Communication Basics

Principles and Dogmas

(C) Herbert Haas

2010/02/15

In this chapter we discuss basic communication issues, such as synchronization,

coding, scrambling, modulation, and so on.

Everything
should be made
as simple as possible,
...but not simpler.
Albert Einstein

Information

What is information?
 Carried by symbols
 Recognized by receiver (hopefully)
 Interpretation is the key

(C) Herbert Haas

2010/02/15

What is information? This question may sound quite easy but think a bit about it.
Obviously we need symbols to represent information. But these symbols must
also be recognized as symbols by the receiver. In fact, philosphical considerations
conclude that information can only be defined through a receiver. The same
problem is with art. What is art? Several decades and centuries had their own
definitions. Today most critics use a general definition: art can only be defined in
context with the viewer.
In the following chapters throughout the whole data communication we will
deal with symbols representing information. A symbol is not a 0 or a 1. But this
binary information can be represented by symbols. Be patient...

Symbols

Symbols (may) represent information

 Voice patterns (Speech)
 Sign language, Pictograms
 Scripture
 Voltage levels
 Light pulses
Blue Whale Sonagrams

(C) Herbert Haas

2010/02/15

What is a good information source? From a theoretical point of view a random

pattern is the best because you'll never know what comes next. On the other hand,
if you receive a continous stream of the same symbol this would be boring. More
than boring: there is no information in it, because you can predict what comes
next! From this we conclude that a sophisticated coding representing the
information as efficient as possible using symbols is a critical step during the
communication process.
Throughout these chapters we will mainly deal with symbols such as voltage
levels or light pulses.
Look at the Blue Whale Sonograms. The x-axis represents time, the y-axis
frequency and the color represents power density. This communication pattern is
very complex (those of dolphins is even more complex). It is known that each
herd has their own traditional hymn. And: they like to communicate!

Symbols on Wire

Discrete voltage levels = "Digital"

 Resistant against noise

How many levels?

 Binary (easiest)
 M-ary: More information per time unit!

Binary

(C) Herbert Haas

M-ary
(here 4 levels, e. g. ISDN)

2010/02/15

What symbols do we encounter on wire? Digital binary symbols are commonly

known and widely in use. Why? Consider information transmissions in groups of
symbols (for example the group of 8 binary symbols is called a byte). We have
two parameters: the number base B and the group order C. If you calculate the
"costs" that you get for arbitrary variations of B and C, and if we assume a linear
progress (so that cost =kBC) then for any given (constant) cost the perfect base
would be B=e, that is B=2.7182...
In other words: the perfect base is a number between 2 and 3. The technical
easiest solution is to use B=2. Note that these considerations assume a linear cost
progression.
In many cases we pay the price of higher efforts and use a larger base. This leads
us to m-ary symbols and later to PAM and QAM.

Synchronization

Sender sends symbol after symbol...

When should receiver pick the signal
samples?

 => Receiver must sync with sender's clock !

Sampling instances

Interpretation:

00001

00001100110
000100111111
001010010111
(only this one is correct)
(C) Herbert Haas

2010/02/15

One of the most important issues among communication is that of

synchronization. Nature forbids absolute synchronization of clocks. Suppose you
are a receiver and you see alternating voltage levels on your receiving interface. If
you had no idea about the sending clock then you would never be able to
interprete the symbols correctly. When do you make a sample?

Synchronization

In reality, two independent clocks are
NEVER precisely synchronous

 We always have a frequency shift

 But we must also care for phase shifts
Different
clock
frequencies

Phase shift
(worst case)
????????????

001010011110

001010011011

(C) Herbert Haas

2010/02/15

So we must assume that the receivers clock is approximately identical to the

senders clock. At least we must deal with small phase and frequency gaps. As you
can see in the slide above, we still cannot be sure when to make samples.
What we need is some kind of synchronization method.

Serial vs Parallel

Parallel transmission





Multiple data wires (fast)

Explicit clocking wire
Simple Synchronization but not cost-effective
Only useful for small distances

Serial transmission
 Only one wire (-pair)
 No clocking wire
 Most important for data communication

(C) Herbert Haas

2010/02/15

In case of parallel transmissions there is always a dedicated clock line. This is a

very comfortable synchronization method. A symbol pattern on the data lines
should be sampled by the receiver each time a clock pulse is observed on the
clock line. But unfortunately, parallel transmissions are too costly on long links.
In LAN and WAN data communication there are practically no parallel lines.
The most important transmission technique is the serial. Data is transmitted over
a single fiber or wire-pair (or electromagentic wave). There is no clock line. How
do we synchronize sender and receiver?

Asynchronous Transmission

Independent clocks
 Oversampling: Much faster than bitrate

Only phase is synchronized

 Using Start-bits and Stop-bits
 Variable intervals between characters
 Synchronity only during transmission

Inefficient
Start-Bit

StartEdge

Stop-Bits

Character

Character

Character

Variable
(C) Herbert Haas

2010/02/15

One synchronization method is the Asynchronous Transmission. Actually this

method cannot provide real synchronization (hence the name) but at least a shorttime quasi-synchronization is possible. The idea is to frame data symbols using
start and stop symbols (lets sloppy call them start- and stop bits). Using
oversampling, the receiver is able to get a sample approximately in the middle of
each bit but only for short bit-sequences.
Asynchronous transmission is typically found in older character-oriented
technologies.
Example application: RS-232C
Relative overhead: 3/11

Synchronous Transmission

Synchronized clocks
 Most important today!
 Phase and Frequency synchronized

Receiver uses a Phased Locked Loop (PLL)

control circuit

 Requires frequent signal changes

 => Coding or Scrambling of data necessary to avoid
long sequences without signal changes

Continous data stream possible

 Large frames possible (theoretically endless)

 Receiver remains synchronized
 Typically each frame starts with a short "training
sequence" aka "preamble" (e. g. 64 bits)

(C) Herbert Haas

2010/02/15

10

The most important method is the Synchronous Transmission. Don't confuse this
with synchronous multiplexingwe are still on the physical layer! Two things are
necessary: a control circuit called Phased-Locked-Loop (PLL) and a signal that
consists of frequent transitions. How do we ensure frequent transitions in our data
stream? Two possibilities: coding and scrambling our data.
Synchronous Transmission is found in most modern bit-oriented technologies
nearly anything you know.

10

Line Coding
NRZ

1 0 1 0 1 1 0 0 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0

RZ
Manchester
Differential
Manchester

NRZI
AMI
Code
Violation

HDB3

(C) Herbert Haas

2010/02/15

11

The trivial code is Non Return to Zero (NRZ) which is usually the human naive
approach.
Remarks:
RZ codes might also use a negative level for logical zeroes, a positive level for
logical ones and a zero Volt level inbetween to return to. RZ is for example used
in optical transmissions (simple modulation).
NRZI codes either modulate for logical ones or zeros. In this slide we modulate
the zeroes, that is each logical zero requires a transition at the beginning of the
interval.
NRZI means Non-return to zero inverted or interchanged.
B8ZS: same as bipolar AMI, except that any string of eight zeros is replaced by a
string of two code violations.
Manchester is used with 10 Mbit Ethernet. Token Ring utilizes Differential
Manchester. Telco backbones (PDH technology) use AMI (USA) or HDB3
(Europe). Of course there are many many other coding styles.

11

Power Spectrum Density

Spectral
Density

NRZ,
NRZI

HDB3

1.0
AMI
Manchester,
Differential Manchester

0.5

0.5

(C) Herbert Haas

1.0

1.5

2.0

Normalized
Frequency (f/R)

2010/02/15

12

The slide above compares the power density distribution of some codes
mentioned before. Obvously the code must match the spectral characteristics of
the transmission channel.
Note that these codes are still kinds of baseband transmissions. Each one can be
modulated using a carrier signal at higher frequency to comply to a specific
channel characteristic.

12

Scrambling Example
t(n-7)

t(n-7)

TS

TS

Example:
Feedback Polynomial = 1+x4+x7
Period length = 127 bit

TS

TS

TS

TS

t(n-4)

TS

TS

TS

TS

TS

TS

TS
s(n)
(C) Herbert Haas

t(n-4)

t(n)

Channel

TS
t(n)

s(n)

2010/02/15

13

Another method to guarantee frequent transitions is scrambling. Scramblers are

used with ATM, SONET/SDH for example.
The feedback polynomial above can be written as
t(n) = s(n) XOR t(n-4) XOR t(n-7)
The descrambler recalculates the original pattern with the same function (change
s(n) with t(n))
Period length = 2^R 1 , where R is the number of shift registers
That is, even a single 1 on the input (and all registers set to 0) will produce a 127bit sequence of pseudo random pattern.
This scrambler is used with 802.11b (Wireless LAN).

13

Transmission System Overview

Source
Coding

Filter unnecessary bits

(Compression)

Source
Decoding

Channel
Coding

FCS and FEC (Checksum)

Error
Detection

Line
Coding

Bandlimited pulses
NRZ, RZ, HDB3, AMI, ...

Modulation

(C) Herbert Haas

Descramber
Equalizer
Signal

Noise

Filter
Demodulator

2010/02/15

DIGITAL

Information
Interpretor

10110001...

ANALOGUE

Information
Source

14

Coding is not coding. The above slide gives you an overview about different
coding purposes. Even modulation is sometimes called coding.
Source coding tries to eliminate redundancy within the information. Source
coders must know well about the type of information that is delivered by the
source.
Channel coding protects the non-redundant data stream by adding calculated
overhead. Typically a Frame Check Sequence (FCS) is added. Only on very
errourness and/or long-delay links a Forward Error Correction (FEC) method
might be useful. FEC requires too much overhead in most terrestial applications.
Line coding focuses on the line, that is we want the symbols to be received
correctly, even if noise and distortions are present. Furthermore line coding
provides clock synchronization as discussed earlier.
Finally modulation might be necessary in case the channel has better properties at
higher frequencies.

14

Communication Channels

Usually Low-Pass behavior
 Higher frequencies are more attenuated than
lower

Baseband transmission
 Signal without a dedicated carrier
 Example: LAN technologies (Ethernet etc)

Carrierband transmission
 The baseband signal modulates a carrier to
match special channel properties
 Medium can be shared for many users (different
carriers) e. g. WLAN
(C) Herbert Haas

2010/02/15

15

Each communication channel exhibits a low-pass behaviorat least beyond a

very high frequency. Not only is the signal attenuated; phase shifts occur and
even nonlinear effects sometimes rise with higher frequencies. The result is a
smeared signal with little energy.
In most cases the signals do not need to be modulated onto a carrier. That is, all
the channel bandwidth can be used up for this signal. We call this baseband
transmission.
Carrier and transmission put the baseband signal onto a carrier with higher
frequency. This is necessary with radio transmissions because low frequencies
have a very bad radiation characteristic. Another example is fiber optics, where
special signal frequencies are significantly more attenuated and scattered than
others.

15

Channel utilization examples

Power
Density

Baseband
Transmission
Frequency

Power
Density

Multiple Carriers
fc1

Power
Density

fc3

Frequency

Telephone
Channel
0.3

(C) Herbert Haas

fc2

3.4

Frequency
(kHz)

2010/02/15

16

The above slide shows some examples for baseband and carrierband
transmission. In case we use multiple carriers we may also call it broadbandtransmission.
The third picture (bottom of slide) shows the spectral characteristic of a telephony
channel (signal). The ITU-T defined an "attenuation-hose" in great detail
(dynamics, ripples, edge frequencies, etc). As a rule of thumb we can expect low
attenuation between 300 Hz and 3400 Hz.

16

Maximal Signal-Rate

Maximal data rate proportional to channelbandwidth B

 Raise time of Heavyside T=1/(2B)

 So the maximum rate is R=2B, also called the Nyquist
Rate
 Note: We assume an ideal channel here without noise!

Bandwidth decreases with cable length

 As a dirty rule of thumb: BW Length @ const
 But note that the reality is much more complex
 Solitons are remarkable exceptions
1

Maximum signal rate: At least

the amplitude must be reached

0
(2B)-1
(C) Herbert Haas

2010/02/15

17

Since each channel is a low-pass, and some channels even damp (very) low
frequencies, data can only be transmitted within a certain channel bandwidth B.
If we put a 0 to 1 transition on the line (with ideally zero transition time), the
receiver will see a slope with a rise time of T=1/(2B).
So the maximal signal rate is T=1/(2B) in theory. In practice we need some
budget because there is noise and distortion and imperfect devices.
The longer the cable the more dramatically the low-pass behaviour. In other
words: on the same cable type we can transmit (let's say) 1,000,000,000 bits/s if
the cable is one meter in length, or only 1 bit/s if the cable is one million
kilometers in length.
It is very interesting to mention that some modern fiber optic transmission
methods violate this basic law. This methods base on so-called SolitonTransmission.

17

The Maximum Information Rate

What about a real channel? What's the

maximum achievable information rate in

presence of noise?

Answer by C. E. Shannon in 1948
 Even when noise is present, information can
be transmitted without errors when the
information rate is below the channel capacity

Channel capacity depends only on

channel bandwidth AND SNR
 Example: AWGN-channel
C = B log (1 + S/N)
(C) Herbert Haas

2010/02/15

18

The great information theory guru Claude E. Shannon made a great discovery in
1948. Before 1948, it was commonly assumed, that there is no way to guarantee
an error-less transmission over a noisy channel. However, Shannon showes that
transmission without errors is possible when the information rate is below the socalled channel capacity, which depends on bandwidth and signal-to-noise ratio.
This discovery is regarded as one of the most important achievements in
communication theory.

18

Bitrate vs Baud

Information Rate: Bit/s

Symbol Rate: Baud

The goal is to send many (=as much as possible)
bits per symbol

 => QAM (see next slides)

N bit/s

2N bit/s

N Baud

N Baud

0 0 1 0 1 0 0 1 0 1 1 1

(C) Herbert Haas

00 10 10 01 01 11

2010/02/15

19

Baud is named after the 19th centurey French inventor Baudot, originally referred
to the speed a telegrapher could send Morse Code.
Today the symbol rate is measured in Baud whereas the information rate is
measured in bit/s.

19

Analogue Modulation Overview

EVERY transmission is analogue but there are different methods to
put a base-band signal onto a high-frequency carrier

The most simple (and oldest) is ASK

 The illustrated ASK method is simple "On-Off-Keying" (OOK)

FSK and PSK are called "angle-modulation" methods (nonlinear =>

spectrum shape is changed!)

For digital transmission, almost always QAM is used
 The BER of BPSK is 3 dB better than for simple OOK

Amplitude Shift Keying (ASK)

t
Phase Shift Keying (PSK)

t
Frequency Shift Keying (FSK)

g (t ) = At cos(2pf t t + jt )
These three parameters can be modulated
(C) Herbert Haas

2010/02/15

20

The slide shows a general modulation equation. The 3 parameters of the equation describe the 3 basic modulation
types. All 3 parameters, the amplitude At, the frequency ft and the phase
t, can be varied, even simultaneously. In
nature, there is no real digital transmission; the binary data stream needs to be converted into an analog signal. As
first step, the digital data will be directly transformed into a analog signal (0 or 1), which is called a baseband
signal. In order to utilize transmission media such as free space (or cables and fibers) the base signal must be mixed
with a carrier signal. This analog modulation shifts the center frequency of the baseband signal to the carrier
frequency to optimize the transmission for a given attenuation/propagation characteristic.
Amplitude Shift Keying
A binary 1 or 0 is represented through different amplitudes of a sinus oscillation. Amplitude Shift Keying (ASK)
requires less bandwidth than FSK or PSK since natura non facit saltus. However ASK is interference prone. This
modulation type also used with infrared-based WLAN.
Frequency Shift Keying
Frequency Shift Keying (FSK) is often used for wireless communication. Different logical signals are represented
by different frequencies. This method needs more bandwidth but is more robust against interferences. To avoid
phase jumps, FSK uses advanced frequency modulators (Continuous Phase Modulation, CPM).
Phase Shift Keying
The 3rd basic modulation method is the Phase Shift Keying (PSK). The digital signal is coding through phase
skipping. In the picture above you see the simplest variation of PSK, using phase jumps of 180. In practice, to
reduce BW, phase jumps must be minimized, and therefore PSK is implemented using advanced phase modulators
(e. g. Gaussian Minimum Shift Keying, etc). The receiver must use same frequency and must be perfectly
synchronized with the sender using a Phase Locked Loop (PLL) circuit. PSK is more robust as FSK against
interferences, but needs complex devices.
After understanding these modulation methods QAM shall be introduced, which is the most important modulation
scheme today for both wired and wireless transmission lines.

20

QAM: Idea

"Quadrature Amplitude Modulation"

Idea:
1. Separate bits in groups of words (e. g. of 6
bits in case of QAM-64)
2. Assign a dedicated pair of Amplitude and
phase to each word (A,
)
3. Create the complex amplitude Aej

4. Create the signal Re{Aej
ejt}
= A (cos
cos t - sin
sin t) which
represents one (of the 64) QAM symbols
5. Receiver can reconstruct (A,
)
(C) Herbert Haas

2010/02/15

21

QAM: Symbol Diagrams

Q

10

Standard
PSK

11

Quadrature
PSK (QPSK)
1

01

00

Other example:
Modem V.29

16-QAM
I

Im{Ui}

For noisy and

distorted channels
4800 bit/s

Re{Ui}
1V

3V

5V

For better channels

7200 bit/s
For even better
channels
9600 bit/s
(C) Herbert Haas

2010/02/15

2400 Baud
Max. 9600 Bit/s

22

Worth to know: Simple Phase Shift Keying (PSK) which only uses two symbols,
each representing either 0 and 1. Quadrature Phase Shift Keying (QPSK) with
four symbols.
Usually the assignment of bit-words to symbols is such that the error probability
due to noise is minimized. For example the Gray-Code may be used between
adjacent symbols to minimize the number of wrong bits when an adjacent symbol
is detected by the receiver.
The above slide also shows the symbol distribution over the complex plane for
the V.29 protocol which is/was used by modems. Depending on the noise-power
of the channel, different sets of symbols are used.
14,400 bit/s requires 64 points
28,800 bit/s requires 128 points

22

Example QAM Applications

One symbol represents a bit pattern
 Given N symbols, each represent ld(N) bits

Modems, 1000BaseT (Gigabit Ethernet),

WiMAX, GSM,

WLAN 802.11a and 802.11g:





(C) Herbert Haas

BPSK @ 6 and 9 Mbps

QPSK @ 12 and 18 Mbps
16-QAM @ 24 and 36 Mbps
64-QAM @ 48 and 54 Mbps

2010/02/15

23

It is important to understand that spread spectrum (or OFDM) techniques are always combined
with a symbol modulation scheme. Quadrature Amplitude Modulation (QAM) is a general method
where practical methods such as BPSK, QPSK, etc are derived from.
The main idea of QAM is to combine phase and amplitude shift keying. Since orthogonal
functions (sine and cosine) are used as carriers, they can be modulated separately, combined into a
single signal, and (due to the orthogonality property) de-combined by the receiver.
And since A*cos(wt + phi) = A/2{cos(wt)cos(phi) sin(wt)sin(phi)} QAM can be easily
represented in the complex domain as Real{ A*exp(i*phi)*exp(i*wt)}.
The standard PSK method only use phase jumps of 0 or 180 to describe a binary 0 or 1. In the
right picture above you see a enhanced PSK method, the Quadrature PSK (QPSK) method. While
using Quadrature PSK each condition (phase shift) represent 2 bits instead of 1. Now it is
possible to transfer the same datarate by halved bandwidth.
The QSK signal uses (relative to reference signal)
- 45 for a data value of 11
- 135 for a data value of 10
- 225 for a data value of 00
- 315 for a data value of 01
To reconstruct the original data stream the receiver need to compare the incoming signal with the
reference signal. The synchronization is very important.
Why not coding more bits per phase jump ?
Especial in the mobile communication there are to much interferences and noise to encode right.
As more bits you use per phase jump, the signal gets more closer. It is getting impossible to
reconstruct the original data stream. In the wireless communication the QPSK method has proven
as a robust and efficient technique.

23

QAM Example Symbols (1)

(C) Herbert Haas

2010/02/15

24

Note that the above QAM signals show different successive QAM-symbols for
illustration purposes. In reality each symbol is transmitted many
hundred/thousand times

24

QAM Example Symbols (2)

(C) Herbert Haas

2010/02/15

25

Note that the above QAM signals show different successive QAM-symbols for
illustration purposes. In reality each symbol is transmitted many
hundred/thousand times
These diagrams have been generated using Octave, a free Matlab clone.

25

The biggest problem

with communication
is the illusion
that it has occured.
Married?

26