Properties of the Laplace transform
ℒ ℒ
❑ Linearity: assume 𝑥 𝑡 𝑋(𝑠) with ROC 𝑅𝑥, 𝑦 𝑡 𝑌(𝑠) with
ℒ
ROC 𝑅𝑦, then 𝑎𝑥 𝑡 + 𝑏𝑦 𝑡 ՞ 𝑎𝑋 𝑠 + 𝑏𝑌(𝑠) with ROC at least
𝑅𝑥 ∩ 𝑅𝑦.
ℒ 1 𝑠
❑ Scaling: 𝑥 𝑎𝑡 𝑋 with ROC = scaled 𝑅𝑥 (i.e. 𝑠 is in the
𝑎 𝑎
𝑠
ROC if 𝑎 is in 𝑅𝑥).
ℒ
❑ Time shifting: 𝑥 𝑡 − 𝜏 𝑒 −𝑠𝜏 𝑋(𝑠) with ROC = 𝑅𝑥.
ℒ
𝑠0 𝑡
❑ s-domain shifting: 𝑒 𝑥 𝑡 𝑋(𝑠−𝑠0 ) with ROC = shifted 𝑅𝑥 (i.e.
𝑠 is in the ROC if 𝑠 − 𝑠0 is in 𝑅𝑥).
ℒ 𝑑
❑ Differentiation in the s-domain: −𝑡𝑥 𝑡 𝑋 𝑠 with ROC = 𝑅𝑥.
𝑑𝑠

𝑑 ℒ
❑ Differentiation in time domain: 𝑥 𝑡 𝑠𝑋 𝑠 with ROC = at
𝑑𝑡
least 𝑅𝑥 .
 Properties of the Laplace transform
(cont.)
𝑡 ℒ 1
❑ Integration in time domain: ‫׬‬−∞ 𝑥 𝜏 𝑑𝜏 𝑋 𝑠 with ROC =
𝑠
at least 𝑅𝑥 ∩ {𝑅𝑒(𝑠) > 0}.
❑ Initial value theorem: If 𝑥(𝑡) = 0 for 𝑡 < 0 and 𝑥(𝑡) contains no
impulses or higher-order singularities at 𝑡 = 0, then
𝑥 0+ = lim 𝑠𝑋(𝑠) .
𝑠→∞

❑ Final value theorem: If 𝑥(𝑡) = 0 for 𝑡 < 0 and 𝑥(𝑡) has a finite
limit as 𝑡 → ∞, then
lim 𝑥 𝑡 = lim 𝑠𝑋(𝑠) .
𝑡→∞ 𝑠→0
Laplace Transforms of Elementary
Functions
Laplace Transforms of Elementary
Functions (cont.)
 Partial fraction expansion
❑ Find the inverse Laplace transform of
𝐵(𝑠) 𝑏𝑀 𝑠 𝑀 + 𝑏𝑀−1 𝑠 𝑀−1 + ⋯ + 𝑏1 𝑠 + 𝑏0
𝑋 𝑠 = = 𝑁
𝐴(𝑠) 𝑠 + 𝑎𝑁−1 𝑠 𝑁−1 + ⋯ + 𝑎1 𝑠 + 𝑎0
➢ Roots of 𝑩(𝒔) and 𝑨(𝒔) are called the zeros and poles of 𝑋(𝑠), respectively.
➢ Without loss of generality, suppose 𝑁 ≥ 𝑀 (if 𝑀 ≥ 𝑁 we can use long division
෨
𝐵(𝑠)
to express 𝑋 𝑠 = σ𝑀−𝑁
𝑘=0 𝑘𝑐 𝑠 𝑘
+ ሚ
, where the order 𝐴(𝑠) ෨
≥ that of 𝐵(𝑠)).
෨
𝐴(𝑠)

❑ Factoring the denominator polynomial as a product of pole

𝑏𝑀 𝑠 𝑀 +𝑏𝑀−1 𝑠 𝑀−1 +⋯+𝑏1 𝑠+𝑏0
factors 𝑋 𝑠 = ς𝑁
𝑘=1 𝑠−𝑑𝑘
➢ If all the poles 𝑑𝑘 are distinct, 𝑋(𝑠) can be expressed as a sum of simple
𝐴𝑘
terms 𝑋 𝑠 = σ𝑁 𝑘=1 where 𝐴𝑘 = 𝑠 − 𝑑𝑘 𝑋(𝑠)ȁ𝑠=𝑑𝑘
𝑠−𝑑𝑘

➢ If a pole 𝑑𝑖 is repeated 𝑟 times, there are 𝑟 terms in the partial fraction

𝐴 𝐴 𝐴
expansion associated with that pole: 𝑖1 , 𝑖2 2 , ⋯ , 𝑖𝑟 𝑟 , where 𝐴𝑖𝑘 =
𝑠−𝑑𝑖 𝑠−𝑑𝑖 𝑠−𝑑𝑖
1 𝑑 𝑟−𝑘
𝑠 − 𝑑𝑖 𝑟 𝑋(𝑠)ฬ .
𝑟−𝑘 ! 𝑑𝑠 𝑟−𝑘 𝑠=𝑑𝑘
 Partial fraction expansion (cont.)
❑ From partial fraction expansion, we can easily obtain the inverse
Laplace transform for each term as:
−1 1 𝑒 𝛼𝑡 𝑢 𝑡 , 𝜎>𝛼
ℒ = ൝ 𝛼𝑡 ,
𝑠−𝛼 −𝑒 𝑢(−𝑡), 𝜎 < 𝛼
𝑡 𝑛−1 𝛼𝑡
1
𝑒 𝑢 𝑡 , 𝜎>𝛼
𝑛−1 !
ℒ −1 = ൞ 𝑡 𝑛−1 .
𝑠−𝛼 𝑛
− 𝑒 𝛼𝑡 𝑢(−𝑡), 𝜎<𝛼
𝑛−1 !
➢ Due to linearity property, i.e. the ROC of 𝑋(𝑠) is at least the intersection of
individual ROCs, the ROC of each term must be correctly inferred from the
ROC of 𝑋 𝑠 .
➢ E.g. if the ROC of 𝑋(𝑠) lies to the left of a particular pole, we pick the left-
sided inverse Laplace transform for that pole and vice versa.

❑ Example:
2𝑠 3 −9𝑠 2 +4𝑠+10
➢ Find the inverse bilateral Laplace transform of 𝑋 𝑠 = , with
𝑠 2 −3𝑠−4
ROC Re(𝑠) < −1.
 Example solution
3𝑠−2
❑ Use partial fraction expansion: 𝑋 𝑠 = 2𝑠 − 3 + = 2𝑠 −
𝑠 2 −3𝑠−4
3𝑠−2 𝐴 𝐵
3+ = 2𝑠 − 3 + + , for all 𝑠: Re(𝑠) < −1.
𝑠−4 𝑠+1 𝑠+1 𝑠−4
3𝑠−2 3𝑠−2 1 2
➢ 𝐴= ቚ = 1, 𝐵 = ቚ = 2 → 𝑋 𝑠 = 2𝑠 − 3 + + .
𝑠−4 𝑠=−1 𝑠+1 𝑠=4 𝑠+1 𝑠−4

➢ Since the ROC of 𝑋(𝑠), {𝑠 ∈ 𝐶: Re(𝑠) < −1}, is to the left of both poles → we
choose left-sided inverse Laplace transform for all terms.
1
➢ Final answer: 𝑥 𝑡 = 2𝛿 𝑡 − 3𝛿 𝑡 − 𝑒 −𝑡 𝑢 −𝑡 − 2𝑒 4𝑡 𝑢 −𝑡 .

𝑗𝜔

𝜎
 Properties of the ROC
❑ The ROC cannot contain any pole.
➢ If 𝑑 is a pole → 𝑋(𝑑) = ±∞ → 𝑋(𝑠) does not converge at 𝑑 → 𝑑 ∉ ROC.

❑ The ROC depends only on the real part of 𝑠.

➢ The ROC consists of strips parallel to the imaginary axis in the 𝑠-plane

❑ The ROC for a finite-duration signal includes the entire s-plane.

❑ ℒ[𝑥 𝑡 ] converges if and only if 𝑥(𝑡) is an exponential-order
signal: ∃𝐴 > 0, 𝜎𝑝 , 𝜎𝑛 : 𝑥 𝑡 ≤ 𝐴𝑒 𝜎𝑝𝑡 , 𝑡 > 0; 𝑥 𝑡 ≤ 𝐴𝑒 𝜎𝑛 𝑡 , 𝑡 < 0.
➢ 𝑥 𝑡 must grow no faster than 𝑒 𝜎𝑝 𝑡 for positive 𝑡 and 𝑒 𝜎𝑛 𝑡 for negative 𝑡.
➢ If 𝑥(𝑡) is a exponential left-sided signal (i.e. 𝑥(𝑡) = 0 for 𝑡 > 𝑏, 𝑏 is an arbitrary
constant), then the ROC of 𝑥(𝑡) is 𝜎 < 𝜎𝑛.
➢ If 𝑥(𝑡) is a exponential right-sided signal (i.e. 𝑥(𝑡) = 0 for 𝑡 < 𝑎, 𝑎 is an
arbitrary constant), then the ROC of 𝑥(𝑡) is 𝜎 > 𝜎𝑝.
➢ If 𝑥(𝑡) is a exponential two-sided signal (i.e. 𝑥(𝑡) infinitely extends in both
directions), then the ROC of 𝑥(𝑡) is 𝜎𝑛 > 𝜎 > 𝜎𝑝.
Illustration of the relationship between the
time extent of a signal and the ROC
𝑗𝜔

𝜎
𝜎𝑛

𝑗𝜔

𝜎
𝜎𝑝

𝑗𝜔

𝜎
𝜎𝑝 𝜎𝑛
 Unilateral Laplace transform
❑ There are many applications where the signals involved are
causal
➢ It is advantageous to define the unilateral Laplace transform that works only
on the non-negative time portion of the signal → no need to consider the
ROC.

❑ Definition
∞
𝐹 𝑠 = ‫׬‬0− 𝑓(𝑡)𝑒 −𝑠𝑡 𝑑𝑡 (3)
➢ The lower limit 0− in the integral implies that discontinuities and impulses at 0
are included.
➢ The unilateral and bilateral Laplace transforms are equivalent for causal
signals, thus the expression for the inverse Laplace transform (2) remains
unchanged.
➢ The unilateral Laplace transform cannot be used for noncausal signals.
ℒ𝑢
➢ Notation: 𝑓 𝑡 𝐹(𝑠)
 Properties of the unilateral Laplace
transform
❑ Similar to those of the bilateral Laplace transform while the ROC
consideration is not needed.
➢ Only a minor condition is needed for time shifting to guarantee causality of
the shifted signals:
ℒ𝑢
𝑥 𝑡−𝜏 𝑒 −𝑠𝜏 𝑋 𝑠 for all 𝜏 such that 𝑥 𝑡 − 𝜏 𝑢 𝑡 = 𝑥 𝑡 − 𝜏 𝑢(𝑡 − 𝜏).
 Application of unilateral LT to solving
differential equations with initial conditions
❑ Consider the following ODE, assuming zero input for 𝑡 < 0
𝑁 𝑑𝑘 𝑀 𝑑𝑘
σ𝑘=0 𝑎𝑘 𝑘 𝑦 𝑡 = σ𝑘=0 𝑏𝑘 𝑘 𝑥 𝑡 .
𝑑𝑡 𝑑𝑡
➢ Taking unilateral Laplace transform on both sides and applying the
𝑑𝑛 ℒ𝑢
generalized differentiation in time property, 𝑥 𝑡 𝑠𝑛𝑋 𝑠 −
𝑑𝑡 𝑛
𝑑 𝑛−1 𝑑 𝑛−2 𝑛−1 −
𝑥 𝑡 ฬ − 𝑠 𝑛−2 𝑥 𝑡 ฬ − ⋯− 𝑠 𝑥 0 , yields:
𝑑𝑡 𝑛−1 𝑡=0− 𝑑𝑡 𝑡=0−
𝐴(𝑠)𝑌(𝑠)– 𝐶(𝑠) = 𝐵(𝑠)𝑋(𝑠) ,
where 𝐴 𝑠 = 𝑠 𝑁 + 𝑎𝑁−1 𝑠 𝑁−1 + ⋯ + 𝑎1 𝑠 + 𝑎𝑜 , 𝐵 𝑠 = 𝑏𝑀 𝑠 𝑀 + 𝑏𝑀−1 𝑠 𝑀−1 +
𝑑𝑙
⋯ + 𝑏1 𝑠 + 𝑏𝑜 , and 𝐶 𝑠 = σ𝑁 𝑘−1
𝑘=1 σ𝑙=0 𝑎𝑘 𝑠
𝑘−1−𝑙
𝑦(𝑡)ฬ .
𝑑𝑡 𝑙 𝑡=0−

𝐵 𝑠 𝑋(𝑠) 𝐶(𝑠)
➢ Hence: 𝑌 𝑠 = + (4)
𝐴(𝑠) ต
𝐴(𝑠)
𝑌 𝑓 (𝑠) 𝑌 𝑛 (𝑠)

❑ Solution:
➢ Solving Eq. (4) for 𝑌(𝑠), then 𝑦(𝑡) is obtained via the inverse LT.
➢ The effects of the initial conditions on 𝑦(𝑡) and the input are separated.
Example: Finding The Forced And Natural
Responses
❑ Use the unilateral Laplace transform to determine the output of a
system represented by the differential equation:
𝑑2 𝑑 𝑑
𝑦 𝑡 +5 𝑦 𝑡 + 6𝑦 𝑡 = 𝑥 𝑡 + 6𝑥 𝑡
𝑑𝑡 2 𝑑𝑡 𝑑𝑡
in response to the input 𝑥(𝑡) = 𝑢(𝑡). Assume that the initial
𝑑
conditions on the system are 𝑦 0− = 1 & 𝑦 𝑡 ቚ − = 2.
𝑑𝑡 𝑡=0

❑ Solution:
➢ Using the differentiation property and the unilateral Laplace transform of both
sides of the differential equation, we obtain
𝑑
𝑠 2 + 5𝑠 + 6 𝑌 𝑠 − 𝑦 𝑡 ቚ − 𝑠𝑦 0− − 5𝑦 0− = 𝑠 + 6 𝑋(𝑠)
𝑑𝑡 𝑡=0−
➢ Solving for 𝑌(𝑠), we get
𝑑
𝑠+6 𝑋(𝑠) 𝑦 𝑡 ቚ +𝑠𝑦 0− +5𝑦 0−
𝑑𝑡 𝑡=0−
𝑌 𝑠 = +
𝑠 2 +5𝑠+6 𝑠 2 +5𝑠+6
𝑓
➢ The first term is associated with the forced response of the system, 𝑌 (𝑠).
The second term corresponds to the natural response, 𝑌 𝑛 (𝑠).
 Example solution (cont.)
❑ Solution (cont.):
➢ Using 𝑋(𝑠) = 1/𝑠 and the initial conditions, we obtain
𝑓 𝑠+6 𝑛 𝑠+7
𝑌 𝑠 = and 𝑌 𝑠 =
𝑠(𝑠+2)(𝑠+3) (𝑠+2)(𝑠+3)

➢ Partial fraction expansions of both terms yields

𝑓 1 −2 1 𝑛 5 −4
𝑌 𝑠 = + + and 𝑌 𝑠 = +
𝑠 𝑠+2 𝑠+3 𝑠+2 𝑠+3
𝑓 𝑛
➢ Next, taking the inverse unilateral Laplace transform of 𝑌 (𝑠) and 𝑌 (𝑠),
we obtain
𝑓
𝑦 𝑡 = 𝑢 𝑡 − 2𝑒 −2𝑡 𝑢 𝑡 + 𝑒 −3𝑡 𝑢 𝑡 and 𝑦 𝑛
𝑡 = 5𝑒 −2𝑡 𝑢 𝑡 − 4𝑒 −3𝑡 𝑢 𝑡
➢ The output of the system is
𝑓 𝑛
𝑦 𝑡 =𝑦 𝑡 +𝑦 𝑡 = 𝑢 𝑡 + 3𝑒 −2𝑡 𝑢 𝑡 − 3𝑒 −3𝑡 𝑢 𝑡
Example solution: Graphical illustration
 Transfer function
❑ LTI system: 𝑦(𝑡) = ℎ(𝑡) ∗ 𝑥(𝑡) (assuming zero initial conditions).
➢ ℎ(𝑡) = the impulse response of the system.
➢ This eqn. applies regardless of whether 𝑥(𝑡) & ℎ(𝑡) are causal or noncausal.

❑ Take the bilateral Laplace transform and apply the convolution

property yields 𝑌(𝑠) = 𝐻(𝑠)𝑋(𝑠).
𝑌(𝑠)
➢ 𝐻 𝑠 = and is called the transfer function of the system.
𝑋(𝑠)
ℒ 𝑌(𝑠) ℒ
➢ ℎ 𝑡 𝐻(𝑠) = , 𝑦 𝑡 𝐻 𝑠 𝑋(𝑠)
𝑋(𝑠)

❑ Suppose the system is described by the differential equation:

𝑁 𝑑𝑘 𝑀 𝑑𝑘
σ𝑘=0 𝑎𝑘 𝑘 𝑦 𝑡 = σ𝑘=0 𝑏𝑘 𝑘 𝑥 𝑡
𝑑𝑡 𝑑𝑡
➢ Taking bilateral Laplace transform, assuming zero initial conditions, yields
σ𝑁 𝑘 𝑀 𝑘
𝑘=0 𝑎𝑘 𝑠 𝑌 𝑠 = σ𝑘=0 𝑏𝑘 𝑠 𝑋 𝑠 .

𝑌(𝑠) σ𝑀
𝑘=0 𝑏𝑘 𝑠
𝑘
➢ 𝐻 𝑠 = = σ𝑁 𝑘
𝑋(𝑠) 𝑘=0 𝑎𝑘 𝑠
 Transfer function examples
❑ Find the transfer function of the LTI system described by the
𝑑2 𝑑 𝑑
differential equation 𝑑𝑡 2
𝑦 𝑡 + 3 𝑑𝑡 𝑦 𝑡 + 2𝑦 𝑡 = 2 𝑑𝑡 𝑥 𝑡 − 3𝑥 𝑡 .
2𝑠−3
❑ Solution: 𝐻 𝑠 = (assuming zero initial conditions).
𝑠 2 +3𝑠+2

❑ The transfer function can also be expressed in pole-zero form

by factoring the numerator and denominator polynomials as
follows
𝑏෨ ς𝑀
𝑘=1 𝑠 − 𝑐𝑘
𝐻 𝑠 = 𝑁
ς𝑘=1 𝑠 − 𝑑𝑘
➢ Knowledge of the poles, 𝑑𝑘 , the zeros, 𝑐𝑘 , and 𝑏෨ = 𝑏𝑀 Τ𝑎𝑁 completely
determines the transfer function 𝐻 𝑠 → another description of LTI systems.
➢ Example: Find a differential equation description of the system described by
2(𝑠+1)(𝑠−1)
the following transfer function 𝐻 𝑠 =
𝑠(𝑠+2)(𝑠+1)
𝑑3 𝑑2 𝑑 𝑑2
➢ Solution: 𝑦 𝑡 + 3 2𝑦 𝑡 + 2 𝑦 𝑡 = 2 2𝑥 𝑡 − 2𝑥 𝑡
𝑑𝑡 3 𝑑𝑡 𝑑𝑡 𝑑𝑡
 Stability
❑ BIBO, or equivalently, absolutely integrable impulse response.
➢ Since ℎ 𝑡 = ℒ −1 𝐻(𝑠) , the ROC of 𝑯(𝒔) determines ℎ 𝑡 , and thus stability
of the system. The ROC of a stable system must contain the 𝑗𝜔-axis (why?).
➢ Differential equation description does not contain ROC information → must
have additional knowledge of the system to determine stability

❑ If the system is causal: ℎ 𝑡 = 0 ∀𝑡 < 0 → ℎ 𝑡 is the right-sided

inverse Laplace transform of the transfer function.
𝐴𝑘 ℒ
➢ 𝐴𝑘 𝑒 𝑑𝑘 𝑡 𝑢(𝑡) → the system is stable if all the poles are on the left half
𝑠−𝑑𝑘
of the complex plane. If there’s at least one pole on the RHP → unstable.

❑ If the system is anti-causal: ℎ 𝑡 is the left-sided iLT of 𝐻 𝑠 .

𝐴𝑘 ℒ
➢ −𝐴𝑘 𝑒 𝑑𝑘𝑡 𝑢(−𝑡) → the system is stable if all the poles are on the right
𝑠−𝑑𝑘
half of the complex plane. If there’s at least one pole on the LHP → unstable.

❑ What if there are poles on the imaginary axis?

➢ Unstable if the 𝑗𝜔-axis poles are repeated; marginally stable if unrepeated
(assuming all other poles are on the LHP).
Stable systems: pole positions &
impulse response
𝑗𝜔

𝑗𝜔

𝜎
Causal systems: pole positions &
impulse response
𝑗𝜔

𝑗𝜔

𝜎
 Example: Inverse Laplace Transform
with Stability and Causality
2 1
❑ A system has the transfer function 𝐻 𝑠 = + . Find the
𝑠+3 𝑠−2
impulse response, (a) assuming that the system is stable and
(b) assuming that the system is causal. Can this system be both
stable and causal?
❑ Solution:
➢ This system has poles at 𝑠 = −3 and 𝑠 = 2.
a. If the system is stable, then the pole at 𝑠 = −3 contributes a right-sided term
to the impulse response, while the pole at 𝑠 = 2 contributes a left-sided
term. We thus have: ℎ 𝑡 = 2𝑒 −3𝑡 𝑢 𝑡 − 𝑒 2𝑡 𝑢 −𝑡 .
b. If the system is causal, then both poles must contribute right-sided terms to
the impulse response, and we have ℎ 𝑡 = 2𝑒 −3𝑡 𝑢 𝑡 + 𝑒 2𝑡 𝑢 𝑡 . This system
is not stable, since the term 𝑒 2𝑡 𝑢 𝑡 is not absolutely integrable.
➢ The system cannot be both stable and causal because the pole at 𝑠 = 2 is in
the right half of the s-plane.
 Application to system reduction

𝐻1 𝑠

𝐻1 𝑠 𝐻2 𝑠 +
𝐻2 𝑠
𝐻 𝑠 = 𝐻1 𝑠 𝐻2 𝑠
𝐻 𝑠 = 𝐻1 𝑠 + 𝐻2 𝑠

- 𝐻1 𝑠 + 𝐻1 𝑠

𝐻2 𝑠 𝐻2 𝑠

𝐻1 𝑠 𝐻1 𝑠
𝐻 𝑠 = 𝐻 𝑠 =
1 + 𝐻1 𝑠 𝐻2 𝑠 1 − 𝐻1 𝑠 𝐻2 𝑠
 Summary
❑ Bilateral Laplace transform is a generalization of FT in that real
frequency 𝜔 is generalized to a complex frequency 𝑠 = 𝜎 + 𝑗𝜔.
➢ Similar properties.
➢ Each LT must be specified with an ROC (region of convergence).
➢ Inverse Laplace transform is usually NOT evaluated directly; instead, by
inverse transforms of the terms of a partial fraction expansion.

❑ Unilateral Laplace transform that works only on the non-

negative time portion of the signal
➢ Cannot applied to non-causal signals, no need to consider ROC
➢ Application: solving differential equations with initial conditions

❑ The transfer function of a CT-LTI system is the ratio of the output

to the input in the 𝑠 −domain where all initial conditions are 0.
➢ Another representation of CT-LTI system
➢ System characteristics are reflected by positions of the poles and zeros
➢ Can be used to simplify interconnected systems