Chapter 7: Analytic Solution of Linear Dynamic Systems

• So far, we have covered the following topics:

– Modeling dynamic systems
– Standard forms for systems (SSR, TFs, etc.)

• Next, we will determine the dynamic response of linear systems using analytical techniques (i.e., the total
solution of linear ODEs)
– In particular, we will investigate the analytical system response to “standard” inputs such as the step, ramp, and
impulse

• We will focus on solving “simple” systems (1st- and 2nd-order ODEs) “by hand” using analytic methods

• Finally, we will establish performance criteria for system response

– Time to reach steady state
– Overshoot and damping, etc.
 Goals of Linear System Analysis

• System designers should understand the basic parameters governing 1st- and 2nd-order
system response
– How does the “resistance” parameter effect time to reach the steady-state response?

– How does the “stiffness” parameter effect oscillation frequency?

• Often the response of a complex, nonlinear model “looks like” a standard 1 st- or 2nd-order
response, and therefore the complex system can be approximated and replaced by a
reduced-order model
– System designers can trade-off model complexity vs. model accuracy, as well as obtain a physical,
intuitive feel for dynamic system response
 7.2 Analytical Solutions to Linear Differential Equations

• Suppose we have an nth-order ODE (I/O equation) with one dynamic variable y, and one input function, u(t)

• As an example, consider the 2nd-order I/O equation (ODE)

• The total solution y(t) is comprised of two parts: the homogeneous (or natural) solution yH(t) and the
Total solution
particular (or forced) solution yP(t)

• Homogeneous solution: the solution to the homogeneous ODE, where the input u(t) is zero. Therefore, the
homogeneous solution is the natural response, and is strictly due to the “natural dynamics”
 Homogeneous Solution
• Homogeneous solution: the solution to the homogeneous ODE, where the input u(t) is zero

• Therefore, solve the unforced ODE with zero input:

• Assume an exponential solution form for yH(t)

• Therefore, and

or,

Because cert cannot be zero for all time t, the bracket term must = 0
 Characteristic Equation
• The Characteristic Equation is the bracket term set to zero:

• The characteristic roots give us the complimentary solution:

• Finally, the homogeneous solution is

Note: in this case, yH(t) decays to zero

as time t  infinity (steady-state)

• The constants c1 and c2 are determined from the two initial conditions (ICs) after the particular response
has been determined
and
 Particular Solution (Forced Response)

• Particular solution: the solution to the ODE with a non-zero right-hand side forcing function

(input)

• The particular solution is usually found by guessing a functional form, which is typically related

to the input function u(t)

• For example, if a linear system is driven by a sinusoidal forcing function u(t ) sin 4t , then the

forced response will also be a sinusoidal function but with a different amplitude and/or phase

angle with respect to the input function, i.e.,

 Solving an LTI System: Example 7.1

• Given the following first-order linear ODE with initial condition y(0) = 3. Determine the
complete (total) response y(t).

4 y  8 y 6
(worked on the board )
 Solving an LTI System: Example 7.2

• Given the following second-order linear ODE with initial conditions y(0) = 2 and y (0)  1
Determine the complete solution y(t) of the system.

2 y  8 y  6 y 10 sin 4t

(worked on the board)

 The Complete Response
• Transient response: part of the total response that goes to zero as time t  infinity (or, the initial part of the
dynamic response that eventually “dies out”)

• Steady-state response: part of the total response that remains as time t  infinity

Transient Steady-state
response response

Complete response = Transient + Steady-state responses

 Characteristic Roots and the Transfer Function

• We have seen that the homogeneous or free response yH(t) depends on the roots of the
characteristic equation

• The roots of the characteristic equation can also be determined from the corresponding
system transfer function G(s) = b(s)/a(s)

• The values of s that make the denominator polynomial a(s) equal zero are called the poles
of the transfer function

• The transfer function poles are identical to the roots of the characteristic equation
 Characteristic Roots and the Transfer Function

• Example 7.3: 3-way valve with applied force f(t)

Characteristic equation:

 Two roots are

r1  200  j 367.42 r2  200  j 367.42

Transfer function Approach :  Poles of TF G(s) are the same as the

characteristic roots
 DC Gain and the Steady State
• An extremely useful analysis technique is the use of the DC gain to compute a system’s steady-state response

to a constant input.

• The term “DC gain” is derived from circuit analysis (DC implies a constant, non-oscillating input)

• Definition: the system DC gain is the steady-state gain to a constant input, for the case when the output has a

constant value at steady state

• The DC gain for a given transfer function is computed by simply setting s = 0

– This result arises from application of the Final Value Theorem in Laplace transform theory (Details in Chapter 8)

• Note that setting s = 0 is similar to setting D = 0 in our D-operator method; in other words, all time

derivatives are zero because the system has reached steady state (constant output)
 DC Gain: Example 7.4
• Consider a solenoid actuator that provides a force input to a mechanical system (spool valve)

• Suppose the input to the solenoid is a constant voltage, u = 2 V

• The DC gain of the Solenoid dynamics is simply

• Therefore, at steady state, the solenoid dynamics can be replaced by a constant gain = , and therefore the

steady-state EM force is fSS =

 DC Gain: Example 7.4 (2)
• We can treat the steady-state output of the Spool valve dynamics in the same manner

• The DC gain of the valve TF is simply

Therefore, the steady-state valve position is

ySS =

DC Gain of the entire system? Of course, the DC gain of the entire system =
 7.3 First-Order System Response
• Recall that many of the models we derived in the earlier chapters resulted in first-order systems

• First-order examples include mass-damper mechanical systems, RL or RC circuits (solenoids and filters),

single-chamber hydraulic systems, single-vessel pneumatic systems, or single-capacitance thermal systems

(example models are below):

Standard Form  y  y  Au (t )

 First-Order System: Standard Form
• All of the various physical models on the prior slide have the following “standard form”

y  y  Au (t )
• The constant τ and “dynamic variable” y is defined for all physical models as shown below

Physical system Dynamic variable y Constant 

Mechanical rotor  , angular velocity J/b
RC circuit eC , capacitor voltage RC
Pneumatic chamber P , chamber pressure RC
Thermal chamber T , chamber temperature RC

• Note that the right-hand side constant A is unity for all models except the mechanical rotor ( A = 1/b for the
mechanical rotor )
 First-Order System: Natural Response
• Recall that the natural response (homogeneous solution) is the response with ZERO input:

Standard form

• We can easily find the natural response by writing the characteristic equation for the standard 1st-order
model:
One characteristic root:

• Therefore, the natural (homogeneous) response is

Homogeneous solution: due to IC
only with ZERO input

• For the case with ZERO input, the constant c is simply the initial condition (IC) for the dynamic variable, c =
 First-Order System: Time Constant t
• Repeat the natural response (for zero input)

• Characteristics of Homogeneous Solution

• If root r < 0 (or, constant 𝜏 > 0), then the solution y(t) decays from its initial condition y0 to zero as time t → ∞
and the response is bounded or stable.
• If r > 0, then y(t) diverges to infinity as time t → ∞ and the response is unbounded or unstable.
• Constant 𝜏 is always positive for these physical systems
• Free response will be an exponential decay to zero at steady state.
• The parameter 𝜏 is called the time constant for the first-order system.
 First-Order System: Time Constant t
• Repeat the natural response (for zero input)
• Note that the 1st-order natural response has the following
characteristics:
– Exponential rise or decay to zero steady-state value (“dies out”)

– When time t = τ , the natural response has decayed to 36.8% of

its initial value x0 since e-1 =

– When time t = 4t, the natural response has essentially “died out”
since e-4 = which is “small” (less than 2% of initial value x0)

– Time constant 𝜏 determines the response time of the first-order

free response.

Definition: the “settling time” tS is the approx. time to reach S-S, For a 1st-order system, tS =
 First-Order System: Step Response
• Consider again the “standard form” of a 1st-order system:

• Suppose the input u(t) = U = step input (constant)

• The total response to the step input will be the natural + forced responses:

• We already know that the natural response yH(t) is an exponential function

• The forced (or steady-state) response to a constant input is also a constant:

Steady-state Steady-state response = constant = AU

 First-Order System: Step Response (2)
• Therefore, put together the natural and forced responses to get the total response of a 1 st-order system to a
step input:
Total response

• Next, apply the IC x(0) = y0 to obtain the constant c

• The total response of the 1st-order system (to a step input)

Transient Steady-state
response response
 Sketching the Step Response of 1st Order System

• Sketching a 1st-order step response is simple

– Re-write the 1st-order model in the “standard form” and compute the time constant, τ

– Estimate the settling time as four-time constants, i.e., tS =

– Compute the steady-state response for the constant input

– Sketch an exponential response from the initial condition y0 to the steady-state value yss

– The response approximately reaches yss at t = tS =

– The total response will “decay down” if y0 > yss , or show an “exponential rise” if yss > y0
 1st-Order System Step Response: Example 7.5

• Consider a solenoid actuator used to position a spool valve in a hydraulic actuation system

• The electrical input is voltage ein(t), and the output is force f (in N). An appropriate first-order model is

• Divide by 1.5 to write the first-order model in our standard form:


Clearly, the time constant is t = 0.002 sec, and the
settling time tS = 4t = 0.008 sec = 8 ms

• If the magnitude of the step input is ein(t) = 2 V, then the steady-state output will be fss =
 1st-Order System Step Response: Example 7.5 (2)
• The step response (from Simulink) is below. The response is clearly exponential, and the settling time is
indeed 8 ms, and the steady-state response is 16 N

Step response of solenoid actuator

(first-order system)
 First-Order System: Pulse Response
• Suppose the input is a pulse function:

 0 for t 0

u(t )  P for 0  t  T
 0 for t T


First-order system
(“standard form”)
 1st-Order Pulse Response: Case 1
• Case 1: Pulse time T is greater than settling time tS
T >> tS

The product of Pulse Looks like zero-input

Looks like step response (homogeneous) response
magnitude & DC gain of TF

Settling time
 1st-Order Pulse Response: Case 2
• Case 2: Pulse time T is less than settling time tS
T < tS

We can get the complete response

analytically by using superposition:
add 2 individual responses
(see textbook)

   
y t   Pb 1  e  t /  Pb 1  e  (t  T )/ U t  T 
 Impulse Response of First-Order System
• We can obtain the impulse response of a system by evaluating the pulse response in the limit as
pulse duration goes to zero
 Impulse Response of First-Order System (2)
• Let pulse duration T go to zero in the limit so that U(t – T) = U(t)

• Now take limit as T  0

Applying l’Hopital’s rule: Impulse response

 1st-Order System Impulse Response: Example 7.6
• Sketch the impulse response of an RL circuit

R = 1.2 W

L = 0.02 H

• The voltage input is an impulse (weight of 0.08 V-s) applied at t1 = 0.1 sec; the circuit has
zero energy at t = 0

RL circuit model: or

The time constant is identified as  =

 1st-Order System Impulse Response: Example 7.6 (2)

• The initial magnitude of the impulse response is

From

• The impulse response exponentially decays to zero in ~four-time constants after the impulse is
applied, or t =

Initial magnitude = 4 A

Discontinuous jump
in energy due to
impulsive voltage
input at t1 = 0.1 sec Exp decay to zero
at t = 0.1667 sec
 7.4 Second-Order System Response

• Many systems are modeled by second-order ODEs

– Mechanical (mass-spring-damper)

– Electrical (RLC circuit) with two storage elements

• Several complex, higher-order systems (like aircraft) have dominant modes that display a

basic second-order system response

• In many cases, we can predict the behavior of a second-order system based on several

key parameters (instead of solving for the exact time response)

7.4 Second-Order System Response
 Second-Order Systems: Roots
• Consider the 2nd-order I/O equation:

• The natural or free response (zero input) will depend on the roots of the characteristic equation

Characteristic Equation

• The two root locations are

Recall:

• Therefore, the two roots can be 1) real and distinct; 2) real and repeated (radicand = 0); 3) complex

conjugates (radicand < 0); 4) purely complex (coefficient a1 = 0 and a0 > 0)

 2nd-Order Root Locations: Case 1
• Both roots are real and distinct (not repeated)

Example: y  4 y  3 y u (t )

r 2  4r  3 0 Characteristic Equation

 Characteristic roots r1 = -1, r2 = -3

y H (t ) c1e  t  c2 e  3t Generally yH (t ) c1e r1t  c2e r2t

Therefore, the natural response (no input) is composed of two

exponential functions, with time constants 1 sec and 1/3 sec

Stable, exponential natural response (dies out to zero)

2nd-Order Roots: Case 1

Two real, distinct, NEGATIVE roots: yH (t ) c1e r1t  c2 e r2t

r1 < 0 , r2 < 0
 2nd-Order Root Locations: Case 1 (con’t)
• Both roots are real, one root is positive

Example: y  2 y  3 y u (t )

r 2  2r  3 0 Characteristic Equation

 Characteristic roots r1 = 1, r2 = -3

y H (t ) c1e t  c2 e  3t
unstable

Therefore, the natural response (no input) is composed of two

exponential functions; but et diverges to infinity (unstable)
 2nd-Order Roots: Case 1

Two real, distinct roots with one root > 0

r1 > 0 , r2 < 0
 2nd-Order Root Locations: Case 1 (con’t)
• Both roots are real, one root is zero, one root is negative

Example: y  3 y u (t )

r 2  3r 0 Characteristic Equation

 Characteristic roots r1 = 0, r2 = -3

y H (t ) c1e 0  c2 e  3t c1  c2 e  3t
Does not “die out”

Therefore, the natural response (no input) is composed of a constant

and one exponential function (marginally stable)
 2nd-Order Roots: Case 1 (con’t)

Two real roots, one negative and one root = zero

r1 = 0 , r2 < 0
 2nd-Order Root Locations: Case 2
• Two real repeated roots (both are negative for this example)

Example: y  6 y  9 y u (t )

r 2  6r  9 0 Characteristic Equation

 Characteristic roots r1 = -3, r2 = -3

y H (t ) c1te  3t  c2 e  3t In general yH (t ) c1te r1t  c2e r1t

2nd-Order Root Locations: Case 2

Two repeated NEGATIVE real roots yH (t ) c1te r1t  c2e r1t

r1 = r2 < 0
2nd-Order Root Locations: Case 2

Two repeated POSITIVE real roots yH (t ) c1te r1t  c2e r1t

r1 = r2 > 0
2nd-Order Root Locations: Case 2

Two repeated ZERO roots yH (t ) c1te r1t  c2e r1t

r1 = r2 = 0
 2nd-Order Root Locations: Case 3
• Roots are complex conjugates (Real + Imaginary parts)

Example: y  4 y  29 y u (t )
r 2  4r  29 0 Characteristic Equation

 Characteristic roots r1, 2  2  j 5


y H (t ) c1e  2 j 5 t  c2 e  2 j 5 t e  2t c1e j 5t  c2 e  j 5t 
Using Euler’s Thm: e j cos   j sin 

y H (t ) e  2t c3 cos 5t  c4 sin 5t 

Therefore, the natural response (no input) is a stable, damped sinusoidal function,
where the real part is the exp decay, and imaginary part is the frequency
2nd-Order Root Locations Case 3

Complex conjugate roots with NEGATIVE real part

r1, 2   j where α < 0

 2nd-Order Root Locations: Case 3 (con’t)
• Roots are complex conjugates, with a positive real part

Example: y  4 y  29 y u (t )
r 2  4r  29 0 Characteristic Equation

 Characteristic roots r1, 2 2  j 5


y H (t ) c1e 2 j 5 t  c2 e 2 j 5 t e 2t c1e j 5t  c2 e  j 5t 
Using Euler’s Thm: e j cos   j sin 

y H (t ) e 2t c1 cos 5t  c2 sin 5t 

unstable

Therefore, the natural response (no input) is an unstable sinusoidal function,

where the exp function grows to infinity as time t  infinity
2nd-Order Root Locations: Case 3 (con’t)

Complex conjugate roots with POSITIVE real part

r1, 2   j where α > 0

 2nd-Order Root Locations: Case 4
• Roots are complex conjugates, with a zero real part

Example: y  9 y u (t )
r 2  9 0 Characteristic Equation

 Characteristic roots r1, 2 j 3

y H (t ) c1e j 3t  c2 e  j 3t

Using Euler’s Thm: e j cos   j sin 

y H (t ) c3 cos 3t  c4 sin 3t

Therefore, the natural response (no input) is an undamped harmonic oscillator,

imaginary part is the frequency, and the sinusoidal response never “dies out”
2nd-Order Root Locations: Case 4

Two imaginary roots (real part is ZERO)

r1, 2 j
 Damping Ratio and Undamped Natural Frequency

• We can characterize the transient response of a second-order system by two basic parameters
– Damping ratio ζ
– Undamped natural frequency, ωn

• To show this, consider the generic 2nd-order system:

y  a1 y  a0 y b0u (t )
• The characteristic equation is

r 2  a1r  a0 0

 a1  a12  4a 0 If 4a0 > a12 then the roots are complex

Root locations: r (damped sinusoid)  Underdamped
2
 Critically Damped System

• 2nd-order root locations of 2nd-order system (repeated)

 a1  a12  4a 0 If 4a0 = a12 then the roots

r are real and repeated Critically Damped
2

If 4a0 < a12 then the roots are real and distinct
(exponential decay  no oscillations) Overdamped

• Therefore, the condition a12 = 4a0 is the critical transition point between a damped oscillating response and a
damped exponential response

If coefficient a1 2 a0 then the system is critically damped

 Damping Ratio
• We choose to define the damping ratio as the ratio of actual damping coefficient to its critically damped
value
a1
  “Damping ratio”
2 a0

• If  = 1, the system is critically damped (no oscillations in the natural or transient response)

• If 0 <  < 1, the system is underdamped (oscillations are present in the natural or transient response)

• If  > 1, the system is overdamped (no oscillations in the natural or transient response)
 Damping Ratio: Mass-Spring-Damper

• Next, consider the classic mass-spring-damper 2nd-order system:

mx  bx  kx  f (t )

• The characteristic equation is mr 2  br  k 0

 b  b 2  4mk
Root locations: r
2m

b
Damping ratio:   Ratio of actual damping (b) to critical damping
2 km
 Undamped Natural Frequency

• Consider the 2nd-order system (no friction, or  = 0)

x  a0 x b0u (t )
• The characteristic equation is

r 2  a0 0

r j a0 Natural response is undamped

Root locations:
harmonic oscillator,

c3 sin a0 t  c4 cos a0 t

Therefore, define the frequency “Undamped natural

of oscillation when no damping  n  a0 frequency”
is present: (rad/s)
 Undamped Natural Frequency: Mass-Spring-Damper

• Consider the mass-spring system (no damper, or no friction)

mx  kx  f (t )
• The characteristic equation is
mr 2  k 0

k Natural response is undamped

Root locations: r j
m harmonic oscillator, k k
c3 sin t  c4 cos t
m m

Therefore, define the frequency k “Undamped natural

of oscillation when no damping n  frequency”
is present: m (rad/s)
 2nd-Order System: Standard Form
• Like the 1st-order system, we can define a “standard form” for the 2 nd-order system:

mx  bx  kx  f (t ) y  2 n y  n2 y u (t )

Substitute for and n and we get the traditional mass-spring-damper

• For the undamped ()/underdamped (0<) case, the (complex) roots are

r  n  jn 1   2

• The natural (free) response is

y H (t ) ce   nt cosd t   
 Underdamped 2nd-Order System
• The natural (free) response is (repeated)

y H (t ) ce   nt cosd t    r  n  jn 1   2

2nd-order root locations

• Note that the real part of the complex roots determines the exponential decay rate of the transient
response

• Note that the imaginary part of the complex roots determines the frequency of oscillation of the
transient response

“Damped frequency”
d n 1   2
frequency of oscillation of
the transient response
 Underdamped System: Root Locations
• Root locations in the “complex plane”, where +x is the Real-number axis, and +y is the Imaginary-number axis

Imaginary
n , n 1   2

r  n  jn 1   2 X
2nd-order underdamped Complex Plane
root locations b c

θ
Right triangle: a2 + b2 = c2 a 0
Real
a = n Real part

b n 1   2 Imaginary part

X
 n
Angle: cos  
 c = n distance from origin to root n
Note: θ = 0  damping = 1
 Underdamped Root Locations (2)
Imaginary
Radial lines: X
All roots have X
the same damping
ratio near negative real Complex Plane
axis increase, and decreases
X
near imaginary axis

X X Horizontal lines: all roots have the

same damped frequency ωd

0 Real
X

X
Circles:
All roots have
the same undamped X Complex roots ALWAYS appear
natural frequency ωn, increasing as conjugate pairs  symmetric
radius increases ωn about the Real axis
 Step Response of an Underdamped 2nd Order System
• Performance criteria for the transient response of an underdamped second-order system are
often based on the step response

y  2 n y  n2 y  A = constant (step)

y t   yH t   yP t  1  e t  c3 cos  t  c4 sin  t  , Unit Step response of an underdamped

system
  n ;   n 1   2

 Derivation as home practice

c3  1; c4 

  
y t  1  e  cos  t  sin  t 
t Unit Step response of an underdamped
   system
 Step Response of an Underdamped 2nd Order System

  
y t  1  e  cos  t  sin  t 
t Unit Step response of an underdamped
   system

  n ;   n 1   2

For n 1 and  0.2, 0.4

 Step Response of an Underdamped 2nd Order System

• Performance equations for step response

– Period of Oscillation
2
Tperiod  where  d  n 1   2
d

– Peak Time: time to reach maximum output tp 
n 1  2
– Peak/Maximum output to unit step input

  
y t  1  e  cos  t  sin  t 
t
  

 / d   
ymax 1  e  cos   sin   Put   n , d n 1   2
ymax 1  e   / 1  2
 d 
Function of ζ only
 Step Response of an Underdamped 2nd Order System

• Performance equations for step response

– Maximum Overshoot Mos: difference between peak and steady state normalized by
steady state
ymax  yss ymax 1  e   / 1  2
M os  with
yss
• As steady state response to unit step input is 1, therefore

  / 1  2
M os e

• For general case

ymax  yss 1  M os 
 Step Response of an Underdamped 2nd Order System

• Performance equations for step response

– Settling time: Time to steady state e t e  4  t  4 / 
4
ts  = 4 time constants
 n
– Number of cycles during transient response = N ×Tperiod = tS N= tS / Tperiod

2 1  2
N cycles  Function of ζ only

 Step Response of an Underdamped 2nd Order System

The concepts and formulation discussed so far are for underdamped system only
 Therefore, first determine if the 2nd-order system is underdamped (check roots or ζ ) before applying these
transient-response equations!
 2nd-Order Underdamped System: Example 7.8

• Sketch the step response of the rotational mechanical system

0.2  1.6  65 Tin (t )

Step input is 2.5 N-m (torque)

System is initially at rest

Divide by J = 0.2 kg-m2 to obtain

standard form
  8  325 5Tin (t )
2n n2 ss = (5)(2.5)/325 = 0.0385 rad
 Undamped frequency n  325 18.0278 rad/s,
and  = 0.2219 (underdamped)
 2nd-Order Underdamped System: Example 7.8 (2)

• Use undamped natural frequency n and damping ratio  to compute all performance metrics
for underdamped step response


Peak time: tp  = 0.1787 sec
n 1   2

  / 1  2
Max overshoot: M OS e = 0.498 (48.9% overshoot)

4
Settling time: tS  = 1.0 sec
 n
2
Period: Tperiod  = 0.3574 sec
n 1   2
2nd-Order Underdamped System: Example 7.8 (3)

48.9% overshoot
Peak response θmax = 0.0573 rad
Peak time tp = 0.179 sec

Period = 0.3574 sec Steady-state

response qss = 0.0385 rad

Settling time ts = 1.0 sec

Ncycles = 2.8
 Estimating Damping Ratio: Log Decrement Method
• Usually, the damping coefficient b is the parameter most difficult to estimate

• Mass m and stiffness k can be measured from static tests, but measuring b requires a
dynamic test (system response)
– See seat-suspension system (ASME article, Chapter 11)

• The damping ratio 𝜁 for an underdamped system can be estimated from the peak values of
the transient response and the “amplitude envelope.

• We can perform a dynamic test where we measure the decaying sinusoidal y(t) (i.e.,
impulse response)
– Provide a sudden, short input (impulse)
 Estimating Damping Ratio: Log Decrement Method (2)

• Consider the impulse response, which would be measured from a dynamic test:

0.06
Period, T  Damped frequency
0.04
is easy to estimate:
2
0.02
d  rad/s
T
Impulse response

-0.02 Impulse response:

-0.04
x (t ) ce  nt sin d t
-0.06
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Time, sec
 Estimating Damping Ratio: Log Decrement Method (3)

• Now, form the ratio of the amplitudes between two successive peaks (note that the sine term is the same at
each peak):
0.06 x1
Ratio of peaks:
x2  n t1
0.04
x1 ce 1
  n ( t1 T )   nT enT
0.02 x2 ce e
Impulse response

-0.02

-0.04

-0.06
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Time, sec
 Estimating Damping Ratio: Log Decrement Method (4)

• Next, define the logarithmic decrement as the natural log of the ratio of successive amplitude peaks:

0.06 x1
Log decrement:
x2
x1
 
0.04

 ln ln e  nT  nT

0.02
x2
Impulse response

-0.02
Sub for period T using ωd  n 1   2

2 2
-0.04
T 
d n 1   2
-0.06
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
Time, sec
 Estimating Damping Ratio: Log Decrement Method (5)

• Subbing for T in the log decrement equation, we get

2
 
1  2
• Finally, we solve the above equation for damping ratio:


 
4 2   2

x1
Recall:  ln is from the impulse response (measured)
x2
 Estimating Damping Ratio: Log Decrement Method (6)

• We can obtain greater accuracy if we compute the ratio of peak amplitudes separated by n cycles

• Therefore, form the log of the peak ratio separated by n periods:

 x1   x1 x2 x3 xn   x1   x2   x3   xn 
ln   ln    ln    ln    ln      ln  
 xn 1   x2 x3 x4 xn 1   x2   x3   x4   xn 1 
       n

• Therefore, the log decrement is

1 x1 Ratio of peak values separated

  ln by n cycles
n x n 1
(Use  in the prior equation for
damping ratio  )
 7.5 Higher-Order Systems
• As an example, consider the unit-step response of a third-order LTI system:

y  2.5 y  38 y  18.5 y u (t )

• By inspection, the third-order characteristic equation is

r 3  2.5r 2  38r  18.5 0 or r  0.5r 2  2r  37 0

• The 3 roots are r1 = –0.5, r2,3  1  j 6

• Therefore, the homogeneous or free response will be the sum of an exponential function
(due to the real root at r1 = –0.5) and a damped sinusoidal function (due to complex
conjugate roots r2,3)

y H (t ) c1e  0.5t  c2e  t cos(6t   ) From a qualitative feel

“slowest” root: dies out in ~8 sec dies out in ~4 sec

 Higher-Order Systems (2)
• From this example we can see that any higher-order system always breaks down into first- and second-order
terms (including roots at r = 0 and roots on the Imaginary axis)

• There is no such thing as a “third-order” response!

• Therefore, if we can identify the transient response associated with each root (pole), we can get a qualitative
feel for the total transient response

• Furthermore, we can choose to eliminate the “fast” roots (or poles, or “modes”) from the system dynamics
and work with a simpler “reduced-order model”
 7.6 SSR and Eigenvalues
• We can derive the characteristic equation (and therefore, find the roots) using state-space representation
(SSR)

• To begin, consider the unforced (homogeneous) state equation

x  Ax
• Following our previous methods for solving ODEs, let’s assume the solutions are exponential functions of
time:

x1 (t ) c1e t
x2 (t ) c2e t
or, in compact
x(t ) ce t
 notation:

t
xn (t ) cn e
 SSR and Eigenvalues (2)
• The time-derivative of the assumed solution is
x (t ) ce t
• Sub this expression into the state equation

x  Ax 
t
ce  Ace t

• Move all terms to the left-hand side and factor out cet

I  Ace t
0

Because cet cannot be always I  A detI  A  0

be zero we get:
 SSR and Eigenvalues (3)
• Expanding the determinant provides an nth-order polynomial in   the characteristic
equation

I  A detI  A  0

• Work example (on board)

• The n values of that satisfy the determinant equation are called the eigenvalues of
matrix A

• Remember: the eigenvalues provide information about the natural or free response
 SSR and Eigenvalues (4)
• If a SISO system is represented by an I/O equation, it is easy to compute the
characteristic equation and solve for the n characteristic roots

• If the SISO system is represented by a transfer function G(s) it is easy to find the poles
(set denominator = 0)

• If the system is represented by a SSR, it is easy to compute the eigenvalues

( MATLAB: >> eig(A) )
I  A det I  A  0

Characteristic roots = poles of G(s) = eigenvalues

 7.7 Approximate Models

• Often times during the preliminary design stage, it is beneficial to use a simplified or approximate system

model instead of the complete (more complex) higher-order model

• For example, control system analysis and design can be performed using approximate subsystem models

without significant loss of accuracy

• The subsequent simulation run times will be greatly reduced by replacing higher-order, nonlinear, complex

models with simplified, linear, reduced-order models

• Often times a properly designed first- or second-order model can adequately represent the full complex

model
 Approximate Models (2)

• A schematic block diagram is shown below, where the complex system is replaced by an approximate

model. The block diagram shows an input/output relationship:

 Approximate Models: Example 7.11

• The figure shows the measured step

response of a solenoid actuator-valve system

• This response has been experimentally

obtained by measuring the valve

displacement that results from a 38-mA

current step input applied at t = 1 s

 Approximate Models: Example 7.11 (2)
• Note that the valve response displays a
very small amount of random “noise” or
“chatter” due to imprecise measurements in
the experimental setting
“noise” in measurements

• If possible, derive an approximate

solenoid-actuator model from the
experimental data

• To begin, note that the valve response is an

~exponential rise to S-S  First-order model: y  y au (t )
 Approximate Models: Example 7.11 (3)

• For our assumed first-order model y is position (in mm) and u(t) is input current (in mA)

y  y au (t )

• The experimental data shows that the steady-state position is ~0.44 mm for a 38 mA step
input; hence coefficient a = 0.44/38 = 0.0116

• The settling time (see plot on prior page) is tS ~ 1.2 sec, so the time constant is
sec
 t S / 4 0.3
• Hence an approximate 1st-order model is

0.0116
0.3 y  y 0.0116u (t ) or, transfer function G (s) 
0.3s  1
 Approximate Models: Example 7.11 (4)

• Use the approximate 1st-order model and apply a 38-mA step input:

• The step-response (via Simulink) is shown on the next slide…

Approximate Models: Example 7.11 (5)

Some error
in transient
response

Approximate first-order model shows

“good” match with experimental data
(predicts steady-state output and
settling time)
 Approximate Models Via System Identification

Details from MATLAB Documentation

 Chapter 7: Summary
• The complete solution of an ODE is composed of the homogeneous (“free”) and particular
(“forced”) responses, where the free response is dictated by the characteristic roots, and the
forced response depends on the nature of the input function
• First- and second-order systems were studied in depth
– First-order system response depends on the time constant 

– Underdamped second-order system transient response depends on damping ratio and undamped
natural frequency n

• The characteristic roots are computed from the characteristic equation, which can be derived from
the system’s I/O equation, transfer function, or state-space representation
– The location of the roots in the complex plane determines the system’s response speed and damping
characteristics