sampling theorem :

• A continuous-time signal can be represented by its samples at points equally spaced in time
• e.g, in moving pictures, which consist of a sequence of individual frames, each one represents
an instantaneous view (i.e., a sample in time) of a continuously changing scene. When these
samples are viewed in sequence at a sufficiently fast rate, we perceive an accurate
representation of the original continuously moving scene
• The importance of the sampling theorem lies in its role as a bridge between continuous-time
signals and discrete-time signals.
• In general, processing discrete-time signals is more flexible and is often preferable than
processing continuous-time signals.
In general, a signal could be uniquely specified by a sequence of equally spaced samples. For
example, in Figure 7.1 we illustrate three different continuous-time signals, all of which have
identical values at integer multiples of T; that is,

If a signal is band limited and the samples

T
Sampling rate
 p(t)
Conversion from
xp(t) impulse train to x(n)= xc(nT)
xc(t) 
discrete-time
sequence

xp(t) x(n)
xc(t)

t t 3 2 1 0 1 2 3 4
n
0 3T 2T T 0 T 2T 3T 4T
 xc(t) xc(t)

3T 2T T 0 T 2T 3T 4T
t 8T 4T 2T 0 2T 4T 8T 10T
t

x(n) x(n)

3 2 1 0 1 2 3 4
n 6 4 2 0 2 4 6 8
n
 p(t)
Conversion from
xp(t) impulse train to x(n)= xc(nT)
xc(t) 
discrete-time
sequence

𝑛=−∞
xp(t) = xc(t) p(t)
n 
𝑝 𝑡 = ෍ δ(𝑡 − 𝑛𝑇ሻ  xc (t )  (t  nT )
𝑛=−∞ n  
n 
  x (nT )(t  nT )
n  
c
 p(t)
Conversion from
xp(t) impulse train to x(n)= xc(nT)
xc(t) 
discrete-time
sequence

1
𝑋𝑝 𝑗𝜔 = 𝑋𝑐 (𝑗𝜔ሻ ∗ 𝑃(𝑗𝜔ሻ
2𝜋
∞
2𝜋 2𝜋
𝑃 𝑗𝜔 = ෍ 𝛿 𝜔 − 𝑘𝜔𝑠 , 𝜔𝑠 =
𝑇 𝑇
𝑘=−∞
• That is, 𝑋𝑝 𝑗𝜔 is a periodic function of 𝜔 consisting of a
superposition of shifted replicas of 𝑋𝑐 𝑗𝜔 , scaled by 1/T

∞
1 2𝜋
𝑋𝑝 𝑗𝜔 = ෍ 𝑋𝑐 𝜔 − 𝑘𝜔𝑠 , 𝜔𝑠 =
𝑇 𝑇
𝑘=−∞
Figure 7.3 Effect in the frequency domain of sampling in the time domain: Figure 7.3 Continued
(a) spectrum of original signal; (c) spectrum of sampled signal with 𝜔𝑠 > 2 𝜔𝑀 .
(b) spectrum of sampling function; (d) spectrum of sampled signal with 𝜔𝑠 < 2 𝜔𝑀 .
A bandlimited continuous-time signal can be sampled and perfectly
reconstructed from its samples if the waveform is sampled over twice
as fast as it's highest frequency component.

The sampling frequency is known as the Nyquist Rate

ωM ωM ω

Band-Unlimited Yc(jω)

ω
 Xc(jω)
Band-Limited 1

ωM ωM ω
Sampling with
Higher Frequency
2/T
P(jω)
ω
3ωs 2ωs ωs ωs 2ωs 3ωs
Sampling with
Lower Frequency 2/T P(jω)

6ωs 4ωs 2ωs 2ωs 4ωs 6ωs

ωM ωM ω
Sampling with
Higher Frequency
ωs > 2ωM
2/T
P(jω)
3ωs 2ωs ωs ωs 2ωs 3ωs
ω
Sampling with
Lower Frequency 2/T P(jω) ωs < 2ωM
ω
6ωs 4ωs 2ωs 2ωs 4ωs 6ωs
 ωs > 2ωM
Xc(jω)
1

ωM ωM ω

2/T
P(jω)
3ωs 2ωs ωs ωs 2ωs 3ωs
ω
Xs(jω)
1/T
ω
3ωs 2ωs ωs ωs 2ωs 3ωs
 ωs > 2ωM Xs(jω) is a periodic function
Xc(jω) with period ωs
Passing Xs(jω) through a low-pass 1
filter with cutoff frequency
ωM < ωc< ωs ωM ωM ωM ω
the original signal can be recovered.
2/T
P(jω)
3ωs 2ωs ωs ωs 2ωs 3ωs
ω
Xs(jω)
1/T
ω
3ωs 2ωs ωs ωs 2ωs 3ωs
ωs < 2ωM
Xc(jω)
1

ωM ωM ω

2/T P(jω)

2ωs
ω
6ωs 4ωs 2ωs 4ωs 6ωs

1/T Xs(jω)
ω
6ωs 4ωs 2ωs 2ωs 4ωs 6ωs
 ωs < 2ωM Xs(jω) is a periodic function
Xc(jω) with period ωs
No way to recover the signal 1
Aliasing
ωM ωM ω

2/T P(jω)

2ωs
ω
6ωs 4ωs 2ωs 4ωs 6ωs

1/T Xs(jω)
ω
6ωs 4ωs 2ωs 2ωs 4ωs 6ωs