06.

Harmonic Analysis
Although harmonic motion is simplest to handle, the motion of many vibratory systems is not
harmonic. However, in many cases the vibrations are periodic for example, the type shown below
on left. Fortunately, any periodic function of time can be represented by Fourier series as an
infinite sum of sine and cosine terms.

Periodic Function

6.1 Fourier Series Expansion

If x  t  is a periodic function with period  , its Fourier series representation is given by:

a0
x t    a1 cos t  a2 cos 2t  ...  b1 sin t  b2 sin 2t  ... 
2

a
 0    an cos nt  bn sin nt 
2 n 1

2
In the equation above  is the fundamental frequency and a0 , a1 ,...b1 , b2 ... are constant

coefficients. To determine the coefficients an and bn we multiply the equation above by
2
cos nt and sin nt respectively, and integrate over one period   for example, from 0 to

2
. Then we notice that all terms except one on the right-hand side of the equation will be zero,

and we obtain:
2

 
a0 
  x  t  dt
0
2

 
2
an   x  t  cos ntdt   x  t  cos ntdt ,
 0
 0
 2

 
2
bn   x  t  sin ntdt   x  t  sin ntdt
 0
 0

a0 
The physical interpretation of the equation x  t      an cos nt  bn sin nt  is that any
2 n 1
periodic function can be represented as a sum of harmonic functions. Although the series in the
above equation is an infinite sum, we can approximate most periodic functions with the help of
only a few harmonic functions. For example, the triangular wave represented above on left can be
represented closely by adding only three harmonic functions, as shown above on right.

Fourier series can also be represented by the sum of sine terms only or cosine terms only. For
example, the series using cosine terms only can be expressed as:

x  t   d 0  d1 cos t  1   d 2 cos  2t  2   ... ,

a0
d0  ,
2
1
d n   an2  bn2  2 ,
bn
n  arctan
an

 Gibbs Phenomenon
When a periodic function is represented by a Fourier series, an anomalous behaviour can be
observed. For example, the illustration below shows a triangular wave and its Fourier series
representation using a different number of terms. As the number of terms n increases, the
approximation can be seen to improve everywhere except in the vicinity of the discontinuity P .
Here the deviation from the true waveform becomes narrower but not any smaller in amplitude. It
has been observed that the error in amplitude remains at approximately 9 percent, even when
k   . This behaviour is known as Gibbs phenomenon, after its discoverer.

Gibbs phenomenon

2
6.2 Complex Fourier Series
The Fourier series can also be represented in terms of complex numbers using the following
relationships:

eit  cos t  i sin  t

e it  cos t  i sin  t
eit  e  it
cos t 
2
e  e  it
it
sin t 
2i

a0 
Equation x  t      an cos nt  bn sin nt  and introducing b0  0 can be written as:
2 n 1

a0   eint  e  int eint  e  int 

x  t      an  bn 
2 n 1  2 2i 
 a ib     a ib   a ib  
 ei0t  0  0    eint  n  n   e  int  n  n  
 2 2  n 1   2 2   2 2 

Complex Fourier coefficients cn and cn can be introduced here as:

an  ibn
cn  ,
2
a  ibn
c n  n
2

 a0 ib0    int  an ibn  int  an ibn  

Then the equation x  t   e  2  2    e  2  2   e
i0t
 2  2   can be
  n 1     
written as:

x  t    cn eint

 
2 2
Using equations an 
  x  t  cos ntdt and bn 
0
  x  t  sin ntdt , the Fourier coefficients
0

cn can be determined as:

an  ibn 1  1 
cn    x  t    cos nt  i sin nt  dt   x  t   e  int dt
2  0  0

3
6.3 Frequency Spectrum
The harmonic functions a0 cos nt or b0 sin nt in
a0 
x t      an cos nt  bn sin nt  are called Harmonics of Order n of the periodic
2 n 1

function x  t  . The harmonic of order n has a period . These harmonics can be plotted as
n
vertical lines on a diagram of amplitude an and bn or d n and n versus frequency n , called
frequency spectrum or spectral diagram. The illustration below shows a typical frequency
spectrum.

Frequency spectrum of a typical periodic function of time

6.4 Time-Domain and Frequency-Domain Representations

The Fourier series expansion permits the description of any periodic function using either a time-
domain or a frequency-domain representation. For example, a harmonic function given by
x  t   A sin t in time domain – see illustration below, top left - can be represented by the
amplitude A and the frequency  in the frequency domain - see top right. Similarly, a periodic
function such as a triangular wave can be represented in time domain, as shown in the illustration
below – bottom left, or in frequency domain, as indicated in the same illustration – bottom right.
Note that the amplitudes d n and the phase angles n corresponding to the frequencies can be
used in place of the amplitudes an and bn for representation in the frequency domain.

Representation of a function in time and frequency domains

4
The use of Fourier integrals permits the representation of even non-periodic functions in either a
time domain or a frequency domain. The illustration above shows that the frequency-domain
representation does not provide the initial conditions. However, in many practical applications the
initial conditions are often considered unnecessary and only the steady-state conditions are of
main interest.

6.5 Even and Odd Functions

An even function satisfies the relation:

x  t   x  t 

In this case, the Fourier series expansion of x  t  contains only cosine terms:

a0 
x t     an cos nt
2 n 1

An odd function satisfies the relation

x  t    x  t 

In this case, the Fourier series expansion of x  t  contains only sine terms:


x  t    bn sin nt
n 1

The coefficients a0 , an and bn are calculated – as already explained above – with:

2

 
a0 
  x  t  dt
0

2
an 
  x  t  cos ntdt ,
0

2
bn 
  x  t  sin ntdt
0

In some cases, a given function may be considered as even or odd depending on the location of
the coordinate axes. For example, the shifting of the vertical axis from (a) to (b) or (c) below on
top will make it an odd or even function. This means that we need to compute only the
coefficients bn or an . Similarly, a shift in the time axis from (d) to (e) amounts to adding a
constant equal to the amount of shift.

5
In the case illustrated above in the middle, when the function is odd, the Fourier series expansion
becomes:

4A 
1 2  2n  1 t
x1  t  

 2n  1 sin
n 1 

On the other hand, if the function is even, as shown above at the bottom, its Fourier series
expansion becomes:

 1 2  2n  1 t
n 1

4A
x2  t  


n 1 2n  1
sin


Since the functions x1  t  and x2  t  represent the same wave except for the location of the
origin, there exists a relationship between their Fourier series expansions also because
 
x1  t    x2  t  .
 4

6.6 Half-Range Expansions

In some practical applications, the function x  t  is defined only in the interval 0 to  as shown
below on top.

6
In such cases there is no condition of periodicity of the function, since the function itself is not
defined outside the interval 0 to  . However, we can extend the function arbitrarily to include the
interval  to 0 as shown in the above illustration in the middle or the bottom. The extension of
the function indicated in the middle results in an odd function x1  t  , while the extension of the

function shown at the bottom results in an even function x2  t  . Thus the Fourier series
expansion of x1  t  yields only sine terms and that of x2  t  involves only cosine terms. These
Fourier series expansions of x1  t  and x2  t  are known as half-range expansions. Any of these

half-range expansions can be used to find x  t  in the interval 0 to  .

Summary
In this lecture we discussed harmonic analysis, including:

 Fourier Series Expansion

 Gibbs Phenomenon
 Complex Fourier Series
 Frequency Spectrum
 Time-Domain and Frequency-Domain Representations
 Even and Odd Functions
 Half-Range Expansions

References
 Rao, S. – Mechanical Vibrations, 5th Edition, Prentice Hall, New York, 2011
 Gans, R.- Mechanical Systems - A Unified Approach to Vibrations and Controls,
Springer, 2015
 Thomson, W., Dahleh, M. - Theory of Vibrations with Applications, 5th Edition, Prentice
Hall, 1998