Signals & SYSTEMS RECAP:

SIGMALS
IMPULSE RESPONSE
 Signals
SIGNALS
A signal is a quantity that varies with time.
 SYSTEMS
Systems are objects or processes that transform signals.

x(t) y(t)
system
 SYSTEMS
Anything can be a system, if you can think of suitable inputs and
outputs to describe it that way.

resistor
car robot
motor chihuahua
x(t) y(t)
quadcopter

chemical plant avocado

circuit guy covered in bees
 SYSTEMS
In control engineering, “the system” could refer to:

+
part of the thing
the thing we’re trying
to control the combination of the
thing and the controller
 SYSTEMS
A system is an input-output relationship between signals.

x(t) y(t)
input output
 SYSTEMS
Multiple-input, multiple-output (MIMO) system:

x1(t) y1(t)
x2(t) y2(t)
x3(t) y3(t)
 SYSTEMS
Single-input, single-output (SISO) system:

x(t) y(t)
 SYSTEMS
We can think of a signal as a
mathematical function.

x(t) y(t)
 SYSTEMS
…But we won’t
always know what
the function is.

x(t) y(t)
 SYSTEM ID.
System identification = finding
a model that can calculate the
response to any input.

x(t) y(t)

F(x)?
 SYSTEM ID.
If we know something about the
system, e.g. that it contains a spring
with a constant of 5, we can use our
knowledge of dynamics to work out
the system model. k=5

x(t) y(t)
displacement force

y = -5x This is called white box

modelling.
 SYSTEM ID.
If we don’t know anything about the contents of
the system, we can identify the model through
experiments, by applying known inputs.

This is called black box modelling.

y = -5x
 SYSTEM ID.
BLACK BOX method =
determining the system model
through experiments.

y = -5x
greyish SYSTEM ID. Even when we can work out the
general form of the system
White Box
model, there will probably be
method = some unknown parameters (e.g.
determining the system the spring constant, here) that
we have to find through
model based on known k=? experiments.

components.
and experiments

y = -kx
 LTI Systems
Linear and Time-Invariant (LTI) systems: only need to know
the response to one signal to fully define the system.

LTI
system
If a system is LTI, we only need to know the
response to one input (SPOILER: it’s the
impulse) to calculate the response to any input.
 LTI Systems
homogeneity: response to scaled input
= response scaled by same factor

Linear
system

ax(t) → ay(t)
 LTI Systems
Additivity: response to sum
= sum of responses

Linear
system

x1 (t) + x2 (t) → y1 (t) + y2 (t)

 LTI Systems
Linear systems preserve
weighted sums.

Linear
system

ax1 (t) + bx2 (t) → ay1 (t) + by2 (t)

 LTI Systems
Time invariance: time-shifting the input causes the
same shift in the response.

T.I.
system

x(t − λ) → y(t − λ)
 LTI Systems
We can approximate any input using a sum of shifted, scaled pulses.

∞
= ෍ x λ p(t−λ)
λ=−∞
 LTI Systems
…so, if we know the response to the pulse, we can approximate the
response based on the LTI properties.

∞ ∞
= ෍ x λ p(t−λ)
Linear = ෍ x λ g(t−λ)
λ=−∞ system λ=−∞
 LTI Systems
Narrower pulses spaced more closely together give a more accurate
approximation.

∞
= ෍ x λ p(t−λ)
λ=−∞
 LTI Systems
Narrower pulses spaced more closely together give a more accurate
approximation.

∞
1 1
= ෍ x λ p(t− λ)
2 2
λ=−∞
 LTI Systems
Narrower pulses spaced more closely together give a more accurate
approximation.

∞
1 1
= ෍ x λ p(t− λ)
4 4
λ=−∞
 LTI Systems
Narrower pulses spaced more closely together give a more accurate
approximation.

∞
1 1
= ෍ x λ p(t− λ)
8 8
λ=−∞
 LTI Systems
Narrower pulses spaced more closely together give a more accurate
approximation.
…so we’ll get a perfect replica
of the signal if we use an
infinitesimally narrow pulse.
∞
1 1
= ෍ x λ p(t− λ)
∞ ∞
λ=−∞
 IMPULSE
The Dirac delta (δ) or impulse function is an infinitesimally narrow
pulse.

∞
= න x λ δ t−λ dλ
−∞
height =∞

area = 1
 IMPULSE
…so if we know the impulse response h(t),
we can calculate any output.

∞
= න x λ δ t−λ dλ
LTI ∞
= න x λ h t−λ dλ
−∞ −∞
system
 IMPULSE
…so if we know the impulse response h(t),
we can calculate any output.

∞ ∞
= න x λ δ t−λ dλ
−∞ h(t) = න x λ h t−λ dλ
−∞

The impulse response therefore

completely defines the system
model for an LTI system.
The operation we use to calculate the output from the
input and impulse response is called convolution.

…But this is a computationally-intensive process, so we

prefer to avoid it if possible.