3.0 fourier series representation of periodic signals 3.1 exponential/sinusoidal signals as building...
TRANSCRIPT
3.0 Fourier Series Representation of Periodic Signals
3.1 Exponential/Sinusoidal Signals as Building Blocks for Many Signals
Time/Frequency Domain Basis SetsTime Domain
Frequency Domain
𝑥 (𝑡)
𝑡
𝑡
𝑡
𝑛
𝑛𝑛
𝑥 [𝑛 ]
{𝑒 𝑗𝜔𝑛 ,𝜔∈ [0 ,2𝜋 ]}
Response of A Linear Time-invariant System to An Exponential Signal
Initial Observation
– if the input has a single frequency component, the output will be exactly the same single frequency component, except scaled by a constant
time-invariant
scaling property
tyety
tyetxe
tytxe
tytxe
j
jj
tj
tj
0
00
0
0
tjeyty 00 0 t
Input/Output Relationship𝑦 (𝑡)𝑥 (𝑡)
Time Domain
Frequency Domain
𝛿(𝑡) h (𝑡)0 0𝑡 𝑡
matrix vectors
eigen valueeigen vector
Response of A Linear Time-invariant System to An Exponential Signal
More Complete Analysis– continuous-time
stsst
sstts
st
esHdehe
dehedeh
dtxhty
jrsetx
, 0
Response of A Linear Time-invariant System to An Exponential Signal
More Complete Analysis– continuous-time
sH
etx
dehsH
st
s
Transfer FunctionFrequency Response
: eigenfunction of any linear time-invariant system
: eigenvalue associated with the eigenfunction est
Response of A Linear Time-invariant System to An Exponential Signal
More Complete Analysis– discrete-time
k
k
nk
k
n
k
kn
k
jn
zkhzH
zzHzkhz
zkhknxkhny
cezznx
, 0
Transfer FunctionFrequency Response
eigenfunction, eigenvalue
System Characterization
Superposition Property– continuous-time
k k
tskk
tsk
kk esHatyeatx
k
nkkk
k
nkk zzHanyzanx
– discrete-time
– each frequency component never split to other frequency components, no convolution involved
– desirable to decompose signals in terms of such eigenfunctions
3.2 Fourier Series Representation of Continuous-time Periodic Signals
Fourier Series Representation
Harmonically related complex exponentials
,Ttxtx
T
et otjk
k 2 ,..... 2, 1, 0,k ,0
k
Ttk period with
all with period T
T: fundamental period
Harmonically Related Exponentials for Periodic Signals
[n]
[n](N)
(N)integer multiples of ω0
‧Discrete in frequency domain
Fourier Series RepresentationFourier Series
k
kkk
tjkk taeatx 0
ta jj : j-th harmonic components
real
kk aa
kkkk
kk
k
kjkkkk
jCBatkCtkBa
eAatkAatx
,sincos2
,cos2
1000
100
tx
Fourier Series RepresentationDetermination of ak
, 1
, 1
0
0
dttxT
a
dtetxT
a
T
T
tjnn
Fourier series coefficients
dc component
nk
nkTdte
dteeadtetx
T
tnkj
Tk
tjntjkk
tjn
T
0,
,
0
000
Fourier Series RepresentationVector Space Interpretation
– vector space
Ttxtx period with periodic is :could be a vector space
some special signals (not concerned here) may need to be excluded
dttxtxtxtxT
2121
Fourier Series RepresentationVector Space Interpretation
– orthonormal basis
......,2 ,1 ,0 ,1
,
,0
21
kttT
jiT
jitt
kk
ji
is a set of orthonormal basis expanding a vector space of periodic signals with period T
Fourier Series RepresentationVector Space Interpretation
– Fourier Series
dtetxT
tT
txT
ttxT
a
tatxT
tatx
tjn
T
n
nn
kkk
kkk
0 1
11
1
1
21
21
'21
'21
Fourier Series RepresentationCompleteness Issue
– Question: Can all signals with period T be represented this way?
Almost all signals concerned here can, with exceptions very often not important
Fourier Series RepresentationConvergence Issue
– consider a finite series
22
,0
tedtteE
txtxteeatx
NT NN
NN
N
Nk
tjkkN
It can be shown
NkdtetxT
aE tjk
TkN ,...1,0 , 1 if min 0
ak obtained above is exactly the value needed even for a finite series
Fourier Series RepresentationGibbs Phenomenon
– the partial sum in the vicinity of the discontinuity exhibit ripples whose amplitude does not seem to decrease with increasing N
See Fig. 3.9, p.201 of text
Fourier Series RepresentationConvergence Issue
– x(t) has no discontinuities
Fourier series converges to x(t) at every t
x(t) has finite number of discontinuities in each period
Fourier series converges to x(t) at every t except at the discontinuity points, at which the series converges to the average value for both sides
Fourier Series RepresentationConvergence Issue
– Dirichlet’s condition for Fourier series expansion
(1) absolutely integrable,
(2) finite number of maxima & minima in a period
(3) finite number of discontinuities in a period
dttxT
3.3 Properties of Fourier Series
kFS atx
Linearity
kk
FS
kFS
kFS
BbAatBytAx
btyatx
,
Time Shift
ktjkFS aettx 00
0
phase shift linear in frequency with amplitude unchanged
Time Reversal
tkjtke
atxtjk
kFS
00 sincos0
the effect of sign change for x(t) and ak are identical
Time Scaling
::
tx positive real number
k
tjkkeatx 0
periodic with period T/α and fundamental frequency αω0
ak unchanged, but x(αt) and each harmonic component are different
Differentiation
kFS ajk
dttdx
0
Parseval’s Relation
k
kTadttx
T221
2201
kT
tjkk adtea
T
average power in the k-th harmonic component in a period T
total average power in a period T
but
3.4 Fourier Series Representation of Discrete-time Periodic Signals
Fourier Series Representation
Harmonically related signal sets
NNnxnx period with periodic ,
..... 2, 1, 0,k ,2
n
Njk
k en
all with period N
N 2 , 0
,nn krNk only N distinct signals in the set
else 0,
..2 ,0 ,2
N,....N, kNeNn
nN
jk
Harmonically Related Exponentials for Periodic Signals
[n]
[n](N)
(N)integer multiples of ω0
‧Discrete in frequency domain
(P. 11 of 3.0)
Exponential/Sinusoidal SignalsHarmonically related discrete-time signal sets
all with common period N
This is different from continuous case. Only N distinct signals in this set.
}2,........1,0,k , ][{)
2(
n
Njk
k en
][][ nn kNk
(P.42 of 1.0)
Orthogonal Basis
‒ N different values in
N-dimensional vector space
‒
𝑎𝑘= �⃗� ⋅𝑣𝑘
�⃗�=∑𝑘
𝑎𝑘𝑣𝑘 (合成 )
(分析 )
Nn
nN
jk
Nn
njkk
Nk
nN
jk
kNk
njkk
enxN
enxN
a
eaeanx
2
2
1 1 0
0
,krNk aa repeat with period N
Note: both x[n] and ak are discrete, and periodic with period N, therefore summed over a period of N
Fourier Series RepresentationFourier Series
Fourier Series RepresentationVector Space Interpretation
Nnxnx period with periodic is ,
is a vector space
else 0,
,
2121
rNjiNnn
kxkxnxnx
ji
Nk
Fourier Series RepresentationVector Space Interpretation
Nknn
N kk ,1 21
a set of orthonormal bases
njk
Nn
kk
Nkkk
enxN
nN
nxN
a
nanx
0 1
11 21
21
Fourier Series RepresentationNo Convergence Issue, No Gibbs
Phenomenon, No Discontinuity
– x[n] has only N parameters, represented by N coefficients
sum of N terms gives the exact value
– N odd – N even
2
1 if ,
2
NMnxnx
eanx
M
M
Mk
nN
jk
kM
2
if ,
1
2
NMnxnx
eanx
M
M
Mk
nN
jk
kM
See Fig. 3.18, P.220 of text
PropertiesPrimarily Parallel with those for continuous-time
Multiplication
kFS anx
Njjkjk
FS
kFS
kFS
badnynx
bnyanx ,
periodic convolution
PropertiesFirst Difference
Parseval’s Relation
kN
jkFS aenxnx
2
11
22
1
Nk
kNk
anxN
average power in a period
average power in a period for each harmonic component
3.5 Application ExampleSystem Characterization
y[n], y(t)h[n], h(t)
x[n], x(t)δ[n], δ(t)
n
n
zn , e st H(z)zn, H(s)est, z=e jω, s=j e jωn , e jωt H(e jω)e jωn, H(j)e jωt
Superposition Property– Continuous-time
– Discrete-time
0
000
jkHaa
ejkHatyeatx
kk
tjk
k kk
tjkk
Njk
kk
nN
jk
Nk
Njk
kNk
nN
jk
k
eHaa
eeHanyeanx
2
222
– H(j), H(ejω) frequency response, or transfer function
Filteringmodifying the amplitude/ phase of the different frequency components in a signal, including eliminating some frequency components entirely
– frequency shaping, frequency selective
Example 1
j
j
aeeH
nxnayny
111
See Fig. 3.34, P.246 of text
Examples
• Example 3.5, p.193 of text
dte T
a
TT
dt T
a
T
-T
T
-T
tjkk
0
1
1
1
21 10
0,
sin 10 kk
Tk
Examples
• Example 3.8, p.208 of text
0,
sin ,0,
2
1
,
,
,,,
1010
0
111010
kk
Tkck
TT
c
Ta
cjkbtgdtdtq
aeaebTtxTtxtq
ctgbtqatx
k
k
kk
kTjk
kTjk
k
kkk
• Example 3.17, p.230 of text
Examples
Nj
j
nN
jN
jnN
jN
j
jnj
n
nj
nN
jnN
j
n
ere
Nnr
eeHeeHny
eeeH
eeN
nnx
nunh
2
)2(2)2(2
0
)2()2(
1
1 where
2cos
21
21
11
21
212cos
1 ,
)(
)()(
)(
y[n], y(t)h[n], h(t)
x[n], x(t)δ[n], δ(t)
x[n] y[n]h[n]
• .
• .
• .
dtteE
txtxtetatxbatx
dttt
bait
b
a NN
NNi
N
NiiN
ij
b
a ji
i
2
ˆ , ˆ , ,over signal afor
,over functions lorthonorma ofset a ,...2,1,0,
Problem 3.66, p.275 of text
...2,1,0 , 0 , 0
j
when minshown becan It
ic
E
b
E
cba
dtttxaE
i
N
i
N
iii
b
a iiN
dtttxA
a
Adttt
b
a ii
ij
b
a ji
1
normalizednot functions basisFor
Problem 3.70, p.281 of text
• 2-dimensional signals
212112
212121
220
110
)(21
21221121
220
2 1
110
220110
1 2
220110
),(11
) ,(1
2,2
,
, all ,,,
dtedtettxTT
dtdteettxTT
a
TT
eattx
ttTtTtxttx
tjn
T T
tjm
tjntjm
T Tmn
tntmj
n mmn