Digital Communications

Analog to Digital
Part I Conversion (ADC)

Part II Digital Transmission Sampling

&
Signal Reconstruction
with Applications
Dr. Ali Hussein Muqaibel

ver. 4.5
Class Objectives
• Why Digital?
• Analog to Digital Conversion (ADC)
• Sampling
• Sampling Theorem.
• What is the spectrum of sampled
signals?
• Signal Reconstruction.
• The interpolation function (sinc)
• Zero hold and first order hold
approximation
• Practical Sampling

Dr. Ali Hussein Muqaibel 2

Advantages of Digital Communication over
Analog Communication
➢Immunity to Noise (possibility of regenerating the original digital signal if signal power
to noise power ratio (SNR) is relatively high by using of devices called repeaters along
the path of transmission).
➢Efficient use of communication bandwidth (through use of techniques like
compression).
➢Digital communication provides higher security (data encryption).
➢Error Control Coding : the ability to detect errors and correct them if necessary.
➢Design and manufacturing of electronics for digital communication systems is much
easier and much cheaper than the design and manufacturing of electronics for analog
communication systems. What is the price for going digital ?
Dr. Ali Hussein Muqaibel 3
 Continuous-time vs. Discrete time signal.

Analog to Digital Conversion Digital vs. Analog

PCM: Pulse Coded Modulation

Source of
Low-pass 01101110101
continuous
Sampler Quantizer Encoder PCM signal applied
time message filter to channel input
signal

Distorted PCM signal Regenerated PCM

Regenerative Regenerative
produced at channel …… signal applied to the
repeater repeater
output receiver

Final channel Regeneration Reconstruction

Decoder Destination
output circuit filter

Dr. Ali Hussein Muqaibel 4

 g(t)
𝑇𝑠 = 1/𝑓𝑠 sampling interval

Sampling Theorem s s s s s s s  s  s s s s
t

A signal whose spectrum is band limited to 𝐵 𝐻𝑧

[𝐺(𝑓) = 0 𝑓𝑜𝑟 |𝑓| > 𝐵] can be reconstructed exactly from its samples taken
1
uniformly at a rate 𝑅 > 2𝐵 𝐻𝑧 (samples/ sec). i.e 𝑇𝑆 <
2𝐵

Minimum sampling frequency is 𝑓𝑠 = 2𝐵 𝐻𝑧 𝑁𝑦𝑞𝑢𝑖𝑠𝑡 𝑅𝑎𝑡𝑒

1
𝑇𝑠 = Nyquist interval
2𝐵

The 𝑛𝑡ℎ impulse located at 𝑡 = 𝑛𝑇𝑠 has a strength 𝑔 𝑛𝑇𝑠 the value of 𝑔(𝑡) at 𝑡 = 𝑛𝑇𝑠

Dr. Ali Hussein Muqaibel 5

Math Representation of Sampling
• 𝑔(𝑡) is a continuous–time signal with bandwidth 𝐵 𝐻𝑧 or (2𝐵 𝑟𝑎𝑑/𝑠) .
• Sampling is equivalent to multiplying 𝑔(𝑡) by a train consisting of unit impulses repeating periodically
every 𝑇𝑠 second. A train of delta function 𝑇 (𝑡) that occur every 𝑇𝑠 is given by
𝑠 𝛿𝑇 (𝑡)
+∞

𝑇 𝑡 = ෍ 𝛿(𝑡 − 𝑛𝑇𝑠 )
𝑠
𝑛=−∞

• The sampled signal 𝑔(𝑡)

ҧ
+∞

𝑔ҧ 𝑡 = 𝑔(𝑡) ෍ 𝛿(𝑡 − 𝑛𝑇𝑠 )

𝑛=−∞
g(t)
+∞

= ෍ 𝑔(𝑡)𝛿(𝑡 − 𝑛𝑇𝑠 )
𝑛=−∞ s s s s s s s  s  s s s s
t

+∞

= ෍ 𝑔(𝒏𝑻𝒔 )𝛿(𝑡 − 𝑛𝑇𝑠 )

𝑛=−∞
Dr. Ali Hussein Muqaibel 6
 Using Fourier series representation:
Spectrum of Sampled Signal 𝜹𝑻𝒔
+∞
𝟏
𝒕 = ෍ 𝛿(𝑡 − 𝑛𝑇𝑠 ) = ෍ 𝑒 𝑗𝑛2𝜋𝑓𝑠 𝑡
𝑻𝒔
+∞

𝑛=−∞ −∞
g(t) G()

t 
s s s s s s s  s  s s s s
s s s 2B +2B s s s

+∞
+∞ +∞ 𝟏
𝟏
𝑔ҧ 𝑡 = ෍ 𝑔 𝒏𝑻𝒔 𝛿 𝑡 − 𝑛𝑇𝑠 = ෍ 𝑔 𝑡 𝑒 𝑗𝑛2𝜋𝑓𝑠 𝑡 𝐺ҧ 𝑓 = ෍ 𝐺(𝑓 − 𝑛𝑓𝑠 )
𝑻𝒔 𝑻𝒔
𝑛=−∞ −∞ −∞
G()
g(t)

A/Ts

... ...
t
s s s s s s s  s  s s s s 
s s s–2B s s+2B 2B +2B s–2B s s+2B s s

Dr. Ali Hussein Muqaibel 7

 Signal Reconstruction
Magnitude of LPF should be Ts to cancel G() LPF for reconstructing the origianl
the scaling factor caused by sampling Ts signal from the sampled signal
A/Ts

... ...

s s s–2B s s+2B 2B +2B s–2B s s+2B  s  s

Reconstructed Signal s > 2(2B)  No interference between Images

A/Ts


s s s 2B +2B s  s  s

Dr. Ali Hussein Muqaibel 8

Signal Reconstruction (Interpolation)
• Interpolation: the process of reconstructing continuous-time signal from its samples.
• Use of lowpass filter of BW of 𝐵 𝐻𝑧 (ideal)
𝑓
𝐻 𝑓 = 𝑇𝑆 Π
2𝐵
• In time domain (IFT)
ℎ 𝑡 = 𝟐𝑩𝑻𝒔 𝑠𝑖𝑛𝑐 2𝜋𝐵𝑡
• At Nyquist rate 𝟐𝑩𝑻𝒔 = 𝟏
ℎ 𝑡 = 𝑠𝑖𝑛𝑐 2𝜋𝐵𝑡
𝑛
Observe ℎ 𝑡 = 0 at all Nyquist sampling interval 𝑡 = ± 𝑒𝑥𝑐𝑒𝑝𝑡 𝑎𝑡 𝑡 = 0
2𝐵

Dr. Ali Hussein Muqaibel 9

 Reconstruction
Signal Reconstruction system
ℎ 𝑡 = 𝑠𝑖𝑛𝑐 2𝜋𝐵𝑡
• The output of the reconstruction system (discrete convolution)

𝑔 𝑡 = ෍ 𝑔 𝑘𝑇𝑠 ℎ 𝑡 − 𝑘𝑇𝑠
𝑘
𝑔(𝑡) = ෍ 𝑔 𝑘𝑇𝑠 𝑠𝑖𝑛𝑐 2𝜋𝐵 𝑡 − 𝑘𝑇𝑠
𝑘
1
• At Nyquist rate 𝑇𝑠 =
2𝐵

𝑔(𝑡) = ෍ 𝑔 𝑘𝑇𝑠 𝑠𝑖𝑛𝑐 2𝜋𝐵𝑡 − 𝑘𝜋

𝑘
• Interpolation formula yields the value of 𝑔(𝑡) between samples as a
weighted sum of all sample values.

Dr. Ali Hussein Muqaibel 10

 Example
Find a signal 𝑔(𝑡) that is band-limited to 𝐵 𝐻𝑧 & whose samples are
𝒈 𝟎 = 𝟏 𝑎𝑛𝑑 𝑔 ±𝑇𝑠 = 𝑔 ±2𝑇𝑠 = 𝑔 ±3𝑇𝑠 = ⋯ = 0
where the sampling interval 𝑇𝑠 is the Nyquist interval for 𝑔 𝑡 , that is 𝑇𝑠 = 1/2𝐵.

𝑔(𝑡) = ෍ 𝑔 𝑘𝑇𝑠 𝑠𝑖𝑛𝑐 2𝜋𝐵 𝑡 − 𝑘𝑇𝑠

𝑘
𝑔(𝑡)
𝑔(𝑡) = 𝑠𝑖𝑛𝑐 2𝜋𝐵𝑡

Dr. Ali Hussein Muqaibel 11

 Order of Signal Reconstruction (Reconstruction Filters)
g(t)

𝑦 𝑡 = ℎ 𝑡 ∗ 𝑔ҧ 𝑡 = ℎ 𝑡 ∗ [𝑔 𝑡 𝛿𝑇𝑠 𝑡 ]

Ts
h0(t) g(t)
hOO(t)
1
1

t t
–2Ts –Ts Ts 2Ts
Ts

Zero–Order Hold Ts
Sinc Filter (Infinite–Order Hold)
g(t)
g(t)
h1(t)

t
–Ts Ts
Ts
Ts

First–Order Hold

Dr. Ali Hussein Muqaibel 12

Interpolation and hold Circuits in Frequency
Zero–Order Hold Sinc Filter (Infinite–Order Hold)
Interpolation function

Dr. Ali Hussein Muqaibel 13

 Practical Sampling Pulses
• Finite width practical pulses
• Because it is periodic, using Fourier
series representation
+∞

𝑃𝑇𝑠 = ෍ 𝐶𝑛 𝑒 𝑗2𝜋𝑛𝑓𝑠
−∞
• Not
+∞
𝟏
𝜹𝑻𝒔 𝒕 = ෍ 𝑒 𝑗𝑛2𝜋𝑓𝑠 𝑡
𝑻𝒔
−∞

Dr. Ali Hussein Muqaibel 14

Practical Difficulty in Signal Reconstruction
• To avoid the need for ideal filter 𝑓𝑠 > 2𝐵, we may use a filter with gradual cutoff
characteristics.
• Also we want the filter to be zero outside...(Impossible by Paley-Wiener criterion)
but closely approximated.
Magnitude of LPF should be Ts to cancel G() LPF for reconstructing the origianl
the scaling factor caused by sampling Ts signal from the sampled signal
A/Ts

... ...

s s s–2B s s+2B 2B +2B s–2B s s+2B s s

Reconstructed Signal s > 2(2B)  No interference between Images

A/Ts


s s s 2B +2B s s s

Dr. Ali Hussein Muqaibel 15