The Handbook of Formulas and Tables For Signal Processing

Poularikas A. D.
Band-Limited Functions and Sampling Theorem

The Handbook of Formulas and Tables for Signal Processing.
Ed. Alexander D. Poularikas
Boca Raton: CRC Press LLC,1999
1999 by CRC Press LLC
38
Bandlimited
Functions-Sampling
38.1 Bandlimited Functions
38.2 Sampling and Interpolation
38.3 Sampling with Pulses.
38.4 Frequency Sampling
38.5 n-Variables Sampling
38.6 Sampling and Derivatives
38.7 Papoulis Generalization
38.8 Bounds and Extreme Values of BL Function
References
38.1 Bandlimited Functions

38.1.1 Definition (Bandlimited)
F( ) = 0 for > BL (sigma bandlimited), E( energy) <
38.1.2 Definition (Time-limited)

f (t ) = 0 for t > , E (energy) < TL ( time - limited).
38.1.3 Energy
E=
f (t ) dt =
2
1
2
F( ) d
2
38.1.4 Properties of Bandlimited Functions
1.
f (t ) e jt dt = F( ) = Tp ( )
f (nT )e
jnT
n =
of 2 and centered at = 0
2.
f (t ) dt =
2
1
2
F( ) d = T
2
n =
f (nT )
, T = sampling time, p ( ) = pulse of width
1
3. f (t ) =
2
F ( ) e
jt
4.
F( ) d < ( F( ) is absolutely integrable)
5.
f (t )
6. f (t ) =
E t
e , E = energy (see 38.1.3), f (t ) BL (sigma bandlimited)
1
2
jF( ) e jt d, a bandlimited function is analytic in the entire t plane
7. A bandlimited function has finite energy, is analytic in the entire t plane, and is of exponential
type
( f (t) < Ae
, A and are constants

M
8. Periodic BL: y(t ) =
ae
jn 0t
, E = finite in (0, T ), T = period,
n = M
an = Fourier series coefficients

9. A function f(t) cannot be bandlimited and time-limited
38.1.5 Properties of Time-Limited Functions

1. Derivatives:
f ( k ) ( ) = lim f ( k ) ( + ), 0; f ( k ) ( ) = lim f ( k ) ( ), 0, > 0;
f ( k ) () k th derivative
2. If f(t) is TL and has bounded derivatives of order up to n for every t < , then
F ( ) =
1
2
+
f (t ) e jt dt =
1
[ f ( )e j f ( )e j ] + L
j
1
[ f ( n1) ( ) e j f ( n1) ( ) e j ] + O (1 / n+1 )
( j )n
3. If f(t) is TL and has bounded derivatives of order up to n for every t < and
f ( k ) ( ) = 0, f ( k ) ( ) = 0, k = 0,1,L, n 1, then F( ) = O (1 / n+1 ), .
38.2 Sampling and Interpolation

38.2.1 Interpolation Function
fk (t ) = sampled function with k (t ) =
T f (nT )k(t nT ), Tk(0) = 1, k(nT ) = 0 for n 0 (see Figure
n =
38.1), T = sampling time

fk(t)
f(t)
1
T
fs(t)
k(t)
-T
-T
2T
k(t)
fk(t)
T
FIGURE 38.1
38.2.2 Sampled Function
f s (t ) =
Tf (nT ) (t nT ) = f (t) Tcomb (t) =

T
sampled function with delta functions, combT (t ) =
n =
(t nT )
n =
38.2.3 Convolution Property

f k (t ) = f s (t ) k (t )
38.2.4 Spectrum of Sampled Function
Fs ( ) =
Tf (nT )e
n =
jnT
F( + 2n), = T ;
n =
Fk ( ) = K ( ) Fs ( ) = K ( )
F( + 2n) = K ()[COMB
( ) F( )] (see 38.2.3 and FT proper-
n =
ties), (see Figure 38.2)
38.2.5 Sampling Theorem
f (t ) =
Tf (nT ) sin 2
n =
> 2,
(t nT ) [(t nT ) for s =
T
f (t ) = BL function =
f (nT )
n =
sin (t nT )
, =
and s = 2, F( ) = 0 for >
(t nT )
T
38.2.6 Sampling Frequency (Nyquist Rate)

1
=
T
38.2.7 Truncation Error

N
f N (t ) =
n = N
sin (t nT )
f (nT )
, eN =
(t nT )
f (t ) f N (t ) dt = T
n >N
f (nT ) ,
T (t - nT) = TcombT(t)
f(t)
fs(t) = f(t) TcombT(t)
n=-
( - ns)
n=-
-s
T
F
2 COMB () = 2
F()
FIGURE 38.2
2 =
s
T
fs() = f() * COMB ()

s
s
2
eN
for every t, f (t ) BL
f (t ) f N (t )
38.2.8 Bernstein Inequality

f (t ) M, f (t ) BL and f (t ) M
38.2.9 Bandlimited Interpolation
f (t ) arbitrary function, fi (t ) =
f (nT )
n =
sin (t nT )
(see also 38.2.1 with k (t ) = sin t / t ),
(t nT )
Fi ( ) = Fs ( ) p ( ), p ( ) = centered pulse with total width 2, fi (t ) = BL obtained by passing fs(t)

(see 38.2.2) through an ideal lowpass filter, fi (nT ) = f (nT ), f (t ) fi (t )
F( ) d
>
38.2.10 Bandlimited Mean-Square Approximation

f (t ) f (t )
f(t) = arbitrary function, f (t ) =
1
2
1
2
F( ) d ,
>
F ( ) e
jt
d, F ( ) = F{ f (t )} = F( ) p ( ), p ( ) = centered
pulse with width 2.
38.3 Sampling with Pulses

38.3.1 Train of Rectangular Pulses
Fs ( ) = F{ f (t ) f p (t )} =
n =
n
sin s
2
F( n s ),
n s
f p (t ) = p / 2 (t )
(n nT ), p
/2
(t ) = centered pulse with width , s = 2 / T , F{ f (t )} = F( )
n =
38.3.2 Train of Pulses with Flat Tops
sin
2
2
Fs ( ) =
T
n =
F( n s ), fs (t ) = [ f (t )combT (t )] p / 2 (t ), combT (t ) =
(n nT )
n =
38.4 Frequency Sampling

38.4.1 Frequency Sampling
If f(t) = time limited = 0 for t > TN ,
F ( ) =
sin(TN n)
T n
N
N
F n T
n =
38.5 n-Variables Sampling
f (t1, t2 ,L, tn ) =
m1 =
mn =
m
m sin(1t1 m1) sin( ntn mn )
f 1 ,L, n
L
,
n
1t1 m1
ntn mn
1
f (t1 ,L, t n ) = n variables function, g( y1 , y2 ,L, yn ) = 0 for yk > k , k = 1, 2,L, n, g() = Fn { f ()}
38.6 Sampling and Derivatives

F{ f (t )} = F( ) = 0 for > 2 f N ,

sin h (t kh)
(t kh) ( R)
(1)
(kh)
(kh) + (t kh) (kh) + L +
R!
(t kh)
h
f (t ) =
R +1
k =
R = highest derivative order, h = ( R + 1) /(2 f N ), ( R) (kh) = R th derivative of the function (),

j
( j ) (kh) =
i h
j 1
R( +j 11) f (i ) (kh), a( ) =
i=0
d
[(t / sin t ) ] t =0 , ( 0 ) = 1,
dt
(5 + 2)
(35 2 + 42 + 16)
( 2 ) = , ( 4 ) =
, ( 6 ) =
, ( ) = 0 for odd .
3
63
15
38.7 Papoulis Generalization

38.7.1 One System
1
f (t ) = BL signal =
g(nT ) y(t nT ), y(t ) =
2
n =
1/ T = / = Nyquist rate
e jt d, g(t ) = 1
H ( )
2
F ( ) H ( ) e
jt
d,
Example
g(t ) =
f (t ) e
d, H ( ) = F{e
1
}=
, y( t ) =
+ j
2
1 + e
jt
d = ( t cos t +
(t 1)sin t ) / t ,
2
38.7.2 Multiple System
1
f (t ) =
[g1 (nT ) y1 (t nT ) + L + gm (nT ) ym (t nT )], where yk (t ) =
C
n =
1, 2,L, m; gk (t ) =
+c
Y ( , t ) e
jt
d f o r k =
F()H ()e
1
2
jt
d, 1 / T = / m, c = 2 / m = 2 / T , Yi (,t)s are determined
from the system.

H1 ( )Y1 (, t ) + L + H m ( )Ym (, t ) = 1
H1 ( + c)Y1 (, t ) + L + H m ( + c)Ym (, t ) = e jct
L
H1 ( + ( m 1)c)Y1 (, t ) + L + H m ( + ( m 1)c)Ym (, t ) = e j ( m 1)ct , t = arbitrary
is in the interval ( , + c)
Properties
1. Yk (, t ) periodic in t with period T = 2 / c
2. Yk (, t ) e jt =
y (t nT ) e
jnT
for every in ( , + c) where Yk (, t ) is defined and for
n =
every t.
3.
For every in ( , ),
jt
= H1 ( )
y (t nT ) e
jnT
+ L + H m ( )
n =
y (t nT ) e
jnT
n =
38.8 Bounds and Extreme Values of BL Function

38.8.1 Bounds of Output Function
E
g(t )
2
H ( ) d
1/ 2
jt
equality holds for t = t0 only if F( ) = kH ( )e 0 for < and k is a constant determined from
energy requirements,
1
g(t ) =
2
F ( ) H ( ) e
jt
1
d, f (t ) = BL function, E = energy =
2
F( ) d
2
Properties
1. If H ( ) = 1 ( g(t ) = f (t )), f (t )
2 If H ( ) = j ( g(t ) = f (t )),
f (t )
E
3
38.8.2 Maximum Energy Concentration

y =
1
Ey
y(t ) dt maximum if y(t) is equal to the yeig (t ) = eigen function of the integral equation
2
y( x )
sin (t x )
dx = y(t )
(t x )
(1)
corresponding to the maximum eigenvalue max , y(t ) = BL, E y = energy of y(t ), solutions of (1)
are known as prolate spheroidal wave functions.
References
Jerri, A. J., The Shannon Sampling Theorem its various extensions and applications: a tutorial review,
Proc. IEEE, 65, 1565-1596.
Linden, D. A., A discussion of sampling theorems, Proc. IRE, 47, 1219-1226, 1959.
Papoulis, A., Signal Analysis, McGraw Hill Inc., New York, NY, 1977.

The Handbook of Formulas and Tables For Signal Processing

Uploaded by

Copyright:

Available Formats

The Handbook of Formulas and Tables For Signal Processing

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Handbook of Formulas and Tables For Signal Processing

Uploaded by

Copyright:

Available Formats

Poularikas A. D.

Band-Limited Functions and Sampling Theorem

1999 by CRC Press LLC

38.1 Bandlimited Functions

38.1.2 Definition (Time-limited)

38.1.4 Properties of Bandlimited Functions

1999 by CRC Press LLC

, T = sampling time, p ( ) = pulse of width

F( ) d < ( F( ) is absolutely integrable)

jF( ) e jt d, a bandlimited function is analytic in the entire t plane

, A and are constants

8. Periodic BL: y(t ) =

, E = finite in (0, T ), T = period,

an = Fourier series coefficients

38.1.5 Properties of Time-Limited Functions

38.2 Sampling and Interpolation

fk (t ) = sampled function with k (t ) =

T f (nT )k(t nT ), Tk(0) = 1, k(nT ) = 0 for n 0 (see Figure

38.1), T = sampling time

38.2.2 Sampled Function

Tf (nT ) (t nT ) = f (t) Tcomb (t) =

sampled function with delta functions, combT (t ) =

38.2.3 Convolution Property

38.2.4 Spectrum of Sampled Function

( ) F( )] (see 38.2.3 and FT proper-

ties), (see Figure 38.2)

38.2.5 Sampling Theorem

38.2.6 Sampling Frequency (Nyquist Rate)

38.2.7 Truncation Error

1999 by CRC Press LLC

fs(t) = f(t) TcombT(t)

1999 by CRC Press LLC

fs() = f() * COMB ()

38.2.8 Bernstein Inequality

38.2.9 Bandlimited Interpolation

Fi ( ) = Fs ( ) p ( ), p ( ) = centered pulse with total width 2, fi (t ) = BL obtained by passing fs(t)

38.2.10 Bandlimited Mean-Square Approximation

f(t) = arbitrary function, f (t ) =

pulse with width 2.

38.3 Sampling with Pulses

(t ) = centered pulse with width , s = 2 / T , F{ f (t )} = F( )

38.3.2 Train of Pulses with Flat Tops

1999 by CRC Press LLC

38.4 Frequency Sampling

38.5 n-Variables Sampling

38.6 Sampling and Derivatives

R = highest derivative order, h = ( R + 1) /(2 f N ), ( R) (kh) = R th derivative of the function (),

38.7 Papoulis Generalization

1999 by CRC Press LLC

38.7.2 Multiple System

d, 1 / T = / m, c = 2 / m = 2 / T , Yi (,t)s are determined

from the system.

for every in ( , + c) where Yk (, t ) is defined and for

38.8 Bounds and Extreme Values of BL Function

1999 by CRC Press LLC

38.8.2 Maximum Energy Concentration

1999 by CRC Press LLC

You might also like