Signal Analysis: The Fourier
Transform
 Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
Fourier Transform
• In the previous lecture, we represented periodic signals a sum of sinusoids or
  exponentials using Fourier series.
• In this lecture, we will extend the Fourier method to continuous-time (CT)
  aperiodic signals.
• We can represent aperiodic signals a continuous sum (or integral) of exponentials
  and we can find frequency components of aperiodic signals using Fourier
  transform.
• Fourier transform (CTFT) are used to express both aperiodic and periodic
  CT signals.
      Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
Fourier Transform
• Any signal              is represented by the Fourier integral (rather than a Fourier series)
                                                                                                              𝑥 𝑡 =              𝐷 𝑒
 where              is Fourier transform of                          .
                                                                                                                      1
                                                                                                              𝐷 =             𝑥(𝑡)𝑒    𝑑𝑡
                                                                                                                      𝑇
• We call      the Fourier transform of                                   , and             the inverse Fourier
  transform of     .
• We can plot the spectrum       as a function of w. Since                                                   is complex, we
  have both amplitude and angle (or phase) spectrums.
                                                                               ∠
       Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
Example
• Find the Fourier transform of                                     if a>0.
    →                →                           →
• To plot the spectrum, we should find                                       and
        Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
                             1     𝑎 − 𝑗𝑤                                        𝑎                                     −𝑤
               𝑋 𝑤 =             =        ,                 𝑅𝑒 𝑋 𝑤         =        ,            𝐼𝑚 𝑋 𝑤          =
                           𝑎 + 𝑗𝑤 𝑎 + 𝑤                                        𝑎 +𝑤                                  𝑎 +𝑤
                                                           𝑎                     −𝑤                  𝑎 +𝑤                  1
    𝑋 𝑤   =   𝑅𝑒 𝑋 𝑤          + 𝐼𝑚 𝑋 𝑤           =                      +                    =                       =                  𝑎 +𝑤
                                                         𝑎 +𝑤                  𝑎 +𝑤                  𝑎 +𝑤                𝑎 +𝑤
                                                                                                 1
                                        𝑋 𝑤      = 𝑎 +𝑤                𝑎 +𝑤           =
                                                                                            𝑎 +𝑤
                                                                                   −𝑤
                                                  𝐼𝑚 𝑋 𝑤                         𝑎 +𝑤                            𝑤
                          ∠𝑋 𝑤 = tan                               = tan            𝑎            = tan       −
                                                  𝑅𝑒 𝑋 𝑤                                                         𝑎
                                                                                 𝑎 +𝑤
                                                      Even function                           Odd function
•         is an aperiodic signal and it has a continuous spectrum whereas, remember
    that, if      were periodic signal it would have a discrete spectrum.
• An aperiodic signal can be represented as linear combination of complex
  exponentials at each frequency (               ).
          Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
Comment on amplitude spectrum
                                  𝑥 𝑡 =𝑒   𝑢 𝑡
                                          1
                                  𝑋 𝜔 =
                                        𝑎 + 𝑗𝜔
• How can we interpret amplitude spectrum?
• The Fourier spectrum of a signal indicates the relative amplitudes that are required
  to synthesize that signal. It is a density function.
• The magnitude spectrum reveals the distribution of the energy of x(t) in frequencies.
• Signal energy is a measure to express size of a signal.
• The energy contained in the frequency range [                                    ,      ] can be computed from
       Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
    Transforms of some useful functions                                                                                                  sin 𝜋𝑥
                                                                                                                                                = 𝑠𝑖𝑛𝑐 𝑥
                                                                                                                                            𝜋𝑥
                                                                                                                                         𝑓 = 𝑤/2𝜋
• Find the Fourier transform of
since                          for |t|< /2 and since it is zero for |t|> /2,
                                                                                                                                            𝑤𝜏
                     /
                                            1                             1 −𝑒             +𝑒           2 −𝑒          +𝑒             2 sin 2
        𝑋 𝑤 =             𝑒      𝑑𝑡 = −       𝑒           −𝑒          =                             =                              =
                      /                    𝑗𝑤                             𝑤            𝑗                𝑤            2𝑗                   𝑤
                  2𝜋f𝜏
            2 sin  2     sin 𝜋f𝜏    sin 𝜋f𝜏                    𝑤𝜏                                                                  𝒘𝝉
        =              =         =𝜏         = 𝜏𝑠𝑖𝑛𝑐 f𝜏 = 𝜏𝑠𝑖𝑛𝑐                                       𝒓𝒆𝒄𝒕(𝒕/𝝉) ⇔ 𝝉𝒔𝒊𝒏𝒄
                2𝜋𝑓         𝜋𝑓        𝜋𝑓𝜏                      2𝜋                                                                  𝟐𝝅
                                                 𝑋 𝑤 is a real function.
                                                 How can we find its amplitude
                                                 and phase spectrums?
                                                 Remember!
                                                       1 = 1𝑒 °
                                                      −1 = 1𝑒 = 1𝑒
                Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
The essential bandwidth of a signal
• The spectrum of all practical signals extend to infinity.
• But most of the signal energy is contained within a certain band of B Hz, and the
  energy contributed by the components beyond B Hz is negligible.
• We can therefore suppress the signal spectrum beyond B Hz with little effect on
  the signal shape and energy. The bandwidth B is called the essential bandwidth
  of the signal.
       Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
• Find the Fourier transform of the unit impulse
       Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
• Find the inverse Fourier transform of                                  .
       Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
• Find the inverse Fourier transform of                                            .
                                                    𝒋𝒘𝟎 𝒕
                                                                                  𝟎
                                                 𝒋𝒘𝟎 𝒕
                                                                                       𝟎
• This result shows that the spectrum of an everlasting exponential                                                      is a single
  impulse at         . Based on this
                                                  𝒋𝒘𝟎 𝒕
                                                                                       𝟎
       Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
• Find the Fourier transforms of the everlasting sinusoid                                                     .
using the previous results
                                          𝟎                                𝟎                         𝟎
       Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
      Fourier Transform of a periodic signal
      • We can use a Fourier series to express a periodic signal as a sum of exponentials
        of the form
      • We found Fourier transform of exponential. 𝒋𝒘𝟎𝒕                                                                  𝟎
      • The Fourier series of a periodic signal x(t) with period                                          is given by
      • Taking the Fourier transform of both sides, we obtain
𝑇 = 𝜋, 𝑤 =    =2
                                        1                                               0.504
             𝑥 𝑡 = 0.504                     𝑒          ,      𝑋 𝜔 = 2𝜋                        𝛿(𝑤 − 𝑛2)
                                     1 + 𝑗4𝑛                                           1 + 𝑗4𝑛
               Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
Principles of Linear Systems and Signals by Lathi, Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif
Discrete Fourier Transform
Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Where we use Fourier transform
• Almost every scientific discipline uses Fourier transform.
   •   Telecommunication
   •   Signal processing
   •   Medical imaging
   •   Quantum mechanics
   •   Geology
   •   Astronomy….
• An example: to design a communication system, the distribution of the energy of
  the signal is important. Carrier frequency, sampling frequency, the cut-off
  frequency of a filter depends on energy distribution.
       Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Fourier transform of a practical signal
•
• How can we apply FT to a signal that
  hasn’t a mathematical equation such
  speech signal?
• For real-world signals, generally,
  analytical computation of CTFT is not possible.
• Thus, we discuss its numerical computation.
• Distribution of the energy of this speech signal in frequencies can be found using
  its samples.
• Sampling this continues signal produce its discrete signal counterpart.
• The discrete signal can be used to reveal the energy distribution.
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Discrete-time signals
• If a signal is defined only at discrete values of time, it is called
  a discrete time (DT) signal.
• A DT signal may occur naturally. Example is the
  one-dimensional hourly measurements x[n]
  made with an electronic thermometer.
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Discrete-time signals
• Alternatively, a DT signal may be derived from a CT signal by
  a process known as sampling.
• Alternatively, a DT signal may be derived from a CT signal by
  a process known as sampling.
• Alternatively, a DT signal may be derived from a CT signal by
  a process known as SAMPLING.
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Sampling
• Sampling a signal is acquiring values from a continuous-time signal at discrete
  points in time.
• The set of samples forms a discrete-time signal.
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Example
•                         where T denotes the time interval between two consecutive
  samples and n is sample number.
• Consider the CT signal                  . Sample the signal using a sampling
  interval of           , and sketch the waveform of the resulting DT sequence for
  the range              .
• By substituting t = nT , the DT
  representation of the CT signal
  x(t) is given by
    For
    the DT signal x[n] has the
    following values:
       Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
•              is sampling interval or SAMPLING PERIOD                                             𝑺.
•       is SAMPLING FREQUENCY                                     𝑺.
• Now we can use the samples to obtain spectrum.
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Discrete Fourier transform (DFT)
• We have
• Select the number of to be N, same as the number of samples                                                       .
• We can select N frequency in                . Generally                                                       is preferred.
• The first frequency is located at                             , and the last frequency is located at                               .
 frequency points:                                                        .
• Frequency resolution is
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
    Example                                                                                                       𝑋𝑘 =           𝑥𝑛𝑒
• Find DFT coefficients of                                       .
•
•
•
•
•
•
•
        Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
•
    Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
    Example                                                                                                       𝑋𝑘 =           𝑥𝑛𝑒
• Find DFT coefficients of                                                  .
•
•
•
•
•
•
        Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Spectrum of a sound signal
                                                                    DFT Coefficients              Absolute value of the DFT Coefficients
                                                                            .
                                                                            .
                                                                            .
   Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
• We can compute the magnitude spectrum                                       from
• Most applications use only magnitude spectra because they reveal the distribution
  of energy in frequencies.
                                   𝜋 𝜔
            𝑇𝑋 𝜔        𝑓𝑜𝑟 𝜔 <      =
 𝑋 𝜔 ≈                             𝑇    2
                0             𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
• To obtain frequency value in radian use
• To obtain frequency value in Hertz use
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen
Do frequency spectra play any role during system operation?
• An example: to design a communication system, the distribution of the energy of
  the signal is important. Carrier frequency, sampling frequency, the cut-off
  frequency of a filter depends on energy distribution.
• Do frequency spectra play any role during system operation? Sometimes yes
  sometimes no.
• Although it has been used for system design (like a communication system) or to
  create a file (like an mp3 file), it may not play a role in real-time processing.
• When we speak into a telephone, the speech will appear almost instantaneously at
  the other end. We don’t need its spectra. When we listen an mp3 file, computer
  doesn’t calculate a spectra.
• But if a system uses frequency characteristic of a signal (like speech recognition)
  it always calculate spectra.
     Continuous and Discrete Time Signals and Systems by Mandal and Amir Asif, Signals and Systems: A Fresh Look by Chi-Tsong Chen