The Fourier Transform ES 442 Analog & Digital Communication Systems Spring 2016 Lecture 3 ES 442 Fourier Transform 1 Review: Fourier Trignometric Series (for Periodic Waveforms) Periodic function g(t) is defined in time interval of t1 t t1 T0 g(t ) a0 an cos(n2 f 0t ) bn sin(n2 f 0t ) n1 1 where g(t ) is of period T0 , and f 0 . T0 We can calculate the coefficients from the equations: 2 an T0 t1 T0 2 bn T0 t1 T0 Lathi & Ding pp. 50-61 g(t ) cos(n2 f 0t )dt for n = 1, 2, 3, 4, t1 t1 g(t ) sin(n2 f 0t )dt for n = 1, 2, 3, 4, g(t) T0 ES 442 Fourier Transform t 2 Review: Exponential Fourier Series (for Periodic Functions) Again, periodic function g(t) is defined in time interval t1 t t1 T0 g(t ) n Dn e jn 2 f 0t for n 0, 1, 2, 3, where again g(t ) is of period T0 , and f0 1 . T0 Lathi & Ding pp. 50-61 We can calculate the coefficients Dn from the equation: Dn 1 jn 2 f 0t g ( t ) e dt for n ranging from - to . T0 T0 But, periodicity is too limited for our needs, so we will need to proceed on to the Fourier Transform. The reason is that signals used in communication systems are not periodic (it is said they are aperiodic). ES 442 Fourier Transform 3 Review: Time Domain and Frequency Domain The Fourier series expresses any time-periodic function in terms of sinusoidal functions (only need to specify the frequency, amplitude and phase of each sinusoid). When you think about that it is remarkable that any periodic function can be simply represented by a sum of sinusoidal functions. In other words, the Fourier series of a time-periodic function generates a frequency spectrum of harmonically-related sine and cosine waveforms. “Every signal has a spectrum and is also determined by its spectrum. You can analyze the signal either in the time domain or in the frequency domain.” -- Professor Brad Osgood (Stanford University) This leads to time duration versus frequency bandwidth duality in signals. ES 442 Fourier Transform 4 Sinusoidal Waveforms are the Building Blocks in the Fourier Series Harmonic Motion Produces Sinusoidal Waveforms Past Future Mechanical Oscillation Time t Sheet of paper unrolls in this direction LC Tank Circuit ES 442 Fourier Transform Electrical LC Circuit Oscillation 5 Visualizing Time Domain & Frequency Domain of a Signal Amplitude Source: Agilent Technologies Application Note 150, Spectrum Analyzer Basics ES 442 Fourier Transform 6 Example: Periodic Square Wave Odd function f (t ) 4 sin( t ) 13 sin(3 t ) 15 sin(5 t ) 71 sin(7 t ) Fundamental only Five terms Eleven terms Forty-nine terms t http://ceng.gazi.edu.tr/dsp/fourier_series/description.aspx ES 442 Fourier Transform 7 Example: Periodic Square Wave (continued) ES 442 Fourier Transform 8 Square Wave From Fundamental + 3rd + 5th & 7th Harmonics f (t ) 4 sin( t ) 13 sin(3 t ) 15 sin(5 t ) 71 sin(7 t ) t https://en.wikipedia.org/wiki/Fourier_series ES 442 Fourier Transform 9 Another Example: Both Sine & Cosine Functions Required Period T0 Note phase shift in the fundamental frequency sine waveform. http://www.peterstone.name/Maplepgs/fourier.html#anchor2315207 ES 442 Fourier Transform 10 Gibb’s Phenomena at Discontinuities in Waveforms Amplitude The Gibbs phenomenon is an overshoot (or "ringing") of Fourier series occurring at simple discontinuities. time t About 9% overshoot at discontinuities of waveform. This is an artifact of the Fourier series representation. ES 442 Fourier Transform 11 Fourier Series versus Fourier Transform Continuous time Discrete time Periodic Fourier Series Discrete Fourier Transform Aperiodic Continuous Fourier Transform Discrete Fourier Transform Fourier series for continuous-time periodic signals → discrete spectra Fourier transform for continuous aperiodic signals → continuous spectra ES 442 Fourier Transform 12 Definition of Fourier Transform g(t )e j 2 ft Figure 3.15 Lathi & Ding dt g(t ) G( f ) G( f )e j 2 ft df Time-frequency duality: g(t ) G( f ) and G(t ) g( f ) We say “near symmetry” because the signs in the exponentials are different between the Fourier transform and the inverse Fourier transform. ES 442 Fourier Transform 13 Fourier Transform Produces a Continuous Spectrum FT [g(t)] gives a spectra consisting of a continuous sum of exponentials with frequencies from - to + . G( f ) G( f ) exp j ( f ) where |G(f)| is the continuous amplitude spectrum of g(t) and (f) is the continuous phase spectrum of g(t). Supplemental: Lathi & Ding pp. 97-99 ES 442 Fourier Transform 14 Example: Rectangular Pulse g(t ) rect(t ) (t / ) g(t ) rect(t ) (t / ) 1 for 0 for all t 2 t 2 2 G(f) Lathi & Ding pp. 101-102 Example 3.2 sinc(f ) 1 2 0 2 2 time t 3 2 Remember = 2f ES 442 Fourier Transform 1 0 1 2 3 f 15 Two Definitions of the Sinc Function sin( x ) sinc ( x ) (1) x and sin( x ) sinc ( x ) (2) x Sometimes sinc (x) is written as Sa (x). x Lathi and Ding, page 100, use the definition sinc (x) = ES 442 Fourier Transform sin( x ) x 16 Definition of the Sinc Function Unfortunately, there are two definitions of the sinc function in use. Format 1 (used in Lathi and Ding, fourth edition – pp. 100 – 102) sinc( x ) sin( x ) x Format 2 (used in many other text books) sinc( x ) sin( x ) x Properties: 1. sinc(x) is an even function of x 2. sinc(x) = 0 at points where sin(x) = 0, that is, sinc(x) = 0 when x = , 2, 3, …. 3. Using L’Hôpital’s rule, it can be shown that sinc(0) = 1 4. sinc(x oscillates as sinc(x) oscillates and monotonically decreases as 1/x decrease with increasing |x| 5. sinc(x) is the Fourier transform of a single rectangular pulse ES 442 Fourier Transform 17 Sinc Function in Two-Dimensions (e.g., Optics) “Airy Disc” Pardon the digression! ES 442 Fourier Transform 18 Sinc Function Tradeoff: Pulse Duration versus Bandwidth G1 ( f ) T1 1 T1 G2 ( f ) 1 T1 T2 f g1 (t ) T1 2 T1 > T2 > T3 T1 2 t G3 ( f ) T3 1 1 T3 g3 (t ) T3 T3 T3 2 2 ES 442 Fourier Transform 1 1 T2 g2 (t ) T2 T2 2 T2 2 f f t Lathi & Ding pp. 110-111 t 19 Sinc Function Appears in Both Pulse Train & a Single Pulse Pulse Train – Fourier Series Single pulse – Fourier Transform f (t ) 1 u(t 2 ) u(t 2 ) Discrete Spectrum Continuous Spectrum F f sin( ft ) ( ft ) What does this imply for a digital binary communication signal? ES 442 Fourier Transform 20 Some Insight into the Sinc Function sin( x ) sinc( x ) x 1 x sinc( x) sin( x ) x 1 x Lathi & Ding pp. 100-101 ES 442 Fourier Transform 21 Properties of Fourier Transforms 1. Linearity (Superposition) Property 2. Time-Frequency Duality Property 3. Transform Duality Property 4. Time-Scaling Property 5. Time-Shifting Property 6. Frequency-Shifting Property 7. Time Differentiation & Time Integration Property 8. Area Under g(t) Property 9. Area Under G(f) Property 10.Conjugate Functions Property ES 442 Fourier Transform 22 Linearity (Superposition) Property Given g1(t) G1(f) and g2(t) G2(f) ; Then g1(t) + g2(t) G1(f) + G2(f) also kg1(t) kG1(f) and kg2(t) kG2(f) (additivity) (homogeneity) Combining these we have, kg1(t) + mg2(t) kG1(f) + mG2(f) Hence, the Fourier Transform is a linear transformation. This is the same definition for linearity as used in your circuits and systems course. ES 442 Fourier Transform 23 Time-Frequency Duality Property Given g(t ) G( f ), then g(t - t0 ) G( f ) e and g(t)e j 2 ft0 g(t )e j 2 ft0 G( f f 0 ) j 2 ft dt g(t ) G( f ) G( f )e j 2 ft df Lathi & Ding pp. 106-109 This leads directly to the transform Duality Property ES 442 Fourier Transform 24 Transform Duality Property Given g(t ) G( f ), then g(t ) G( f ) and G(t ) g( f ) Lathi & Ding pp. 108-109 See illustration on next page for example! ES 442 Fourier Transform 25 Illustration of Fourier Transform Duality G1 ( f ) T1 Lathi & Ding Page 109 g1 (t ) T1 2 T1 2 t 1 T1 1 T1 f 2 T1 2 T1 f1 g 2 (t ) G2 ( f ) 1 T1 T1 1 T1 1 T1 t ES 442 Fourier Transform T1 2 T1 2 f 26 Time-Scaling Property Given g(t ) G( f ), then for a real constant a, f 1 g( at ) G( ) a a Lathi & Ding pp. 110-112 Time compression of a signal results in spectral expansion and time expansion of a signal results in spectral compression. ES 442 Fourier Transform 27 Time-Shifting Property Given g(t ) G( f ), then g(t t0 ) G( f )e j 2 ft0 Lathi & Ding pp. 112-113 Delaying a signal by t0 seconds does not change its amplitude spectrum, but the phase spectrum is changed by -2ft0. Note that the phase spectrum shift changes linearly with frequency f. = Frequency f = 2 Frequency 2f ES 442 Fourier Transform 28 Frequency-Shifting Property aka Modulation Property Given g(t ) G( f ), then g(t )e j 2 ft0 G( f f 0 ) Multiplication of a signal g(t) by the factor cos(2 fC ) places G(f) centered at f = fC. g(t ) 2A B g(t )cos(2 fC ) t g(t ) G( f ) g(t ) G( f ) t g(t ) Lathi & Ding pp. 114-119 B f A fC 2B ES 442 Fourier Transform fC f 29 Frequency-Shifting Property (continued) g(t ) G( f ) a g(t )cos(2 f 0 ) c g(t )sin(2 f 0 ) e G( f ) b g ( f ) Lathi & Ding Figure 3.21 Page 115 d Note phase shift f 2 2 Multiplication of a signal g(t) by the factor cos(2 f 0 ) places G(f) centered at f = fC. ES 442 Fourier Transform 30 Time Differentiation & Time Integration Property Given g(t ) G( f ) For time differentiation: dg(t ) j 2 f G( f ) dt d n g(t ) n and j 2 f G( f ) n dt For time integration: t g( )d Lathi & Ding pp. 121-123 G( f ) 1 2 G(0) ( f ) j 2 f ES 442 Fourier Transform 31 Area Under g(t) & Area Under G(f) Properties Given g(t ) G( f ), Then g(t )dt G(0) That is, the area under a function g(t) is equal to the value of its Fourier transform G(f) at f = 0. Given g(t ) G( f ), Then g(0) G( f )df That is, the value of a function g(t) at t = 0 is equal to the area under its Fourier transform G(f). ES 442 Fourier Transform 32 Conjugate Functions Property Given g(t ) G( f ), Then for a complex-valued time function g(t), we have g * (t ) G * ( f ) where the star symbol (*) denotes the complex conjugate operation. The corollary to this property is g * ( t ) G * ( f ) End of Fourier Transform Property Slides ES 442 Fourier Transform 33 Some Important Fourier Transform Pairs Impulse Function Impulse train Unit Rectangle Pulse Unit Triangle Pulse Exponential Pulse Signum Function ES 442 Fourier Transform 34 The Impulse Function (t) (t ) 0 for all t 0 ES 442 Fourier Transform 35 Fourier Transform of the Impulse Function (t) F (t ) (t )e j 2 ft dt e j 2 f (0) 1 because the exponential function to zero power is unity. Therefor, we can write (t) 1 g(t) =(t) t 1 G(f) =1 f (t) is often called the Dirac delta function ES 442 Fourier Transform 36 Inverse Fourier Transform of the Impulse Function (t) F 1 ( f ) ( f )e j 2 ft df e j(0)t 1 because the exponential function to zero power is unity. Therefor, we can write 1 (t) 1 G(f) = (f) g(t) t This is just a DC signal. 0 f A DC signal is zero frequency. ES 442 Fourier Transform 37 Fourier Transform of Impulse Train (t) aka “Dirac Comb Function,” Shah Function & “Sampling Function” Shah function (Ш(t) ): Ш(t) = (t nT ) (t nT ) 0 n n 1 n 1 Ш(t) 2T0 T0 0 2 n 0 Ш (t ) dt 1 2 Period Period T0 T0 2T0 t 2 f 0 f 0 0 ES 442 Fourier Transform f0 2 f0 1 T0 f 38 Shah Function (Impulse Train) Applications The sampling property is given by Ш(t) f (t ) n f (n) (t nT0 ) The “replicating property” is given by the convolution operation: Ш(t) f (t ) n f (t nT0 ) Convolution Convolution theorem: g1 (t ) g2 (t ) G1 ( f )G2 ( f ) and g1 (t )g2 (t ) G1 ( f )G2 ( f ) ES 442 Fourier Transform 39 Sampling Function in Operation Ш(t) f (t ) n f (n) (t nT0 ) f (t ) t (t nT0 ) T0 T0 f (0) t f (T0 ) f (1) f (2T0 ) f (2) t ES 442 Fourier Transform 40 Fourier Transform of Complex Exponentials F 1 ( f f c ) j 2 f t ( f f ) e df C Evaluate for f f c F 1 ( f f c ) e j 2 f ct df e j 2 f ct f fc ( f f c ) e j 2 f c t F 1 ( f f c ) and Lathi & Ding pp. 104 j 2 f t ( f f ) e df C Evaluate for f f c F 1 ( f f c ) e j 2 fct df e j 2 f ct f f c ( f f c ) e j 2 f c t ES 442 Fourier Transform 41 Fourier Transform of Sinusoidal Functions j 2 f t j 2 f t c c Taking ( f f c ) e and ( f f c ) e We use these results to find FT of cos(2 ft ) and sin(2 ft ) Using the identities for cos(2 ft ) and sin(2 ft ), * j 2 f c t j 2 f c t j 2 f c t j 2 f c t cos(2 ft ) 21 e e & cos(2 ft ) 21j e e Therefore, cos(2 ft ) 21 ( f f c ) ( f f c ) , sin(2 ft ) 1 2j and ( f fc ) ( f fc ) cos(2fct) Lathi & Ding pp. 105 sin(2fct) -fc -fc fc f ES 442 Fourier Transform fc f 42 Fourier Transform of Signum (Sign) Function sgn(t ) 1, 0, 1, for t 0 for t 0 for t 0 Sgn(t) +1 -1 t We approximate the signum function using g(t ) exp( at ), for t 0 0, for t 0 exp( at ), for t 0 +1 -1 j 4 f 1 G( f ) 2 , because 2 a (2 f ) j f j 4 f j 1 lim 2 2 a (2 f ) a 0 f j f ES 442 Fourier Transform t Lathi & Ding pp. 105-106 43 Summary of Several Fourier Transform Pairs Lathi & Ding Table 3.1 Page 107; For a more complete Table of Fourier Transforms -----------See also the Fourier Transform Pair Handout ES 442 Fourier Transform 44 Spectrum Analyzer Shows Frequency Domain A spectrum analyzer measures the magnitude of an input signal versus frequency within the full frequency range of the instrument. It measures frequency, power, harmonics, distortion, noise, spurious signals and bandwidth. It is an electronic receiver Measure magnitude of signals Does not measure phase of signals Complements time domain Bluetooth spectrum ES 442 Fourier Transform 45 Fourier Transform of Cosine Signal A cos(2 f c t ) A ( f f c ) ( f f c ) 2 A 2 Re A cos(2 f t ) -fc Not in Lathi & Ding A 2 3D View fc f Blue arrows indicate positive phase direction ES 442 Fourier Transform 46 Fourier Transform of Sine Signal B sin(2 f c t ) j B ( f f c ) ( f f c ) 2 Re B 2 B sin(2 f t ) -fc B 2 fc f We must subtract 90 from cos(···) to get sin(···) ES 442 Fourier Transform 47 Fourier Transform of Sine Signal (as usually shown in books) ( f0 ) sin(t ) t FT Imaginary axis A j 2 f0 f0 j ES 442 Fourier Transform f A 2 48 Fourier Transform of Phase Shifted Sinusoidal Signal j j 2 ft Re e Re A 2 j 2 ft - B 2 Re j j 2 ft e Re -fc Re A 2 j 2 ft B 2 fc f 2 A B R 2 2 2 A and tan 1 B ES 442 Fourier Transform 49 Even, Odd Relationships for Fourier Transform (1) g(t) G(f) Re Im t Im Real; Even g(t) G(f) t Real; Odd f Real; Even Re Im Re Re Im f Imaginary; Odd From: Frederic J. Harris, Trignometric Transforms, Spectral Dynamics Corporation, 1976. ES 442 Fourier Transform 50 Even, Odd Relationships for Fourier Transform (2) g(t) Re G(f) t Im Imaginary; Even g(t) f Im Imaginary; Even Re Im Re G(f) t Re f Im Imaginary; Odd Real; Odd From: Frederic J. Harris, Trignometric Transforms, Spectral Dynamics Corporation, 1976. ES 442 Fourier Transform 51 Effect of the Position of a Pulse on Fourier Transform G( f ) Re(G( f )) Im(G( f )) 2 2 t Even function. f Both must be identical. f t This time shifted pulse is both even and odd. ES 442 Fourier Transform 52 Selected References 1. Paul J. Nahin, The Science of Radio, 2nd edition, Springer, New York, 2001. A novel presentation of radio and the engineering behind it; it has some selected historical discussions that are very interesting. 2. Keysight Technologies, Application Note 243, The Fundamentals of Signal Analysis; http://literature.cdn.keysight.com/litweb/pdf/59528898E.pdf?id=1000000205:epsg:apn 3. Agilent Technologies, Application Note 150, Spectrum Analyzer Basics; http://cp.literature.agilent.com/litweb/pdf/5952-0292.pdf 4. Ronald Bracewell, The Fourier Transform and Its Applications, 3rd ed., McGraw-Hill Book Company, New York, 1999. I think this is the best of all books covering the Fourier Transform (Bracewell gives many insightful views and discussions on the FT). ES 442 Fourier Transform 53