Signal and linear system analysis

• Time-frequency viewpoint
 Signal models
 Signal classifications
 Fourier series
 Fourier transform
 Discrete Fourier transform and fast Fourier transform
• Power spectral density and correlation
• Signals and linear systems
 linear, time-invariant, two-port systems
• Sampling theory
• Hilbert transform
 Signal and linear system analysis- signal models
• Deterministic and Random Signals
 Deterministic signals：
Unit rectangular pulse：(t)

 Random signals：

• Periodic signals：
• Rotating phasor (periodic signal) distinguish it
from the phasor 𝐴𝑒 𝑗𝜃
• Complex conjugate rotating phasors
 Signal and linear system analysis- signal models

even odd

Amplitude and phase spectra for the (a) Single-sided. (b) Double-sided.
• Singularity Functions→ aperiodic signals
 unit impulse function 𝛿(𝑡)
 Signal and linear system analysis- Signal classifications
 unit step function 𝑢(𝑡)

• Signal classifications
periodic signal：

 𝑥(𝑡) ：energy signal iff 0 < 𝐸< ∞ , so that 𝑃 = 0

 𝑥(𝑡) ：power signal iff 0 < 𝑃< ∞ → 𝐸 = ∞
 𝐸 = 𝐴2 ∕2𝛼 → energy signal
 𝑥2 (𝑡) = 𝐴𝑢(𝑡) P=(½ )𝐴2 → power signal
 Rotating phasor signal
→ power signal

 Sinusoidal signal：𝑥𝑝 (𝑡) = 𝐴 cos(𝜔0 𝑡+ 𝜃) P=(½ )𝐴2 → power signal

 Signal and linear system analysis- Fourier series(1)
• Complex Exponential Fourier Series：𝑥(𝑡) is periodic with period 𝑇0 ， 𝜔0 = 2𝜋𝑓0

2 1 𝑗𝜔 𝑡 1 −𝑗 𝜔 𝑡 1 1 𝑗 4𝜔 𝑡 1 −𝑗 4𝜔 𝑡
 Ex：𝑥(𝑡) = cos(𝜔0 𝑡)+ 𝑠𝑖𝑛 (2𝜔0 𝑡)= 𝑒 0 + 𝑒 0 + − 𝑒 0 − 𝑒 0
2 2 2 4 4
• Symmetry Properties of the Fourier Coefficient : 𝑥(𝑡) is real
 𝑋𝑛∗ =𝑋−𝑛 , 𝑋𝑛 =|𝑋𝑛 |𝑒 𝑗∠𝑋𝑛 , |𝑋𝑛 |=|𝑋−𝑛 | , ∠𝑋𝑛 =-∠𝑋−𝑛
Ex : 𝑥(𝑡) = 𝑥(-𝑡) even

Ex : 𝑥(𝑡) = -𝑥(-𝑡) odd

 Signal and linear system analysis- Fourier series(2)
• Trigonometric Form of the Fourier Series
𝑥(𝑡) is real , 𝑋𝑛 𝑒 𝑗n𝜔0𝑡 +𝑋−𝑛 𝑒 −𝑗n𝜔0𝑡 =|𝑋𝑛 |𝑒 𝑗(n𝜔0𝑡+∠𝑋𝑛 ) +|𝑋𝑛 |𝑒 −𝑗(n𝜔0𝑡+∠𝑋𝑛 )
=2|𝑋𝑛 |cos(n𝜔0 𝑡 + ∠𝑋𝑛 )
→ )

𝐴𝑛

𝑋0 : DC 𝑛 = 1 is called the fundamental 𝐵𝑛

𝑛 = 2 is called the second harmonic
• Parseval’s Theorem：
 Average power of a periodic signal 𝑥(𝑡) is the sum of the powers in the phasor
components of its Fourier series
Signal and linear system analysis- Fourier series(3)
Signal and linear system analysis- Line Spectra(1)

Line spectra for half-rectified

sinewave (Double-sided)

Line spectra for half-rectified

sinewave (single-sided)
 Signal and linear system analysis- Line Spectra(2)

1
  T0
4

1
  T0
8

1
  T0
8
 Signal and linear system analysis- Fourier transform(1)
• 𝑥(𝑡) : aperiodic energy signal
 
Fourier transform of 𝑥(𝑡)→ X ( f )  [ x(t )]   x(t )e  j 2 ft
dt or X ( )   x(t )e  jt dt ,   2 f
 
 1 
inverse Fourier transform→ x(t )   [ X ( f )]   X ( f )e
1 j 2 ft
df or x(t )   X ( )e jt d 
2 

• Amplitude and Phase Spectra X ( f ) | X ( f ) | e j ( f ) ,  ( f )  X ( f )

 Amplitude spectrum | X ( f ) || X ( f ) |
 Phase spectrum  ( f )   ( f )
• Symmetry Properties
 t  t0  t0  / 2  j 2 ft
 Ex： x(t )  A    X ( f )  A t0  / 2 e dt  A sin c( f  )e  j 2 ft0
  
 2 t0 f sin c( f  )  0
| X ( f ) | a | sin c( f  ) ，
| ( f )   ,
 2 t0 f   sin c( f  )  0
t0   / 2
 Signal and linear system analysis- Energy spectral density
• Parseval’s theorem for Fourier transforms：

• Energy spectral density of a signal：

 Ex：Energy of 𝑥(𝑡)?

 10
E   | X ( f ) | df  
2
4df  80J
 10


sin c 2 (u )du  1
 
E   | x(t ) | dt   | 40sin c(20t ) |2 dt  80J
2
 

8W W  10
2
W W   f 
EW   | X ( f ) | df  2 2
2   20   df   80 W  10
W 0
   
 Signal and linear system analysis- Fourier transform(2)

Convolution：x(t )  x1 (t ) * x2 (t )   x1 ( )x2 (t   )d • Superposition Theorem
a1 x1 (t )  a2 x2 (t )  a1 X 1 ( f )  a2 X 2 ( f )
Ex：x1 (t )  e t u(t，
) x2 (t )  e   t u(t ),     0
• Time-Delay Theorem
x(t  t0 )  X ( f )e j 2 ft0
• Scale-Change Theorem
1 f
x( at )  X( )
|a| a
• Duality Theorem
X (t )  x(  f )
• Frequency-Translation Theorem
x ( t ) e j 2 f 0 t  X ( f  f 0 )
• Modulation Theorem
x(t )cos(2 f 0t ) 
1 1
X ( f  f0 )  X ( f  f0 )
2 2
 Signal and linear system analysis- Fourier transform(3)
• Differentiation Theorem
d n x(t )
n
 ( j 2 f ) n
X( f )
dt
• Integration Theorem
X( f ) 1
 x( )d    X (0) ( f )
t

j 2 f 2
• Convolution Theorem
 
 x ( )x2 (t   )d   x1 (t   )x2 ( )d
 1

 X1( f ) X 2 ( f )
• Multiplication Theorem
x1 (t ) x2 (t ) 

X 1 ( f )  X 2 ( f )   X 1 ( )X 2 ( f   )d 
 Signal and linear system analysis- Fourier transform(4)
 f 
Ex : Use the duality theorem to show that 2 AW sin c(2Wt )  A  
 2W 
t
 x(t )      A sin c( f  )  X ( f )
 
 f 
 X (t )  A sin c( t )  A     x(  f )，  2W
  
 f 
 A2W sin c(2Wt )  A   Q.E.D.
 2W 
Ex：Fourier-transform pairs:
A (t )  A
A (t  t0 )  Ae j 2 ft0
A  A ( f )
Ae j 2 ft0  A ( f  f 0 )
Signal and linear system analysis- Fourier transform(5)

t  t 
    sin c ( f  )       sin c 2
( f )
   
 Signal and linear system analysis- Fourier transform(6)
• ideal sampling waveform：

Q.E.D.

• cosinusoidal pulse 

Q.E.D.
 Signal and linear system analysis- Fourier transform(7)
• Fourier Transforms of Periodic Signals：in a strict mathematical sense, does not exist, since
periodic signals are not energy signals
• 𝑥(𝑡) is a periodic power signal
 Signal and linear system analysis- Fourier transform(8)


• Output of a radar transmitter：

a periodic cosinusoidal pulse train,
 Power spectral density and correlation
• Power spectral density 𝑆(𝑓)：real, even, nonnegative function of frequency

 Cosinusoidal signal , its average power per ohm 1/2𝐴2

• Time-Average Autocorrelation Function  

inverse Fourier transform of energy spectral density 𝐺(𝑓)

signal energy E=0

• Time-average autocorrelation function 𝑅(𝜏) of a power signal 𝑥(𝑡)

• Wiener-Khinchine theorem：
 Power spectral density and correlation
• Properties of R(𝜏)

• Ex：pseudo-noise or m-sequence 1110010

 Signals and linear systems (1)
• A single-input, single-output system：𝑦(𝑡) = H [𝑥(𝑡)]
• Linear Time-Invariant (LTI) System
 Linear ：

 Time-invariant ：𝑦(𝑡 − 𝑡0 ) = H[𝑥(𝑡 − 𝑡0 )]

 impulse response ℎ(𝑡) of an LTI system ：ℎ(𝑡) ≜ H[𝛿(𝑡)] Signal


Approximation with a sequence of impulses

 
 Signals and linear systems -properties(2)
• Stability：A linear system is bounded-input, bounded-output (BIBO) 
• Transfer Function：
 𝑌(𝑓) = 𝐻(𝑓)𝑋(𝑓) 

• Causality ： ℎ(𝑡) = 0, 𝑡<0

Ex：ℎ (𝑡) = 𝑒 −2𝑡 cos (10𝜋𝑡) 𝑢 (𝑡) is stable and causal.

The system is causal because ℎ (𝑡) = 0 for 𝑡 < 0.

• Symmetry Properties of H (𝑓)：
H ( f ) | H ( f ) | e jH ( f ) ， | H ( f ) || H (  f ) | ，  H ( f )  H (  f )
| H ( f ) | ：Amplitude- (magnitude) response function，even
H ( f )：phase response function，odd
 Signals and linear systems - Low-pass RC filter (3)
• Low-pass RC filter：  (𝑗2𝜋𝑓𝑅𝐶 + 1)𝑌(𝑓) = 𝑋(𝑓)

𝑓3 = 1∕(2𝜋𝑅𝐶)

= 𝐴[𝑢 (𝑡) − 𝑢 (𝑡 − 𝑇 )]

spectra
𝑇 ∕𝑅𝐶 = 0.5

spectra spectra
𝑇 ∕𝑅𝐶 = 2 𝑇 ∕𝑅𝐶 = 10
 Signals and linear systems (4)
• Input-Output Relationships for Spectral Densities：
 Energy spectral densities ： 𝐺𝑥(𝑓) =|𝑋 (𝑓)|2 and 𝐺𝑦(𝑓) = |𝑌 (𝑓)|2 ， 𝑌 (𝑓) = 𝐻 (𝑓)𝑋 (𝑓)
𝐺𝑦(𝑓) = |𝐻 (𝑓)|2 𝐺𝑥(𝑓)
 Power spectral densities：S𝑦(𝑓) = |𝐻 (𝑓)|2 S𝑥(𝑓)
• Response to Periodic Inputs→ steady-state response
 input signal： 𝐴𝑒 𝑗2𝜋𝑓0𝑡

 steady-state output due to an arbitrary periodic input is


y (t )  |X n | | H (nf 0 ) | exp  j  2 nf 0t  X n  H (nf 0 ) 
n 

 X 0 H (0)  2  | X n | | H (nf 0 ) | cos 2 nf 0t  X n  H ( nf 0 )
n 

作業： unit-amplitude triangular signal with period 0.1 s, ,

 Signals and linear systems (5)
• Distortionless Transmission：𝑦(𝑡) = 𝐻0𝑥(𝑡− 𝑡0) delayed replica of the input (𝑡0 > 0 for causality)
→ Frequency response function 𝐻(𝑓) = 𝐻0𝑒 −𝑗2𝜋𝑓𝑡0
→ amplitude response is constant and phase shift is linear with frequency.
• Group Delay

, 𝜃(𝑓) is the phase response of the system

 Distortionless system：𝜃(𝑓) = −2𝜋𝑓𝑡0 , 𝑇𝑔(𝑓) = 𝑡0
Ex：A system with amplitude response and phase shift as shown
 Signals and systems - Nonlinear (6)
• Nonlinear Distortion
Ex：𝑥(𝑡) = 𝐴1 cos (𝜔1𝑡) + 𝐴2 cos (𝜔2𝑡)，
𝑦(𝑡) = 𝑎1 𝑥(𝑡) + 𝑎2 𝑥 2 (𝑡) = 𝑎1 [𝐴1 cos (𝜔1𝑡) + 𝐴2 cos (𝜔2𝑡)] + 𝑎2 [𝐴1 cos (𝜔1𝑡) + 𝐴2 cos (𝜔2𝑡)]2

harmonic distortion terms Intermodulation distortion terms

 distortion terms at harmonics of the input frequencies (in this case, second) as well as distortion terms
involving sums and differences of the harmonics (in this case, first) of the input frequencies

𝑌(𝑓) = 𝑎1𝑋(𝑓) + 𝑎2𝑋(𝑓) ∗ 𝑋(𝑓)

 It is impossible to isolate harmonic and intermodulation distortion components

Ex：If input
𝑌(𝑓) = 𝑎1𝑋(𝑓) + 𝑎2𝑋(𝑓) ∗ 𝑋(𝑓)
distortion term→
 Signals and linear systems - Ideal Filters (7)
• Ideal Filters：low-pass (LPF) , high-pass (HPF) , band-pass (BPF)
 𝐵 ：the single-sided bandwidth
 ideal LPF：𝐻LP(𝑓) = 𝐻0Π(𝑓∕2𝐵)𝑒−𝑗2𝜋𝑓𝑡0
 ideal HPF：𝐻HP(𝑓) = 𝐻0 [1 − Π(𝑓∕2𝐵)] 𝑒−𝑗2𝜋𝑓𝑡0
 Ideal BPF：𝐻BP(𝑓) = [𝐻1(𝑓 − 𝑓0) + 𝐻1(𝑓 + 𝑓0)] 𝑒−𝑗2𝜋𝑓𝑡0 where 𝐻1(𝑓) = 𝐻0Π(𝑓∕𝐵).

ℎLP(𝑡) = 2𝐵𝐻0 sinc [2𝐵(𝑡 − 𝑡0)] ℎBP(𝑡) = 2ℎ1(𝑡 − 𝑡0) cos [2𝜋𝑓0(𝑡 − 𝑡0)]

ℎBP(𝑡) = 2𝐻0𝐵 sinc [𝐵(𝑡 − 𝑡0)] cos [2𝜋𝑓0(𝑡 − 𝑡0)]

 Signals and linear systems -Approximation of Ideal LPF(8)
• Ideal Filters：noncausal and unrealizable
 practical filter types that may be designed to approximate ideal filter characteristics as
closely as desired.
(1) Butterworth LPF：
constant amplitude response in the passband
at the cost of less stopband attenuation

Pole locations and amplitude responses for

fourth-order Butterworth
Ex1：

Ex2：Amplitude response of 𝑛th-order Butterworth filter：

(2) Chebyshev (type 1) LPF：
minimum allowable attenuation in the passband Pole locations and amplitude responses
for fourth-order Chebyshev filters
maximizing the attenuation in the stopband.

Chebyshev polynomial：
Signals and linear systems -Approximation of Ideal LPF(9)
(3) Bessel LPF：
maintain a linear phase response in the passband
at the expense of the amplitude response.
𝑡0 is the nominal delay of the filter

𝐾𝑛 is a constant chosen to yield 𝐻(0) = 1,

𝐵𝑛(𝑓) is a Bessel polynomial

amplitude responses show that the Chebyshev filters have more

attenuation than the Butterworth and Bessel filters do for
frequencies exceeding the 3-dB frequency.

Bessel filter has the most constant group delay Amplitude-response and group-delay characteristics of third order
Butterworth, Bessel, and Chebyshev filters
 Signals and linear systems - Pulse Resolution (10)
t   f 
Ideal time-limited waveform x1 (t )      T sin c ( fT )  X 1 ( f ) ， Ideal band-limited waveform X 2 ( f )  A    2 AW sin c(2Wt )  x2 (t )
T   2W 
• Relationship of Pulse Resolution and Rise-time to Bandwidth

• Pulses with band-pass spectra 𝑥(𝑡) = 𝐴Π(𝑡/𝜏)cos(𝜔0𝑡) ： 𝑊 ≥1/𝑇 Hz

• Rise-time ：time required for a pulse’s leading edge to go from 10% to 90% of final value
 Unit input：𝑢(𝑡− 𝜎)
 step response of a filter
 Signals and linear systems - Rise-time to Bandwidth (11)
Ex：Impulse response of a low-pass RC filter
1
ℎ(𝑡) = 𝑒 −𝑡∕𝑅𝐶 𝑢(𝑡)
𝑅𝐶
step response ：𝑦𝑠(𝑡) =(1 − 𝑒−2𝜋𝑓3𝑡)𝑢(𝑡)
0.35
rise-time 𝑇𝑅 = = 2.2𝑅𝐶
𝑓3

Ex：Step response of an ideal low-pass filter

𝑢 = 2𝜋𝐵(𝜎 − 𝑡0)

sine-integral function：
 Signals and linear systems - Sampling theory(1)
• Sampling theory ：𝑥(𝑡) →𝑥𝛿(𝑡) → 𝑦(𝑡) = 𝑥(𝑡)
 application in control systems and pulse-modulation communication systems

1
 Sampling with 𝑇𝑠 < W → ，W：BW of signal 𝑥(𝑡)
2
 Reconstruction 𝑥𝛿(𝑡) → 𝑥(𝑡)
1
𝑇𝑠 < W (𝑓𝑠 > 2𝑊) and ideal low-pass filter with bandwidth 𝐵 , where W< 𝐵<𝑓𝑠 −𝑊 with 𝑓𝑠 = 1/𝑇𝑠
2
𝑓
𝐻LP(𝑓)= 𝐻0Π( )𝑒−𝑗2𝜋𝑓𝑡0
2𝐵

 Nyquist frequency：2W

→ →𝑌 (𝑓) = 𝑓𝑠𝐻0𝑋(𝑓)𝑒−𝑗2𝜋𝑓𝑡0 →𝑦(𝑡) =𝑓𝑠𝐻0𝑥(𝑡−𝑡0) distortionless

 Signals and linear systems - Sampling theory(2)

aliasing error in the reconstruction of sampled signal error due to non-ideal reconstruction filter
𝑥(𝑡) is not bandlimited or because 𝑓𝑠 < 2𝑊 error can be minimized by choosing reconstruction filters with
sharper roll-off characteristics or by increasing the sampling rate.

1
， ，if 𝐵 = 𝑓𝑠, 𝐻0 = 𝑇𝑠, and 𝑡0 = 0
2

，
 Signals and linear systems - Sampling theory(3)
• If band-pass spectra signal with upper limit on frequency 𝑓𝑢 ( >> bandwidth W )
 uniform sampling theorem for band-pass spectra
 sampling rate 𝑓𝑠=2𝑓𝑢 ∕𝑚，𝑚 is the largest integer not exceeding 𝑓𝑢 ∕𝑊⎿𝑓𝑢 ∕𝑊⏌ ⎿3.5⏌=3

Band-pass signal spectrum

fs=2fu /m=2(3)/2=3
 Signals and linear systems - Hilbert transform(1)
• 𝐻(𝑓) = −𝑗 sgn 𝑓，sgn 𝑓 : signum 𝑓，|𝐻(𝑓)|= 1 ，𝐻(𝑓) : odd

lim 𝐺(𝑓; 𝛼) = 𝑠𝑔𝑛 𝑓

𝛼→0

𝛼→0

• The Hilbert transform of 𝑥(𝑡)

 (−𝑗 sgn 𝑓)2 = −1

 Signals and linear systems - Hilbert transform(2)
1 1
Ex：𝑥(𝑡) = cos (2𝜋𝑓0𝑡)，𝑋(𝑓) = 𝛿(𝑓
− 𝑓0) + 𝛿(𝑓
+ 𝑓0)，
2 2
output spectrum from the Hilbert transformer

Ex：𝑥(𝑡) = sin (2𝜋𝑓0𝑡)，


• Properties：
1. energy (or power) in a signal 𝑥(𝑡) and 𝑥̂(𝑡) are equal

2. A signal and its Hilbert transform are orthogonal

= 0 ，hint ，odd× even
 Signals and linear systems - Hilbert transform(3)
3. 𝑐(𝑡) and 𝑚(𝑡)：non-overlapping spectra, 𝑚(𝑡) is low-pass and 𝑐(𝑡) is high-pass

prove ：


𝑀(𝑓)𝐶(𝑓′) is nonvanishing only for |𝑓|< 𝑊 and|𝑓′|> 𝑊
replace sgn (𝑓 + 𝑓′) by sgn (𝑓′)

Ex：𝑓0 = 𝜔0∕2𝜋 >𝑊，

 Signals and linear systems - Hilbert transform(4)
• Analytic Signals
 analytic signal 𝑥𝑝(𝑡), corresponding to the real signal 𝑥(𝑡), is defined as 𝑥𝑝(𝑡) = 𝑥(𝑡) +j𝑥̂(𝑡)
Ex：Ideal band-pass filter with bandwidth 𝐵, delay 𝑡0, and center frequency 𝑓0 is given
ℎBP(𝑡) = 2𝐻0𝐵 sinc [𝐵(𝑡 − 𝑡0)] cos [𝜔0(𝑡 − 𝑡0)]，𝐵 < 𝑓0,
Hilbert transform of ℎBP(𝑡)
Envelope
|ℎBP(𝑡)|

• Fourier transform of 𝑥𝑝(𝑡)𝑋𝑝(𝑓) = 𝑋(𝑓) + 𝑗 {−𝑗 sgn (𝑓)𝑋(𝑓)}

= 𝑋(𝑓) [1 + sgn 𝑓]

• Fourier transform of 𝑥n(𝑡) = 𝑥(𝑡) - j𝑥̂(𝑡)

𝑋n(𝑓) = 𝑋(𝑓) − 𝑗 {−𝑗 sgn (𝑓)𝑋(𝑓)} = 𝑋(𝑓) [1 - sgn 𝑓]
 Signals and linear systems - Hilbert transform(5)
• Complex Envelope Representation of Band-pass Signals：
• Complex-valued lowpass signal (hereafter referred to as the complex envelope)：

• ，

𝑥𝑅(𝑡) and 𝑥𝐼 (𝑡) ：the in-phase and quadrature components of 𝑥(𝑡).

 Signals and linear systems - Hilbert transform(6)
Ex：The real bandpass signal 𝑥(𝑡) = cos (22𝜋𝑡)，Its Hilbert transform is 𝑥̂(𝑡) = sin(22𝜋𝑡)
𝑥𝑝(𝑡) = 𝑥(𝑡) +j 𝑥̂(𝑡) = cos(22𝜋𝑡) + 𝑗 sin (22𝜋𝑡) = 𝑒𝑗22𝜋𝑡 , 𝑓0 = 10 Hz
= 𝑥𝑝(𝑡) 𝑒−𝑗2𝜋𝑓0𝑡= 𝑒𝑗22𝜋𝑡𝑒−𝑗20𝜋𝑡= 𝑒𝑗2𝜋𝑡= cos(2𝜋𝑡) + 𝑗 sin (2𝜋𝑡)
𝑥𝑅(𝑡) = cos (2𝜋𝑡) & 𝑥𝐼(𝑡) = sin(2𝜋𝑡)
𝑥(𝑡) = 𝑥𝑅(𝑡) cos(2𝜋𝑓0𝑡) − 𝑥𝐼 (𝑡) sin(2𝜋𝑓0𝑡)
= cos(2𝜋𝑡) cos (20𝜋𝑡) − sin(2𝜋𝑡) sin (20𝜋𝑡) = cos(22𝜋𝑡)
• Complex Envelope Representation of Band-pass Systems
 bandpass system with impulse response ℎ(𝑡), complex envelope
 Input x(t) is also bandpass
 Output 𝑦(𝑡) = 𝑥(𝑡) ∗ ℎ(𝑡)
c.c. stands for the complex conjugate
of the immediately preceding term

𝑦(𝑡)
 Signals and linear systems - Hilbert transform(7)
Ex：The bandpass signal 𝑥(𝑡) = Π(𝑡∕𝜏) cos(2𝜋𝑓0𝑡)，impulse response ℎ(𝑡) = 𝛼𝑒−𝛼𝑡𝑢(𝑡) cos(2𝜋𝑓0𝑡)

𝑦(𝑡)

the second integral is approximately zero

 Signals and linear systems – DFT & FFT
• Discrete Fourier transform (DFT) (𝑁2 complex × ) & fast Fourier transform(FFT) (𝑁 log2 𝑁 complex ×)
 compute the Fourier spectrum of a signal by means of a digital computer
 Fourier spectrum of a signal at frequencies 𝑘∕(𝑁𝑇𝑠), 𝑘 = 0, 1,…,𝑁 − 1，𝑓𝑠 = 𝑇1 ，T=𝑁𝑇𝑠
𝑠
 DFT： discrete set of frequencies {0, 1∕𝑇 , 2∕𝑇 ,…, (𝑁 − 1)∕𝑇 }.

 IDFT：

Ex：

𝑇𝑠=T/𝑁

Calculating the DFT,

x(t) multiplied by a window function Π(𝑡∕𝑇 )
X(f)*𝑇sinc (𝑓𝑇) in the frequency domain,(𝑇 = 4+s). Xsp( f )
Sampled signal spectrum and corresponding periodic
repetition of the sampled, windowed signal.