Professional Documents
Culture Documents
HANDOUT 2.
Dr Anil Kokaram
Electronic and Electrical Engineering Dept.
anil.kokaram@tcd.ie www.mee.tcd.ie∼ack
SYSTEM RESPONSE
• Ultimate goal is to control a process or analyze a signal To do this need
to be able to predict the response of a system to various inputs
• Given known system dynamics (i.e. we know enough about the system
to write down ODE’s describing its operation) we can use the Laplace
Transform to work out the response of the system to any input whose
Laplace Transform can be found.
• However, real inputs are not predictable; and it may not be possible to
write down a system’s ODE’s. The process used in Mobile Phones for
canceling echoes is a good example of a system which operates despite
the inability to analytically define the geometry/dynamics of the phone’s
immediate environment.
• Still can identify some basic system characteristics which allow handling
of system response. This is possible using purely time domain analysis.
• Time domain analysis relies on the fact that the response of a system to an
impulse or to a step tells you everything about the system (in principle).
These responses can be calculated if the ODE’s for the system can be
written, or they can be measured by experiment beforehand or on-line
(e.g. mobile phones, noise cancellation etc.)
• We will explore both techniques for analyzing systems. Start with Laplace
analysis as you met this in 2nd year.
even though you don’t know what value s is. Don’t panic, in this
course we are going to assume that the Laplace integral always
converges (i.e. Re(s) > 0) and so
· −st ¸¯
e ¯
− ¯ =0 (4)
s ¯s=∞
BUT YOU NEED TO REMEMBER THAT CONVERGENCE
IS AN ISSUE
4. In addition, assume that e−(s+a)t is 0 for s = ∞, (i.e. you can
always assume that s + a > 0 to make this happen.)
5. Remember to spot the ‘shift’ theorem since it gives a slick way of
using tables to get Laplace transforms of functions (we’ll derive
this later for the Fourier transform as well).
µ ¶
Given L f (t) = F(s)
µ ¶
Then L e−ktf (t) = F(s + k)
δ(t) impulse 1
1
u(t) unit step s
n!
tn sn+1
1
e−at s+a
ω0
sin(ω0 t) s2 +ω02
s
cos(ωo t) s2 +ω02
ω0
sinh(ω0 t) s2 −ω02
s
cosh(ω0 t) s2 −ω02
A(s+a)+Bω0
e−at [A cos(ω0 t) + B sin(ω0 t)] (s+a)2 +ω02
½ ¾ Z ∞
−at
L e sin(βt + φ) = e−at sin(βt + φ)e−st dt
0
1 ¡ jx ¢
Using sin(x) = e − e−jx
2j
Z ∞ µ ¶
1
= e−at ej(βt+φ) − e−j(βt+φ) e−st dt
0 2j
Z ∞ µ ¶
1 −(s+a)t j(βt+φ) −j(βt+φ)
= e e −e dt
2j 0
Z µ ¶
1 ∞ jφ −(s+a−jβ)t
= e e − e−jφ e−(s+a+jβ)t dt
2j 0
µ · −(s+a−jβ)t ¸∞ · −(s+a+jβ)t ¸∞ ¶
1 e e
= ejφ − e−jφ
2j −(s + a − jβ) 0 −(s + a + jβ) 0
µ ¶
1 ejφ e−jφ
= −
2j (s + a − jβ) (s + a + jβ)
µ ¶
1 (s + a + jβ)ejφ − (s + a − jβ)e−jφ
=
2j (s + a)2 + β 2
µ ¶
1 (s + a)(ejφ − e−jφ ) + jβ(ejφ − e−jφ )
=
2j (s + a)2 + β 2
(s + a) sin(φ) β cos(φ)
= +
(s + a)2 + β 2 (s + a)2 + β 2
You will need to find the inverse Laplace Transform of sometimes complicates
expressions. The method of partial fractions allows you to do this by express-
ing a complicated fraction in terms of a sum of simpler fractions. The idea
is that the inverse laplace transform of the simpler fractions is easier to spot.
Basic idea
k3 s + 1 A B C
= + + (1)
s(s + k1)(s + k2) s s + k1 s + k2
Expand r.h.s. and equate coeffs
A(s + k1)(s + k2) + Bs(s + k2) + Cs(s + k1)
=
s(s + k1)(s + k2)
Equating coeffs in s2 ⇒0 = A + B + C
Equating coeffs in s ⇒k3 = A(k1 + k2) + Bk2 + Ck1
Equating constants ⇒1 = Ak1k2
Then solve simultaneous equations for A, B, C. Urrgh, could be a
pain. There’s another, easier way (cover up rule)
x(t)
C y(t)
• Steps are
2.1 Example 1 :
Impulse response of this system is the response y(t) when x(t) = δ(t).
R
x(t)
C y(t)
Impulse response of this system is the response y(t) when x(t) = u(t)
(i.e. x(t) is a step function.
R
x(t)
C y(t)
1
⇒ Y(s) = X(s) (7)
1 + RCs
1
Find Laplace Transform of input x(t) : L{u(t)} = s
0.9
0.8
0.7
Amplitude (volts)
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (secs)
1
0.9
0.8
0.7
Amplitude (volts)
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (secs)
4 TRANSFER FUNCTIONS
F (s) is called the transfer function of the system. Its all we need
to work out everything about the system if the initial conditions were
0. The Transfer function generalizes the idea of ‘Gain’ to dynamic
attributes of the system.
OTHER TERMS
+
(INITIAL CONDITIONS)
R
x(t) C y(t)
R2
R1
A
-
x(t) + y(t)
R
A
-
x(t) + y(t)
-
-
+ +
w(t) z(t)
x(t) y(t)
Assuming that each stage does not load the preceeding one (i.e. each
input impedance >> output impedance of previous stage), Then
X(s) Y(s) 1 z(s)
= −k1; = ; = −k2
w(s) X(s) 1 + sT Y(s)
SIGNALS SYSTEMS
1 ↔ δ(t) Y(s) = 1 · X(s) ↔ y(t) = x(t)
1
Rt
s ↔ u(t) Y(s) = 1s · X(s) ↔ y(t) = 0 x(τ )dτ
1 −at 1
s+a ↔ e Y(s) = s+a · X(s) ↔ ẏ(t) + ay(t) = x(t)
ω ω
s2 +ω 2
↔ sin(ωt) Y(s) = s2 +ω 2
· X(s) ↔ ÿ(t) + ω 2y(t) = ωx(t)
−sT −sT
e ↔ δ(t − T ) Y(s) = e · X(s) ↔ y(t) = x(t − T )
It just turns out that the Laplace transform of some signals have a
form which occurs also as a Transfer function of particular systems.
Suppose
n(s)
G(s) = (11)
d(s)
Then the roots of n(s) are called the ZEROS of the system G(s)
And the roots of d(s) are called the POLES of the system G(s)
EXAMPLE:
4s2 − 8s − 60
G(s) = 3
s + 2s2 + 2s
4(s + 3)(s − 5)
=
s(s + 1 + j)(s + 1 − j)
Zeros of G(s) are the values of s that make the numerator = 0 hence
How does θ(t) (the rotational angle of the axle) depend on e(t) (the voltage
applied to the field winding)? The following system equations apply:
Kirchoff
∂i
e(t) = Ri(t) + L (3)
∂t
θ(s)
1. What is the the system transfer function E(s) ?
Assume θ(0) = 0, θ̇(t) = 0, i(0) = 0; First have a think about what you
expect . . .
From eqn 1
From eqn. 2
˙
⇒ τ (s) − B[sΘ(s) − θ(0)] = J[s2 Θ(s) − sθ(0) − (θ)(0)]
From eqn. 3
So
Θ(s) K 1
=
E(s) R(1 + sTf ) Bs[1 + Tm s]
Therefore, the time domain impulse response h(t) is just L−1 (H(s))
Remember: System transfer function is Laplace transform of the
impulse response, or the impulse response is the inverse Laplace
Transform of the system transfer function. So
½ ¾
K/(RB)
h(t) = L−1 (11)
s(1 + sTf )(1 + Tm s)
Need to use partial fractions to express the transfer function in a
simpler form to make taking the inverse easier (i.e. to use tables).
K/(RB) A B C
= + +
s(1 + sTf )(1 + Tm s) s (1 + sTf ) (1 + sTm )
Using “Cover Up” rule
K/(RB) K
A= =
(1 + 0 × Tf )(1 + Tm × 0) RB
¯
K/(RB) ¯¯
B=
s(1 + Tm s) ¯ 1
s=− T
f
K/(RB)
=
− T1 (1 − TTm )
f f
−KTf /(RB)
= )
(1 − TTm
f
−KTf2 /(RB)
=
(Tf − Tm )
¯
K/(RB) ¯¯ K/(RB)
C= =
s(1 + sTf ) ¯s=− 1 − T1m (1 −
Tf
Tm )
Tm
So now we can use tables to find the inverse Laplace Transform, hence
· Ã ! Ã !¸
K 1 Tf 1 Tm 1
H(s) = − 1 +
RB s (Tf − Tm ) s + Tf Tf − Tm s + T1m
• Step response g(t) (say) : can get this either i) integrate the impulse
response OR ii) set e(t) = u(t) and then calculate output. You can spot
that the only terms that involve t in h(t) have simple exponentials, so
integration seems straightforward
Z t
g(t) = h(τ )dτ
0
·Z t µ ³ ´ ¶ ¸
K Tf − T1 τ T
e ( Tm ) dτ
m 1
− τ
= u(τ ) − e f +
RB 0 (Tf − Tm ) Tf − Tm
R 1
remember e−t/a = −(1/a) e
−t/a
= −ae−t/a
µ ¶
· Tf2 ¸t
K − τ
Tf −1 Tm2 −( Tτ −1)
= τ+ e − e m
RB (Tf − Tm) Tf − Tm 0
· ³ ´ ¸
K Tf2 − Tt −1 T 2
e−( Tm −1)
t
m
= t+ e f −
RB (Tf − Tm ) Tf − Tm