You are on page 1of 27

4 Damped Harmonic Motion and Resonance

Damped harmonic motion is when the amplitude of the periodic motion of an oscillator (e.g., a
mass on a spring) decreases because of some kind of damping force which opposes velocity
(e.g., air resistance). When disturbed (i.e., moved away from equilibrium and/or given a
velocity), an oscillator responds harmonically at its natural frequency (unless the damping is too
high), dissipating energy away due to the damping and eventually returning to rest at its
equilibrium position.

Check out this link for a damped harmonic motion simulator for a mass on a spring:
https://www.mathwarehouse.com/harmonic-motion/interactive-damped-oscillator.php

And check out this Efficient Engineer video for a great introduction to the topic:
https://youtu.be/vLaFAKnaRJU

Resonance describes what happens when an external periodic force is applied to a system at a
frequency equal to (or very close to) the "natural frequency" of the system (the frequency it
would oscillate at naturally if excited with an impulse), causing an increased amplitude of
vibration in the system compared to what this same force would produce at other frequencies.
Here's a video showing 3 equal masses but supported using structures with different lengths (and
therefore different spring constants):
Resonance: A Perfect Experiment https://www.youtube.com/watch?v=iyw4AcZuj5k
Notice that as the oscillation frequency of the platform gradually increases each mass has a
frequency where it responds a lot. If you look closely, you'll see 3 regimes which happen at
different frequencies for each mass:
1. Low Frequency (well below resonance): the mass responds in phase and with the same
amplitude as the platform
2. Resonance: the mass responds with a much bigger amplitude and 90° behind the platform
in phase
3. High Frequency (well above resonance): the mass responds with further and further
decreasing amplitude

Though we'll talk mostly about masses on springs, the math and language we'll develop for
oscillators and resonance are broader concepts that apply to many systems, including resistor-
inductor-capacitor circuits, a magnetic pendulum over an aluminum plate, musical instruments,
earthquake protection for buildings, audio amplifiers and filters, and piezoelectric quartz tuning
forks for keeping time in electronic systems to name a few.

4.1 Damped Linear Oscillator Systems

In one dimension, a mass on a spring subject to linear damping and no other forces has a net
force on it given by:

where k is the spring constant, x is the mass' displacement from the spring's equilibrium position,
b is the damping constant, and v is its velocity. Applying Newton's second law:

Rearranging and recognizing that gives the equation of motion:

This is a linear homogeneous second order differential equation. In Math 2Z03, you learn that to
solve this you can substitute an assumed solution form , leading to:

The relative size of the damping determines the form of the solutions.

Case 1 - Overdamped:

When we have two real roots, both negative: .


The general solution is a linear combination of two decaying exponential functions:
Case 2 - Critically damped:

When we have one repeated root of .

A second solution comes from reduction of order, and the general solution is:

Case 3 - Underdamped:
When we have a complex conjugate pair of roots:

, leading to two solutions:

where is the damped natural frequency, so called because it

shows how the damping constant b has modified the natural frequency the system
would otherwise have. Note that if is very close to then can be much lower than
. In most underdamped resonant systems so that .

These solutions are inconvenient because they're complex, but because the DE is linear, we know
that any linear combination of solutions is also a solution, and it turns out the following linear
combinations let us replace these solutions with more-convenient sinusoids:

(you can prove this using Euler's identity, ).

So, the general solution we use for underdamped is:

4.1.1 Analytically exploring the damped response:


Examine the difference damping makes by finding the response of overdamped, critically
damped, and underdamped mass-damper-spring systems when you pull the mass a distance
from its equilibrium position and release it.
This specifies initial conditions ("released" means no initial velocity).
Case 1 - Overdamped:

→ , where
Take the derivative, then apply the ICs:

You can substitute this back into directly in maple, or rearrange if you like:

It's hard to notice it from this general form, but since the decaying exponential

term always dominates the rising exponential components of the sinh & cosh functions
meaning that this solution decays exponentially in time from its starting position.

Case 2 - Critically damped:


Take the derivative, then apply the ICs:
Because the exponential term dominates the linear term for large enough time, this also acts
as a decaying exponential.

Case 3 - Under damped:

→ , where
Take the derivative, then apply the ICs:

This solution is oscillatory with frequency due to the sin and cos terms, but modulated by a

decaying exponential function .

4.1.1.1 Setup & Solve Harmonic DEs with Maple


Suppose m = k = 1, and b = 4, 2, and 1 for over, critically, and underdamped, respectively. Write
down the differential equations of motion for releasing the mass from , and solve them in
Maple (hint, use dsolve). Verify that the solutions from Maple match the form of the analytical
solutions. Plot the motion and describe what happens in each damping case.

Solution:
DE is
→ , where .

In maple, the system solutions work like this:


Sets up variables, solve the DEs, and assigns the results:
restart;
m:=1: k:=1: x0:=1:
ode:= m*diff(x(t), t,t)+b*diff(x(t),t)+k*x(t)=0;
ics:= x(0)=x0, D(x)(0)=0;
assign(dsolve(subs([x=overdamped, b=4], {ode, ics})));
assign(dsolve(subs([x=critdamped, b=2], {ode, ics})));
assign(dsolve(subs([x=underdamped,b=1], {ode, ics})));

2
d  d
ode :=  2 x( t ) b  x( t ) x( t )0
 dt  dt 
 
ics := x( 0 )1, D( x )( 0 )0
Demonstrate that it worked:
overdamped(t);
critdamped(t);
underdamped(t);
x(t);
1 3  ( ( 2 3 ) t)  3 1  ( ( 2 3 ) t)
  e     e
2 3   3 2
( t ) ( t )
e e t
 t  t
     
1  2 t 3   2 t 3 
3e 
sin 
e cos 
3  2   2 
x( t )
Plots the solutions:
> plot([overdamped(t), critdamped(t), underdamped(t)], t=0..20, , legend =
["overdamped", "critdamped", "underdamped"]);

Overdamped: slowly returns to the equilibrium position


Critically Damped: returns to the equilibrium position quickly, and never crosses it.
Underdamped: returns even quicker, but oscillates.

4.1.1.2 Repeat around Critical damping


Plot the solutions for the case of slightly (5%) more and slightly (5%) less than critical damping.

Solution:
Since critical damping is b = 2, we can try b = 2.1 & b = 1.9 for slightly more and slightly less
than critically damping:
restart;
m:=1: k:=1: x0:=1:
ode:= m*diff(x(t), t,t)+b*diff(x(t),t)+k*x(t)=0:
ics:= x(0)=x0, D(x)(0)=0:
assign(dsolve(subs([x=overdamped, b=2.1], {ode, ics})));
assign(dsolve(subs([x=critdamped, b=2], {ode, ics})));
assign(dsolve(subs([x=underdamped,b=1.9], {ode, ics})));
plot([overdamped(t), critdamped(t), underdamped(t)], t=0..20,x=-.001..0.01,
legend = ["overdamped", "critdamped", "underdamped"]);

Indeed, critical damping is the fastest it can return to the equilibrium position without actually
ever crossing it. Slightly underdamped may be preferable for certain applications, e.g., when
you want to minimize time spent more than 0.002 away from equilibrium.
4.1.2 Damping Ratio
In the underdamped case, the oscillations of the harmonic response occur at a frequency of

, where is the natural frequency of an undamped mass-

spring system and is the damping ratio.

Question:
Rewrite the DE in terms of and (rather than ), and then use maple to solve the
damper-mass-spring system for m = 1, k = 1, x0 = 1, and damping ratios of 1/4, 1/8, and 1/16.
Compare with a cosine at the natural frequency.

Solution:

In Maple:
restart;
m:=1: k:=1: x0:=1:
w0:=m/k:
ode:= diff(x(t), t,t)+2*zeta*w0*diff(x(t),t)+w0^2*x(t)=0:
ics:= x(0)=x0, D(x)(0)=0:
assign(dsolve(subs([x=X1, zeta=1/4], {ode, ics})));
assign(dsolve(subs([x=X2, zeta=1/8], {ode, ics})));
assign(dsolve(subs([x=X3, zeta=1/16], {ode, ics})));
plot([X1(t), X2(t), X3(t), cos(w0*t), exp(-t/16)], t=0..20,
legend=[zeta=1/4,zeta=1/8,zeta=1/16, cosine, exp]);
Higher damping ratios make the oscillations decay faster and shift their frequency down.

The motion in any case decays with exponent , as you can see from the
exponential plot.

4.2 Driven Resonance and Quality Factor

4.2.1 Forced oscillations

Suppose we "drive" the mass-spring system by applying a harmonic force to it.


The new equation of motion is the inhomogeneous DE:

Solution derivation, Math2Z03 material:


To solve this, use the Method of Undetermined Coefficients: assume a particular solution of
the form: .

Then get the table:


Table 0-1: Derivatives for Method of Undetermined Coefficients for Linear System Forced Response DE

So, the total solution (for the underdamped case) is now:


Where , is the natural frequency of an undamped mass-

spring system, and is the damping ratio.

Check this result with maple:


restart;
ode:= m*diff(x(t), t,t)+b*diff(x(t),t)+k*x(t)=F0*sin(omega*t):
dsolve(ode);

 ( b b 24 k m ) t   ( b b 24 k m ) t 


     
 2m   2m 
x( t )e _C2e _C1
2
F0 ( cos(  t )  bsin(  t ) ksin(  t )  m )

b 2  2k 22 k m  2 4 m 2

Analysis: The solution has two parts:

1. The first part ( ), called the "transient" term, decays with

time according to .
2. The second part, called the "steady state" term, doesn't decay with time, is a harmonic
solution at the same frequency as the driving force ( ) but phase shifted from the

original sine function, and has an amplitude of .

4.2.1.1 Limiting Cases of Driving Frequency


Low Frequency Response: What is the response of the system when ? (the "low-
frequency response" or "DC response")

Solution:

A= ,

In this limit, the amplitude becomes:


While the phase shift is:
 The response to low-frequency force is determined by the spring constant (and not
the damping or mass), is independent of frequency, and is perfectly in phase with the excitation
Note also that is the extension a force produces in a spring with spring constant k.

High Frequency Response: What is the response of the system when ?

Solution:

Now the amplitude limits to:

While the phase shift is:


 The response to high frequency force is determined by the mass (and not the
damping or spring constant), decays with , and is lags 180o behind the input.

Resonant Response: What is the response of the system when ?

Solution:

Amplitude is:

Phase shift is:


 The response to a driving signal at the natural frequency of the system is the low-

frequency response modulated by the relative damping and lags 90o behind the
input.
(Note: if damping is high, this is not exactly the resonant frequency that causes maximum

amplitude; that would be where the denominator of the amplitude term is


minimized. We'll derive this in section 2.2.4).
4.2.2 Resonance
To summarize the third case in the previous section, we found that in steady state, the response

of the system to the force when is:

, where

For sufficiently low damping ( ), driving the system at its natural frequency puts it into
resonance: the harmonic driving force pushes on the mass in perfect synchronization with its
movement and builds on the velocity from the cycle before, leading to a higher response
amplitude than the same force would produce at low frequency. For this reason, the natural
frequency of the system is often called its resonant frequency (though technically the true
resonant frequency is slightly lower).

4.2.2.1 Example: Resonance


Plot the position vs. time for a mass-spring system with driven with a
force corresponding to at three frequencies, .

(take )

Solution:
DE is:

They gave us F0/k (which should be the low-frequency response) but in the form of the DE with
we need F0/m.

If , then .
restart;
w0:=10: x0:=1: zeta:=1/16:
ode:= diff(x(t), t,t)+2*zeta*w0*diff(x(t),t)+w0^2*x(t)=x0*w0^2*sin(omega*t):
ics:=x(0)=0, D(x)(0)=0;
dsolve({ode, ics});
assign(%):
plot([subs(omega=2, x(t)), subs(omega=10, x(t)), subs(omega=50, x(t))],
t=0..10);

ics := x( 0 )0, D( x )( 0 )0


  5 t 
 8   5 255 t 
e sin   ( 32  23175 ) 255
16  8 
x( t )
51 1600003175  216  4
  5 t 
 5 255 t 
 8 
cos
2000 e  
 8 

1600003175  16  4
2

400 ( 5 cos(  t ) 400 sin(  t )4  2 sin(  t ) )



1600003175  216  4

The low frequency response is basically a sine function with amplitude 1, the resonant response
has an amplitude around 8, and the high-frequency response is reduced to very low amplitude.
Our predictions were:

1. Low frequency response should be:

2. Resonant frequency response should be:


3. High frequency response should be:

4.2.2.2 Superposition and Resonant Filters


In the previous example,
a) determine how the system would respond in response to all three excitation forces
simultaneously.
b) repeat part (a) at 1/10th and 1/100th the damping.

Solution:
With a superposition of all 3 individual responses (because it's a linear DE); this means that the
combined response is essentially just the response the resonant driving force would have by
itself, since it's so much larger than the others - especially at those lower damping values.

Example code:
> restart:
zeta:=1/16: omega0:=10:
DE:=diff(x(t),t,t)+2*zeta*omega0*diff(x(t),t)
+omega0^2*x(t)=omega0^2*sin(omega*t);
ICs:=x(0)=0,D(x)(0)=0:
dsolve([subs(omega=2, DE), ICs]): xLF:=rhs(%):
dsolve([subs(omega=50, DE), ICs]): xHF:=rhs(%):
dsolve([subs(omega=10, DE), ICs]): xRF:=rhs(%):
dsolve([diff(x(t),t,t)+2*zeta*omega0*diff(x(t),t)
+omega0^2*x(t)=omega0^2*(sin(2*t)+sin(10*t)+sin(50*t)), ICs]):
xMF:=rhs(%):
plot([xLF, xHF, xRF, xMF], t=0..20);
 d2  5 d
DE :=  2 x( t )   x( t ) 100 x( t )100 sin(  t )
 dt  4  dt 
4.2.2.3 Amplitude & Phase Shift Log Plots
Plot the log of the steady-state amplitude of an rad/s mass-spring-damper system with

to a driving force at  m vs the log of frequency. Repeat for the phase vs. log of
frequency.

Solution:
restart;
w0:=10: x0:=1: zeta:=1/16:
A:=x0*w0^2/sqrt((w^2-w0^2)^2+(2*zeta*w0*w)^2);
phi:=-arctan((w^2-w0^2)/(2*zeta*w0*w))-Pi/2;
with(plots): #Includes the plots package in Maple, which is required for
advanced plotting like loglogplot() and semilogplot()
loglogplot(A, w=1..100, numpoints=1000); #loglogplot is log horizontal, log
vertical
semilogplot(phi*180/Pi, w=1..100); #semilogplot is log horizontal, linear
vertical
400
A :=
16 w 3175 w2160000
4

2
 4 ( w 100 )  
 := arctan  
 5w  2
4.2.3 Quality Factor
The amplitude amplification ratio at resonance compared to the DC response is also called the

quality factor, . From case 3 above:

However, this isn't the only definition of quality factor you'll find. In fact, quality factor has 3
commonly-used definitions:
1. Q is the amplitude amplification factor at resonance (i.e., steady-state amplitude at
resonance is Q multiplied by the amplitude the same excitation would produce at zero

frequency):
2. Q is 2* the ratio of the peak energy stored to the energy dissipated per cycle at

resonance:
3. Q is the ratio of the resonant frequency to the half-width (full width at half maximum) in
the power response vs. frequency graph (and power goes like the square of amplitude):

Let's explore these and see what these each say about the quality factor of our linear system, and
whether they're truly equivalent.

Definition 1: (previous section)

Definition 2:

We can find peak energy stored from at the max displacement of the spring, which is the
amplitude of the motion (we could also have used max kinetic energy by finding the velocity as

it crosses the 0 point):


On the other hand, the dissipated energy per cycle is the work done on the damping force:

Where is the period of oscillations.

Using the steady-state resonant response only ( ) and that

We have:
Therefore:
Definition 3:
Where is the full-width-at-half-maximum of the of the relative power response curve.

Returning to the frequency-dependent steady-state response:

The amplitude amplification compared to DC is:

The relative power response is then:

Where is the relative frequency and are two convenience variables we'll use
for this proof. Using our new convenience variables, the relative frequency at the half-width is:
For small damping:

and

so:

To rewrite the difference of slightly different terms it's a good time to use the binomial series

approximation: , i.e., :

Therefore,

4.2.3.1.1 Exercise: Total Response in terms of Q


Rewrite the total response of the system in terms of quality factor and natural frequency.

Recall,

Where , is the natural frequency of an undamped mass-

spring system, and is the damping ratio.


Since , ,

Wher

e .

4.2.3.1.2 Exercise: Transient decay time


In terms of Q, how many cycles does it take for the transient response amplitude to drop to 1/e of
it's initial value?

Solution:

The transient response has the exponential decay term . This has an exponent of -1 when

4.2.3.2 Example: Quality Factor from Response Curve


Given the following plot of a mass-spring system with linear damping's response vs. driving
frequency (in rad/s), determine the resonant frequency, quality factor, and state whether the
system is underdamped, critically damped, or overdamped.
Solution:
(peak amplitude)
Q is Ares/ADC = 16/2 = 8. (Could also try to read the ωres/FWHM, but would need to know it in
the power spectrum; therefore look for the FW@1/sqrt(2) of max instead, which is A = 11.35;
width looks like about 2, meaning Q is about 10/2 = 5. Too rough to get an accurate number but
consistent with the other calc of 8.)

Critically damped was defined as . And also:

Since , the system has much smaller b (i.e., less damping) than it would if
critically damped, so underdamped.
Note that for critically or overdamped systems it's hard to even see a resonance peak; critical
damping is at Q = ½ (meaning ). Since this peak is prominent we know immediately that
it's underdamped.

4.2.4 Further exploration:


Here is that same resonance curve but also plotted for other Q's besides Q = 8; specifically, Q =
4, 2, 1, ½, and ¼ (the final two correspond to critically damped and over damped, respectively):

Note that when Q is low (2 or less) is no longer the frequency with the maximum amplitude.

This is because the amplitude is given by , and while is certainly


the minimum value of the first term in the denominator, that second term in the denominator is
increasing with , so moving a bit to the left of will increase the amplitude. The effect
is more important when the damping is higher (meaning Q is lower, meaning the second term in
the denominator is larger compared to the first one near resonance.) To be a bit more
quantitative about it we can find where the denominator is a minimum:
When Q is 8 we get so we can usually ignore the correction, but when Q is 1

this becomes which is a significant effect.

(Code for this follows. Note that the gridlines might not work for all maple worksheet modes.)
restart;
w0:=10: x0:=1: #zeta:=1:
A:=zeta->x0*w0^2/sqrt((w^2-w0^2)^2+(2*zeta*w0*w)^2):
phi:=-arctan((w^2-w0^2)/(2*zeta*w0*w))-Pi/2:
with(plots):
plot(2*[A(1/16), A(1/8), A(1/4), A(1/2), A(1), A(2)], w=0.1..40,
numpoints=1000, labels=["omega", "Amplitude"], axis=[gridlines=[10,
color=blue]]);

4.2.4.1
To clarify the three frequencies of a mass-spring system with linear damping:
1. is the natural frequency. This is the frequency the system would respond
at if there was no damping. In that case, it's also the driving frequency that will make the
response grow to infinite amplitude over time.

2. is the damped natural frequency; this is the frequency of the damped


system's transient response

3. is the resonant frequency; this is the frequency where the system's


response has the largest amplitude.
In most undamped cases of interest, is so small that the difference between these frequencies
is negligible, so people often will just call the resonant frequency.

For example, compare the response amplitude at all 3 frequencies when :


> restart;
omega0:=10: x0:=1: zeta:=1/16:
ode:= diff(x(t), t,t)+2*zeta*omega0*diff(x(t),t)
+omega0^2*x(t)=x0*omega0^2*sin(omega*t):
ics:=x(0)=0, D(x)(0)=0;
dsolve({ode, ics});
x:=rhs(%):
omega_d:=evalf(omega0*sqrt(1-zeta^2));
omega_max:=evalf(omega0*sqrt(1-2*zeta^2));
evalf(subs(omega=omega0,x));
evalf(subs(omega=omega_d, x));
evalf(subs(omega=omega_max, x));
ics := x( 0 )0, D( x )( 0 )0
 5 t   5 t 
   
 8   5 255 t  2  5 255 t 
 8 
e sin   ( 32  3175 ) 255 2000 e cos  
16  8  8
x( t ) 
51 1600003175 216 4 1600003175 216 4
400 ( 5 cos(  t ) 400 sin(  t )4 2 sin(  t ) )

1600003175 216 4
omega_d := 9.980449638
omega_max := 9.960860906
( 0.6250000000t ) ( 0.6250000000
t)
0.5009794328 e sin( 9.980449638 t )8. e cos( 9.980449638 t )
8. cos( 10. t )
( 0.6250000000t )
0.2507345621 e sin( 9.980449638 t )
( 0.6250000000t )
8.007820382 e cos( 9.980449638 t )
8.007820384 cos( 9.980449638 t )0.2507345982 sin( 9.980449638 t )
( 0.6250000000t )
7.999938404 e cos( 9.980449638 t )7.999938404 cos( 9.960860906 t )
0.5019607792 sin( 9.960860906 t )
Looking at the steady-state cos and sine term amplitudes and combining them using
we can find the total response amplitude:
> Amp_omega0:=8;
Amp_omega_d:=sqrt(.2507345739^2+8.007820122^2);
Amp_omega_m:=sqrt(.5019607843^2+7.999938484^2);
Amp_omega0 := 8
Amp_omega_d := 8.011744563
Amp_omega_m := 8.015670925

Thus, the response at is indeed the largest amplitude (8.016) but is very similar to the
response at (8).
4.3 Resonance Summary
A mass m on a spring with constant k subject to linear damping has a DE of:

Or, if it's also subject to a driving force , has a DE of:

The relative size of the damping determines how the system responds:

1. Overdamping ( ):
a. When disturbed the system will slowly return to equilibrium.
b. No resonance in response to driving force.

2. Critical damping ( ):
a. When disturbed the system will return to equilibrium as fast as possible without
ever crossing the equilibrium point.
b. No resonance in response to driving force.

3. Underdamping ( ):
a. When disturbed the system will harmonically oscillate at its natural resonant
frequency, with an amplitude that decays exponentially in time.
b. If the driving force has the same frequency as the system's natural frequency, the
response amplitude can be larger than the driving force would produce at low
frequency: this causes the system to resonate in response to the driving force.

Other Terms in Common Use:

1. natural frequency (undamped resonant frequency).


a. With damping, the system naturally responds at the damped resonant frequency:

(though this is virtually the same unless damping is


large).
b. If driven, the system responds at the same frequency the driving force has. The
driving force that makes it respond the most is the slightly different

, which also is virtually the same as unless damping is


large.
2. damping ratio. Critical damping is at .
3. Quality factor:

a. : The amplitude amplification at resonance (i.e., steady-


state amplitude at resonance is Q multiplied by the amplitude the same excitation
would produce at zero frequency).

b. : 2* the peak energy stored over the energy dissipated per
cycle at resonance.

c. If Q is large, then also : The resonant frequency over the half-


width (full width at half maximum) in the power response vs. frequency graph.

d. For the linear mass-spring system, .

corresponds to critical damping, however, resonance doesn't exist unless


the system is underdamped so it's strange to talk about the quality factor of the
resonance when there isn't a resonance. Therefore, Q should be > ½ if you're
referring to it.

Typically solving a DE for a linear system will work like this:


restart;
m:=1: y0:=.05: g:=9.81: W:=m*g: k:=W/y0: x0:=+0.02: v0:=0:
b:=sqrt(4*m*k)/8:
ode:= m*diff(x(t), t,t)+b*diff(x(t),t)+k*x(t)=3*sin(omega*t);
ics:= x(0)=x0, D(x)(0)=v0;
interface(displayprecision=4): #Makes Maple only show 4 decimals in its
displayed output
evalf(dsolve({ode,ics}));
assign(%):
subs(omega=3, x(t));
plot([%, 0.016*cos(3*t)], t=0..8);
In the solution, there are two terms that include an exponential decay term ( ). These are the
transient terms because they disappear after enough time. The other term sticks around and is
called the steady state term. The steady state term has the form, . The

amplitude of this term is while the phase is ,


so you can rewrite the steady state term as .
Trial and error with the plot would also allow you to determine the amplitude and phase shift.

You might also like