TNE30003 Communication Principles: Sampling

Download as pdf or txt
Download as pdf or txt
You are on page 1of 41

COMMONWEALTH OF AUSTRALIA

Copyright Regulations 1969

WARNING

This material has been reproduced and communicated to you by or on


behalf of Swinburne University of Technology pursuant to Part VB of the
Copyright Act 1968 (the Act).

The material in this communication may be subject to copyright under the


Act. Any further reproduction or communication of this material by you may
be the subject of copyright protection under the Act.
Do not remove this notice.

TNE30003
Communication Principles
Sampling
Updated March 2018
Tony Cricenti

L9
Aims

• Understand the process of Sampling a Signal.


• State the Sampling Theorem.
• Describe three different forms of Sampling
(Impulse, Natural and Flat-top).
• Draw the frequency spectrum for a sampled
signal.
• Describe Aliasing.
• Refer Chap 8.1 of Mesiya

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 2


Why Digital?
• Some advantages of Digital Communication:
• Immune to channel noise and channel distortion.
• Regenerative repeaters
• Flexibility- Digital hardware, DSP, uProcessors, digital
switching, storage …
• Can be coded- error correction codes, high fidelity and
privacy, encryption, compression …
• Several sources of signals, can be transmitted together by
multiplexing- output of digital computer, voice, video,
telemetry, etc, ..
• Using only two symbols (binary). Each symbol is
represented by a pulse (line code).

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 3


Why Digital?
• The Digital Communication Process:
Anti- 100 010 010 ... Line
Alias Quan Enco Code
Filter tizer der pulse coded
Baseband
Analog signal
S/H
signal How to
ADC
represent a
0 and an 1?

V
Pulse Digital
Shaping Modulator
Transmitted
RCF: Modify thro' radio
ASK/ PSK/
pulse shape to channel
FSK
reduce ISI.

The digital communication process.

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 4


Sampling

• Thus far we have considered continuous signal


modulation schemes.
• We now turn to signals which are characterized
by a series of pulses (Sampled Signals).
• These signals are common place in Digital
communications systems, e.g., Compact disc
players and digital cellular phones.

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 5


Sampled Signal: Time Domain

• Sampling: Beginning of Digitalization process:


• A sampled signal consists of a regular sequence of
samples of a continuous time or analogue signal.
• The sampled signal is obtained by periodically
(interval of Ts) observing the signal x(t) during
short intervals of time τ seconds wide.
• The period at which the signal is sampled is called
the Sampling Period Ts.

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 6


Sampled Signal: Time Domain

• The sampling rate or sampling frequency is


related to the sampling period by:

1
• Sampling Rate fs = Samples/s or Hz.
Ts

• That is, if one sample is taken in each Ts


seconds, then the number of samples per
second is given by fs.
• The width of the sampling pulse is τ s.

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 7


Sampled Signal: Time Domain
• Consider the following analogue signal
The sampled signal is as shown:

Sampling Period

Ts
t

Sampling Pulse
Width

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 8


Sampling Process: Natural Sampling
• Natural sampling is produced using the
following system:
Analog
x s (t ) =
A
x (t )
x (t)
signal xs(t)
The signal is sampled B τ
for τ seconds in A
every Ts seconds.
B
Ts
• The pulse train makes the switch change from
point A to point B.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 9
Sampling Process: Natural Sampling

• When the switch is in position B:


A
x (t)
xs(t)
B
A

• The output is: x s (t ) = 0

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 10


Natural Sampling
• The output xs(t) can be expressed as the
product of x(t) and the pulse train p(t), where
p(t) is a pulse train given by:
τ
1

0
Ts

and x s (t ) = x (t ) × p (t )
• We can use this to determine the spectrum of
the sampled signal.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 11
Spectrum of Naturally Sampled Signal

• To obtain the spectrum of the naturally sampled


signal we could determine the Fourier Series
coefficients of xs(t), however this may be difficult
if the signal x(t) is non trivial.
• Recall xs(t)= x(t)×p(t)
• Diagrammatically this is:
Hey, this looks a bit
like DSB-SC!
x(t) xs(t)

p(t)
τ
1

0
Ts
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 12
Spectrum of Naturally Sampled Signal

• If we know the spectrum of x(t) and of p(t) we


could proceed as we did to obtain the spectrum of
DSB-SC.
• The coefficients of the Fourier series for a periodic
pulse train are given by (two-sided):

Aτ sin (nπτ / Ts ) τ
cn = 1
A
Ts nπτ / Ts
0
n = 0,±1,±2 ... Ts

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 13


Spectrum of Pulse Train

• The magnitude spectrum (single-sided) is plotted


by taking the magnitude of cn.
2 Aτ sin (nπτ / Ts ) sin (nπτf s )
Aτ cn = = 2 Aτf s
co = Ts nπτ / Ts nπτf s
Ts
n = 0,1,2, ...
When nfs=k/τ,
Sinc function (k=1,2, …) cn =0.
envelope
Zero Crossing
τ
1
A

0
Ts
One-sided spectrum
f
0 fs 2fs 3fs nfs=1/τ nfs=2/τ nfs=3/τ

Department of Telecommunications, Electrical and Robotics Engineering TNE30003


Spectrum of Pulse Train
• Note that the spectrum consists of impulses at nfs.
i.e., at fs, 2fs, 3fs, 4fs,...
• There is no impulse at f = k/τ (zero crossing),
i.e., 1/τ, 2/τ, 3/τ,…
• To obtain the spectrum of the sampled signal, xs(t),
we can convolve the spectrum of x(t) with that of
p(t).
• But we will take a shortcut by using the modulation
theorem and treating each impulse in the spectrum
of p(t) as a carrier.
• (As per DSB, but with several carriers).
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 15
Spectrum of Naturally Sampled Signal

|X(f)|
x(t) Message signal in
frequency domain
t
Message signal f
in time -W 0 W
domain
P(f) Sampling signal in
Sampling τ frequency domain
signal , p(t) in 1
time
domain 0
Ts
xs(t)
0 fs 2fs 3fs f
Xs(f)
Sampled signal in
t
frequency domain

Sampled signal in
0 W fs 2fs 3fs f
Called Natural sampling since the top of the sample follows
time domain
the shape of the analogue signal.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 16
Spectrum of Naturally Sampled Signal
• The spectrum of a naturally sampled signal is
the baseband message spectrum shifted and
centered about each multiple of the sampling
frequency.
• The strength of each copy is determined by the
strength of the corresponding impulse in the
sampling signal spectrum.

• Spectrum is X(f)*P(f)

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 17


Impulse Sampling

• Impulse sampling: In this case, the signal is


sampled using impulses, that is zero width
pulses (τ=0).
Ts
t

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 18


Spectrum of Impulse Sampled Signal

• We could determine the spectrum using


Fourier techniques, however it can be
obtained by letting τ →0.
c = f , n = 0, 1, 2, ...
• Note as τ →0 then 1/ τ→∞ n s

• This implies that the first zero crossing of the


sinc envelope moves out to f=∞
• This also implies that all spectral components
will have equal amplitude.

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 19


Spectrum of Impulse Sampled Signal

|X(f)|
x(t) Message signal in
frequency domain
t
f
Message signal -W 0 W
in time
domain P(f)
Sampling signal in
frequency domain

xs(t) 0 fs 2fs 3fs f


X (f)
s

t Sampled signal in
frequency domain

0 W fs 2f s 3f s f
Sampled signal in
time domain Called Impulse sampling since the width of the sample is zero.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 20
Flat Top Sampling
• Flat top sampling: In practice, the sampling is
done by a “sample and hold (S/H)” circuit.
• This circuit produces samples that have “flat
tops”.
Sampling
Ts Period
t

Sampling Pulse
Width

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 21


Spectrum of Flat Top Sampled Signal

• It can be shown that the flat top samples are


given by: Impulse samples at t = kT s
Single Rect pulse

xs (t ) = p (t ) * ∑ x(kTs )δ (t − kTs ) = p (t ) * xδ (t )
k
Convolution Impulse samples

p(t) = a single sampling pulse (a rectangular pulse).


xδ(t) = x(t) at the begin of sampling instant.
That is, xs(t) = convolution of a single Rect. Pulse with
impulse sampled samples.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 22
Spectrum of Flat Top Sampled Signal

• Convolution becomes multiplication in Frequency


domain.
• Therefore, the spectrum of the flat-top sampled
signal is given by:
X s ( f ) = P( f ) × X δ ( f )

where, P(f) = spectrum of a single rectangular pulse


and Xδ(f) = spectrum of impulse-sampled samples.

Ref: Mesiya 8.1


Department of Telecommunications, Electrical and Robotics Engineering TNE30003 23
Spectrum of Flat Top Sampled Signal
• That is, to get the spectrum simply multiply the
spectrum of a single pulse (sinc(f)) by the spectrum of
the impulse sampled version of x(t).
Spectrum of a single
|X(f)| sampling pulse
Spectrum of signal to be
|P(f)| (rectangular pulse)
sampled

f 0 f
0 W 1/τ

Spectrum of Flat Top


Spectrum of Impulse-sampled Multiply sampled signal
signal
|Xs(f)|
|Xs(f)|
1/τ
0 f 0 f
W fs-W fs fs+W 2fs-W 2fs 2fs+W W fs-W fs fs+W 2fs-W 2fs 2fs+W

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 24


Spectrum of Flat Top Sampled Signal
• Note the distortion in the spectrum due to the
flat top sampling!
• The amplitude of the spectral components
follows the sinc(f) envelope.
Note distortion

|Xs(f)|
Sinc Envelope

1/τ

0 f

W fs 2fs
fs-W fs+W 2fs-W 2fs+W

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 25


Recovery of Sampled Signals
• We have seen that we can obtain a sampled
version of a signal.
• How do we recover the original signal?
• If we look at the spectrum of the sampled signal,
we can figure out how to extract the original
signal. (Choose impulse sampling for simplicity).
Xs(f)

Spectrum of
Baseband Signal

0 W f s 2f s 3f s f
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 26
Recovery of Sampled Signals: LPF
• We see that the original message can be obtained by
removing all the spectral components centered about
fs, 2fs, 3fs etc.
• Use a reconstruction filter (an LPF) with fcutoff = fs/2.
Xs(f)

This will also work with


Natural Sampling

0 W f s 2f s 3f s f
Use an LPF to
remove components

• Problem occur with Flat top sampling as the


Baseband Spectrum is distorted.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 27
Recovery of Sampled Signals: Holding
• The sampled signal can also be recovered using a
technique called holding.
• Consider a flat top sampled signal:

Sampled signal Spectrum of Flat Top


sampled signal

t
|Xs(f)|

1/τ
0 f
W fs-W fs fs+W 2fs-W 2fs 2fs+W

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 28


Recovery of Sampled Signals: Holding

• If we can “hold” the sampled value, then τ will


increase and 1/ τ will decrease consequently reducing
the amplitude of the higher order components.
• The current sample is held until the next sample.
Thus, τ= Ts.
“Held” sampled signal
Spectrum of a held flat
|Xs(f)| top sample
t
1/τ
0 f
W fs-W fs fs+W 2fs-W 2fs 2fs+W

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 29


Recovery of Sampled Signals: Holding

• Holding increases the energy of the recovered


signal.
• However the recovered signal will be distorted.
Staircase shape.

Distortion
t
|Xs(f)|
1/τ
Distortion 0 f
W fs-W fs fs+W 2fs-W 2fs 2fs+W

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 30


Recovery of Sampled Signals
• What effect does the sampling frequency have on
the recovery of the message signal?
• Consider the spectrum below:
Xs(f)

0 W fs -W fs fs +W f
• As fs is made smaller eventually the
component centered about fs will join with the
spectrum about 0 Hz.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 31
Recovery of Sampled Signals

• The minimum value of fs when the spectral overlap


just occurs is:
Xs(f)

f
0 fs -W fs fs +W
W

• fs-W=W
• Or fsmin=2W (Nyquist Rate)

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 32


Aliasing
• If the sampling frequency is less than the
Nyquist Rate, then we have spectral overlap or
“Aliasing”.
• In this case we cannot recover the message
using a LPF.
• To avoid Aliasing we use a LPF to limit the
message bandwidth to W≤fs/2.

x(t) LPF xs(t)


Anti Alias Sampler
Filter

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 33


Sampling Theorem- Nyquist Criterion

• A signal can be recovered from its sampled


version using a LPF,
if and only if the sampling frequency is greater or
equal to twice the highest frequency of the
baseband signal.

• fs>= 2W Xs(f) Guard


band

0 W fs -W fs fs +W f

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 34


Sampling Theorem- Nyquist Criterion

• fsmin = 2W
• (Called Nyquist sampling rate)

• Nyquist Sampling Theorem: A signal with a


maximum frequency content at W Hz, can be
completely recovered from its samples taken at a
rate of fs>=2W. That is, fsmin = 2W.

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 35


Sampling Theorem- Nyquist Criterion

• Usually, fs is chosen to be greater than fsmin so that


a guard band is available.
• This enables a real LPF (called reconstruction
filter, fcutoff =fs/2) to be used to recover the
baseband signal.
Xs(f) Guard
band

W fs/2 fs -W fs fs +W f
0

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 36


Sampling Theorem- Example

• When a voice signal is sampled for transmission


along telephone lines, its bandwidth is limited to
3.3kHz. If the signal is sampled at 8 kHz, what is
the guard band?
Xs(f) Guard
band

0 W fs -W fs fs +W f

• From data given W = 3.3 kHz and fs=8 kHz


• Guard band fgb=fs-W-W=8-3.3-3.3=1.4kHz.
Department of Telecommunications, Electrical and Robotics Engineering TNE30003 37
Sampling Theorem- Example

• For the sampled signal shown what is:


• The sampling frequency
• Nyquist Rate
• Sampling Type
• Spectrum
1µs
t

10 µ s

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 38


Sampling Theorem- Example

• The sampling frequency is determined


from the sampling period:

Ts = 1µs 1 1
∴ fs = = −6
= 1 MHz
Ts 1× 10

• Since have a sinusoid W=1/10us=100 kHz


• Nyquist Rate=2W Hz
• Thus, Nyquist Rate=200 kHz

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 39


Sampling Theorem- Example

• The sampling type is impulse since the width of


the sampling pulse is zero.
• The one sided spectrum is as shown:
|XS(f)| What happens if
|X(f)|
fs = 150 kHz??

f f
kHz 100 900 1100 1900 2100 kHz
100 500
Baseband Sampled Message
Message Signal Signal

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 40


References

• Mesiya Chapter 8 sections 8.1 and 8.2


• Carlson Chapter 6 sections 6.1 and 6.2
• You Tube….
• https://www.youtube.com/watch?v=WUCMavXbJo4

Department of Telecommunications, Electrical and Robotics Engineering TNE30003 41

You might also like