Introduction to Communication Networks

Unit 2
Fundamentals of Signal
Transmission

EECS 122 SPRING 2007

Spring 2007

Acknowledgements

For this unit I have used mostly slides from theBook by WIliam
Stallings (our supplementary texbook)

Some figures/tables have been taken from Communication

Networks by Leon_Garcia and Widjaja marked GW

Other sources are referenced...

Prof. Adam Wolisz

2 of 46

Information Transmission

Important transformations:

Source coding

Channel coding

Input g:
information

Input representation:

Transmitted signal
digital or analog

g(t)
Input
device

Received signal generally

different from transmitted

s(t)
Transmitter

Sending

gr(t) ideally (but not

necessarily) identical to g(t)

r(t)
Transmission
medium

gr(t)
Receiver

g
Output
device

Receiving

Source information: Files, Pictures, Speech, Video....

Input signal: a representation of the data (a time function)
Transmitted signal: another representation of the data, proper for transmission

Prof. Adam Wolisz

3 of 46

Input information ...see examples for speech...

Prof. Adam Wolisz

[GW]

4 of 46

Continuous & Discrete

Prof. Adam Wolisz

5 of 46

Sinusoidal Signals

s(t) = A sin(2 f t + )

A: Amplitude (max strength of signal)

f: frequency (rate of change of signal [Hz] repetitions/s)

: Phase (relative position in time)

Two signals with

90o phase shift:

Prof. Adam Wolisz

6 of 46

Transmission of Sinusoidal Signals

Sender

Sender signal
s(t) = As sin(2 f t )

The transmitter has a given power Ps [mW]

Power to Amplitude relation is Ps As2

Receiver

Received signal:
r(t) = Ar sin(2 f t - ), = Distance/ (Velocity of propagation)

In most cases: Ar < As

The receiver receives a power Pr [mW] < Ps [mW]

During

the transmission signals are always attenuated. The attenuation

is strongly frequency dependent.

The

delay may also be frequency dependent...

Prof. Adam Wolisz

7 of 46

Attenuation of Signals

The attenuation for the transmission is

af [dB] = 10 log ( Ps /Pr ) = 20 log (As/Ar)

Decibel (dB)

is a ratio (not a value)

expresses a logarithmic relationship

is used to indicate relative magnitude (gain or loss)

dB = 10 log(P1/P2)
Power

ratio n of: (n=2) => 3 dB,

(n=10) => 10 dB,

(n=100) =>20 dB
(n=0.1) => -10 dB

For expressing power, frequently dBW is used, expressing the

signal strength with reference to 1 W

Power [dBW] = 10 log ( Power [W] /1 W )

Prof. Adam Wolisz

8 of 46

Power Equations

The transmitter has a given power Ps [dBW]

The medium has an attenuation of af [dB/km] for a unitary

distance

The receiver has a sensitivity Sf [dBW]

The smallest incoming power which it will detect with acceptable error see discussion later...

To receive a signal properly the power equation (in dB) must

be considered
Ps - afx - Sf > 0
where x: distance [km]

What can be influenced?

Prof. Adam Wolisz

Ps- sometimes...

X - usually...
9 of 46

Fourier Analysis of Signals

Every signal s(t) can be represented as a (possibly infinite!)

sum of component sinusoidal signals with different
frequencies
Example in the next slide:

We do ususally ignore sinusoids with rather small amplitude

(thus carrying little energy) consider only SIGNIFICANT
sinusoids.

Signals with QUICK changes of amplitude will contain

significant sinusoids with HIGHER frequencies than signals
with slow changes....

Prof. Adam Wolisz

10 of 46

Square Wave Frequency Components

Prof. Adam Wolisz

11 of 46

Spectrum and Bandwidth

There is a clear and unique equivalence between time representation and

frequency spectrum representation
s(t) S(f)

rapid signal changes bigger bandwidth has to be used: smaller

rectangles => higher basic frequency...

Spectrum

Absolute bandwidth

range of frequencies contained in signal

width of spectrum

Effective bandwidth

Often just bandwidth

Band of frequencies containing most of the energy

DC Component

Component of zero frequency

Prof. Adam Wolisz

12 of 46

Example: Spectrum with DC Component

Prof. Adam Wolisz

13 of 46

Acoustic Spectrum (Analog)

Prof. Adam Wolisz

14 of 46

Low-pass Filter
- Any transmission system supports a limited band of
frequencies
- A frequent system:
- low-pass filter

af = a for f < flim ,

af for f flim
flim

= upper limiting frequency

Gf = 1/af = transmission gain

Be careful:
We will mix gain and attenuation
since both are popular!
Prof. Adam Wolisz

15 of 46

Characteristics of a Typical Voice Channel

Attenuation:

Attenuation - relative to
attenuation at 1000 Hz!

1 - unequalized
2 - equalized

Frequency [Hz]
Note: Attenuation relative to attenuation at 1000 Hz!
Prof. Adam Wolisz

16 of 46

Transmission of Digital Signals

Transmitting a rectangular pulse over a low-path filter with different
bandwidth (the R-C System).

The Response of an R-C Transmission System to a Rectangular

Input Pulse for Several Values of =1/(RC)
Prof. Adam Wolisz

17 of 46

Even cloaser to reality...

Prof. Adam Wolisz

[from Comer]

18 of 46

Transmission of Complex Signals

Analog
signals
subject to
distortion
by noise

Digital
signals
reconstru
cted in
spite of
noise

Regenera
tion
possible
for digital
signals
Prof. Adam Wolisz

19 of 46

Evailable bandwidth influences signal quality!!!

Prof. Adam Wolisz

[GW]

20 of 46

Digital Representation of Analog Information (1)

Source coding
Sampling (discretization
in time)
- The Nyquist sampling
theorem says that any
signal s(t), where the
spectrum has no
frequency components
at a frequency higher
than B[Hz] can be
precisely reconstructed
from sample values
taken 2B times per
second !
For voice we assume
that : B=4000 Hz,
corresponding to 8000
samples/second.

Prof. Adam Wolisz

21 of 46

Digital Representation of Analog Signals(2)

Quantization (discretization in value)

A proper quantization unit has to be chosen. For voice (not CD quality !!!) 256
levels give an excellent quality (8 bits coding, usually sign plus 7 bits value).

This is PCM (Pulse Code Modulation) coding

The discretization can be linear (left side) or nonlinear

Effect of nonlinear coding

Prof. Adam Wolisz

22 of 46

Transmission Media and their Features

We will shortly discuss some of them

Guided Media:
Twisted

Pair

Coaxial

cable

Fiber

Unguided Media:
Radio

Prof. Adam Wolisz

23 of 46

Twisted Pair (TP)

2 parallel wires -> simple antenna

A way to limit electromagnetic influence twisted pair

Prof. Adam Wolisz

24 of 46

Twisted Pair (TP) (2)

UTP- Unshielded TP (speed limited by radiation)

considerations

STP- Shielded TP (speed limited by higher capacity)

Prof. Adam Wolisz

25 of 46

Comparison of Shielded and Unshielded TP

Attenuation (dB per 100 m)

Frequency
(MHz)

Near-end Crosstalk (dB)

Category 3
UTP

Category 5
UTP

150-ohm STP

Category 3
UTP

Category 5
UTP

150-ohm STP

2.6

2.0

1.1

41

62

58

5.6

4.1

2.2

32

53

58

16

13.1

8.2

4.4

23

44

50.4

25

10.4

6.2

41

47.5

100

22.0

12.3

32

38.5

300

21.4

31.3

Prof. Adam Wolisz

26 of 46

Attenuation of typical twisted pair cables

Prof. Adam Wolisz

[GW]

27 of 46

Coaxial Cable

[GW]

Used for analog and digital transmission

Low noise
Huge bandwidth (several 100 MHz)
Prof. Adam Wolisz

28 of 46

Attenuation of the Coax.

Prof. Adam Wolisz

[GW]

..

29 of 46

Fiber optics...

Prof. Adam Wolisz

[GW]

30 of 46

Attenuation of the fiber...

[GW]

Attractive windows in the Wavelengths: 1.3, 1.5

Prof. Adam Wolisz

31 of 46

Fiber optics (3)

Refractive index and some
representative light path in a multimode
stepped-index optical fiber

Refractive index and some representative

light path in a multimode graded-index
optical fiber. The light is continually
refocused as it travels along the fiber.

Single-mode optical fibers has a core that

is so thin that only one ray (only the
fundamental mode) of light travels along
it. In this way, all the rays of light arrive at
the end at the same time, and modal
dispersion is nonexistent.
Prof. Adam Wolisz

32 of 46

A comparison of the guided media.

Prof. Adam Wolisz

33 of 46

Electromagnetic Spectrum
The light wave region is
expanded for further detail.

Prof. Adam Wolisz

34 of 46

Radio propagation (1)

Prof. Adam Wolisz

35 of 46

Radio Propagation (2)

Prof. Adam Wolisz

36 of 46

RadioPropagation (3) - the relevant for us...

Prof. Adam Wolisz

37 of 46

Propagation path attenuation the theory

Attenuation in free space is inverse proportional to the
square of distance
Attenuation

is stronger for higher frequencies lower

frequencies cover bigger distances with the same
transmission power
Pt 2
Pr =
(4d )2

Pr received power [mW]

Pt transmitted power [mW]

Other environments introduce even stronger attenuation

Pr = P0 / d
Prof. Adam Wolisz

38 of 46

The simplification and the reality...

[Kotz & al]

In addition:
-Reciprocity does not hold (your hear me but I do not read you
-Individual identical radios are NOT identical
See additional reading for details!
Prof. Adam Wolisz

39 of 46

Multipath Fading

The received signal is the sum of the signals arriving along different paths,
Except for the LOS path all paths are the result of reflection and diffraction,
Equalization
Diversity
Prof. Adam Wolisz

40 of 46

Rayleigh fading: Received Signal

10

received signal [dB]

5
0
-5
-10
-15
-20
-25
6km/h
50 km/h

-30
-35
0

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

time [s]

Prof. Adam Wolisz

41 of 46

Doppler Shift
As the transmitter moves towards

the receiver, the propagation time

will change with time t as (1)
The original frequency fc changes
to fc+fd (2)
fd is a shift in the frequency
observed at the receiver
(Doppler frequency shift)
fd is positive if the receiver and
sender move towards each other,
else negative

d (t ) d 0 vmt
(t ) =
=
c
c

vm
fd =
fc
c

(1)

(2)

The Doppler effect constitutes a

source of signal fading.

Prof. Adam Wolisz

42 of 46

REMEMBER: the transmissions my interfere!

[FCC]

Main issue: Interfering with others....

Licencing of the frequencies .....

Competition in (few!) bands dedicated to unlicenced sharing

Recent research: secondary usage of licenced frequencies,
temporarily not used by the owner

Prof. Adam Wolisz

43 of 46

Fiber, radio... Something in common?

In both cases only specific frequency bands are used, different

form the natural spectrum of the signal. Think of voice....

The useful signl has to be shifted where desired...

We will call this pass-band transmission

We do this, using modulation ...

Prof. Adam Wolisz

44 of 46

Pass-band Transmission (Analog Data and Signals!)

We shift the frequency spectrum

elsewhere...

The different kinds of modulation

can be combined

Prof. Adam Wolisz

45 of 46

Why to Modulate Analog on Analog?

Take Amplitude Modulation AM: If the modulated signal is a

sinusoid with frequency fc, and data are within [0 , B] Hz then
the resulting spectrum is from fc to (fc +B) or from (fc -B) to fc

One possible justification: smaller differences in attenuation...

Prof. Adam Wolisz

46 of 46

Digital Data / Digital Signal Encoding

Digital Data:

Sequence of bits to transmit

Physical signal:Physical value e.g. voltage, changing in discrete time

epochs.

Encoding (channel coding): Mapping of bit sequences to signals. This

can be done in many ways...

Example:
1

Unipolar (unbalanced) signaling

Polar (balanced) signaling

Prof. Adam Wolisz

47 of 46

Digital Signal Encoding (2)

Signaling period

Number of signal levels - M

Within the signaling period K changes of the signal level are

possible, the number of changes per second = baud rate !

There are MK signal shapes possible. Usually only some

N < MK signal shapes are permitted.

Achievable bit rate R=(lg2N)/T

Bit rate is not necessarily equal to baud rate

Prof. Adam Wolisz

- T

48 of 46

Digital Signal Encoding (3)

Criteria for selection

Signal spectrum :
lack

of high-frequency component

lack

of dc component (transformers!)

transmitted

power concentrated in the middle of bandwidth

Clocking reconstruction at the receiver

long

periods of constant values avoided

Some error detection features

Polarity of the signal negligible

(only

the change counts)

Noise immunity

Prof. Adam Wolisz

49 of 46

Digital Signal Encoding (4a)

NRZ-L(Nonreturn-to-Zero-Level)

0 = high level

1 = low level

NRZI (Nonreturn-to-Zero-Inverted)

0 = no transition at beginning of interval

1 = transition at beginning of interval

Bipolar -AMI

0 = no line signal

1 = positive or negative level, alternating

Note:

3 level of signal, bit rate=baud rate

Thus: the possible bit rate could be log2 (3)= 1.58,

but really only one bit /signaling time is send....
Prof. Adam Wolisz

50 of 46

The Discussed Codes..

Prof. Adam Wolisz

51 of 46

Digital Signal Encoding (4b)

Pseudoternary

0 = positive or negative level, alternating zeros

1 = no line signal

Manchester

0 = transition high > low in middle of interval

1 = transition low > high in middle of interval

Note: Bit rate lower than baud rate!!

Differential Manchester

Always a transition in middle of interval

0 = transition at beginning of interval

1 = no transition at beginning of interval

Prof. Adam Wolisz

52 of 46

NRZ vs. Manchester...

Prof. Adam Wolisz

53 of 46

Spectral Efficiency

Consider a simple pulse like:

The distribution of the signal energy over the frequency

spectrum is:

Prof. Adam Wolisz

54 of 46

Spectral Efficiency of Codes

Prof. Adam Wolisz

55 of 46

Multilevel Schemes

Note: The bit rate is twice (or 3 times !) the baud rate...
Prof. Adam Wolisz

56 of 46

Digital goes also on analog

Prof. Adam Wolisz

57 of 46

Also in multilevel way...

Think in terms of having multiple

Amlitudes

Frequencies

Phase changes

Think in term of mixing them

(e.g. Amplitude and Phase)

Prof. Adam Wolisz

58 of 46

Which Bandwidth is needed?

Nyquist result: Using M signal levels, In absence of

noise; we could get at most :
C [Bps] = 2B log2 M
where

B is the bandwidth in Hz.

Great idea: just increase the number of signal levels?

Not really... There is noise...

Shannon result: In presence of noise, theoretical

upper bound is:
C [Bps] = B log2 (1 + SNR)
for error free transmission without limits on delay!
Prof. Adam Wolisz

59 of 46

The reality...

Unfortunately limits are usually not achievable....

And we DO have time limits...

As result: there are errors in transmission.

We do ususally look at the measure expressed as

BIT ERROR RATE the ratio of bits in error....

We plot it usually as a function of :

Eb/No

the ratio of (signal energy / bit)

to (noise power density/ Hz)

Prof. Adam Wolisz

60 of 46

Probability of error (BER)

Bit Error Rate for Multilevel PSK (example)

Average energy-per-bit to noise density ratio (Eb/E0) in dB

Prof. Adam Wolisz

61 of 46