Chapter 3

Digital Transmission
Fundamentals
Digital Representation of Information
Why Digital Communications?
Digital Representation of Analog Signals
Characterization of Communication Channels
Fundamental Limits in Digital Transmission
Line Coding
Modems and Digital Modulation
Properties of Media and Digital Transmission Systems
Error Detection and Correction
Digital Networks
z Digital transmission enables networks to
support many services

TV E-mail

Telephone
Questions of Interest
z How long will it take to transmit a message?
z How many bits are in the message (text, image)?
z How fast does the network/system transfer information?
z Can a network/system handle a voice (video) call?
z How many bits/second does voice/video require? At what
quality?
z How long will it take to transmit a message without
errors?
z How are errors introduced?
z How are errors detected and corrected?
z What transmission speed is possible over radio,
copper cables, fiber, infrared, …?
 Chapter 3
Digital Transmission
Fundamentals
Digital Representation of
Information
Bits, numbers, information
z Bit: number with value 0 or 1
z n bits: digital representation for 0, 1, … , 2n
z Byte or Octet, n = 8
z Computer word, n = 16, 32, or 64
z n bits allows enumeration of 2n possibilities
z n-bit field in a header
z n-bit representation of a voice sample
z Message consisting of n bits
z The number of bits required to represent a message
is a measure of its information content
z More bits → More content
Block vs. Stream Information
Block Stream
z Information that occurs z Information that is
in a single block produced & transmitted
z Text message continuously
z Data file z Real-time voice
z JPEG image z Streaming video
z MPEG file
z Size = Bits / block z Bit rate = bits / second
or bytes/block z 1 kbps = 103 bps
z 1 kbyte = 210 bytes z 1 Mbps = 106 bps
z 1 Mbyte = 220 bytes z 1 Gbps =109 bps
z 1 Gbyte = 230 bytes
Transmission Delay
z L number of bits in message
z R bps speed of digital transmission system
z L/R time to transmit the information
z tprop time for signal to propagate across medium
z d distance in meters
z c speed of light (3x108 m/s in vacuum)

Delay = tprop + L/R = d/c + L/R seconds

Use data compression to reduce L

Use higher speed modem to increase R
Place server closer to reduce d
Compression
z Information usually not represented efficiently
z Data compression algorithms
z Represent the information using fewer bits
z Noiseless: original information recovered exactly
z E.g. zip, compress, GIF, fax
z Noisy: recover information approximately
z JPEG
z Tradeoff: # bits vs. quality
z Compression Ratio
#bits (original file) / #bits (compressed file)
 Color Image
W W W W
Red Green Blue
Color component component component
H image = H image + H image + H image

Total bits = 3 × H × W pixels × B bits/pixel = 3HWB bits

Example: 8×10 inch picture at 400 × 400 pixels per inch2

400 × 400 × 8 × 10 = 12.8 million pixels
8 bits/pixel/color
12.8 megapixels × 3 bytes/pixel = 38.4 megabytes
Examples of Block Information

Type Method Format Original Compressed

(Ratio)
Text Zip, ASCII Kbytes- (2-6)
compress Mbytes

Fax CCITT A4 page 256 5-54 kbytes

Group 3 200x100 kbytes (5-50)
pixels/in2
Color JPEG 8x10 in2 photo 38.4 1-8 Mbytes
Image 4002 pixels/in2 Mbytes (5-30)
Stream Information
z A real-time voice signal must be digitized &
transmitted as it is produced
z Analog signal level varies continuously in time

Th e s p ee ch s i g n al l e v el v a r ie s w i th t i m(e)
Digitization of Analog Signal
z Sample analog signal in time and amplitude
z Find closest approximation
Original signal
Sample value

7Δ/2 Approximation
5Δ/2
3 bits / sample

3Δ/2
Δ/2
−Δ/2
−3Δ/2
−5Δ/2
−7Δ/2

Rs = Bit rate = # bits/sample x # samples/second

Bit Rate of Digitized Signal
z Bandwidth Ws Hertz: how fast the signal changes
z Higher bandwidth → more frequent samples
z Minimum sampling rate = 2 x Ws

z Representation accuracy: range of approximation

error
z Higher accuracy
→ smaller spacing between approximation values
→ more bits per sample
Example: Voice & Audio
Telephone voice CD Audio
z Ws = 4 kHz → 8000 z Ws = 22 kHertz → 44000
samples/sec samples/sec
z 8 bits/sample z 16 bits/sample

z Rs=8 x 8000 = 64 kbps z Rs=16 x 44000= 704 kbps

per audio channel
z Cellular phones use z MP3 uses more powerful

more powerful compression algorithms:

compression 50 kbps per audio
algorithms: 8-12 kbps channel
 Video Signal
z Sequence of picture frames
z Each picture digitized &
compressed
z Frame repetition rate
z 10-30-60 frames/second
depending on quality
z Frame resolution
z Small frames for 30 fps
videoconferencing
z Standard frames for
conventional broadcast TV
z HDTV frames

Rate = M bits/pixel x (WxH) pixels/frame x F frames/second

Video Frames
176
QCIF videoconferencing 144 at 30 frames/sec =
760,000 pixels/sec

720
Broadcast TV at 30 frames/sec =
480
10.4 x 106 pixels/sec

1920
HDTV at 30 frames/sec =

1080 67 x 106 pixels/sec

Digital Video Signals
Type Method Format Original Compressed

Video H.261176x144 or 2-36 64-1544

Confer- 352x288 pix Mbps kbps
ence @10-30
fr/sec
Full MPEG 720x480 pix 249 2-6 Mbps
Motion 2 @30 fr/sec Mbps

HDTV MPEG 1920x1080 1.6 19-38 Mbps

2 @30 fr/sec Gbps
Transmission of Stream
Information
z Constant bit-rate
z Signals such as digitized telephone voice produce
a steady stream: e.g. 64 kbps
z Network must support steady transfer of signal,
e.g. 64 kbps circuit
z Variable bit-rate
z Signals such as digitized video produce a stream
that varies in bit rate, e.g. according to motion and
detail in a scene
z Network must support variable transfer rate of
signal, e.g. packet switching or rate-smoothing
with constant bit-rate circuit
Stream Service Quality Issues
Network Transmission Impairments
z Delay: Is information delivered in timely
fashion?
z Jitter: Is information delivered in sufficiently
smooth fashion?
z Loss: Is information delivered without loss? If
loss occurs, is delivered signal quality
acceptable?
z Applications & application layer protocols
developed to deal with these impairments
 Chapter 3
Communication
Networks and Services
Why Digital Communications?
A Transmission System
Transmitter Receiver

Communication channel

Transmitter
z Converts information into signal suitable for transmission
z Injects energy into communications medium or channel
z Telephone converts voice into electric current
z Modem converts bits into tones
Receiver
z Receives energy from medium
z Converts received signal into form suitable for delivery to user
z Telephone converts current into voice
z Modem converts tones into bits
Transmission Impairments
Transmitted Received
Transmitter
Signal Signal Receiver

Communication channel

Communication Channel Transmission Impairments

z Pair of copper wires z Signal attenuation
z Coaxial cable z Signal distortion
z Radio z Spurious noise
z Light in optical fiber z Interference from other
z Light in air signals
z Infrared
 Analog Long-Distance
Communications
Transmission segment

Source Repeater ... Repeater Destination

z Each repeater attempts to restore analog signal to

its original form
z Restoration is imperfect
z Distortion is not completely eliminated
z Noise & interference is only partially removed
z Signal quality decreases with # of repeaters
z Communications is distance-limited
z Still used in analog cable TV systems
z Analogy: Copy a song using a cassette recorder
Analog vs. Digital Transmission
Analog transmission: all details must be reproduced accurately
Distortion
Sent Attenuation Received

Digital transmission: only discrete levels need to be reproduced

Sent Distortion Received

Simple Receiver:
Attenuation
Was original
pulse positive or
negative?
 Digital Long-Distance
Communications
Transmission segment

Source Regenerator ... Regenerator Destination

z Regenerator recovers original data sequence and

retransmits on next segment
z Can design so error probability is very small
z Then each regeneration is like the first time!
z Analogy: copy an MP3 file
z Communications is possible over very long distances
z Digital systems vs. analog systems
z Less power, longer distances, lower system cost
z Monitoring, multiplexing, coding, encryption, protocols…
Digital Binary Signal
1 0 1 1 0 1
+A

0 T 2T 3T 4T 5T 6T
-A

Bit rate = 1 bit / T seconds

For a given communications medium:
z How do we increase transmission speed?

z How do we achieve reliable communications?

z Are there limits to speed and reliability?

Pulse Transmission Rate
z Objective: Maximize pulse rate through a channel,
that is, make T as small as possible

Channel
T t t

z If input is a narrow pulse, then typical output is a

spread-out pulse with ringing
z Question: How frequently can these pulses be
transmitted without interfering with each other?
z Answer: 2 x Wc pulses/second

where Wc is the bandwidth of the channel

 Bandwidth of a Channel
X(t) = a cos(2πft) Channel Y(t) = A(f) a cos(2πft)

z If input is sinusoid of frequency f, A(f)

1
then
z output is a sinusoid of same frequency f
f
z Output is attenuated by an amount A(f) 0 Wc
that depends on f
z A(f)≈1, then input signal passes readily Ideal low-pass
z A(f)≈0, then input signal is blocked channel

z Bandwidth Wc is range of
frequencies passed by channel
Multilevel Pulse Transmission
z Assume channel of bandwidth Wc, and transmit 2 Wc
pulses/sec (without interference)
z If pulses amplitudes are either -A or +A, then each
pulse conveys 1 bit, so
Bit Rate = 1 bit/pulse x 2Wc pulses/sec = 2Wc bps
z If amplitudes are from {-A, -A/3, +A/3, +A}, then bit
rate is 2 x 2Wc bps
z By going to M = 2m amplitude levels, we achieve
Bit Rate = m bits/pulse x 2Wc pulses/sec = 2mWc bps

In the absence of noise, the bit rate can be increased

without limit by increasing m
Noise & Reliable Communications
z All physical systems have noise
z Electrons always vibrate at non-zero temperature
z Motion of electrons induces noise
z Presence of noise limits accuracy of measurement
of received signal amplitude
z Errors occur if signal separation is comparable to
noise level
z Bit Error Rate (BER) increases with decreasing
signal-to-noise ratio
z Noise places a limit on how many amplitude levels
can be used in pulse transmission
Signal-to-Noise Ratio
Signal Noise Signal + noise
High
SNR
t t t

No errors

Signal Noise Signal + noise

Low
SNR
t t t

Average signal power error

SNR =
Average noise power

SNR (dB) = 10 log10 SNR

Shannon Channel Capacity
C = Wc log2 (1 + SNR) bps
z Arbitrarily reliable communications is possible if the
transmission rate R < C.
z If R > C, then arbitrarily reliable communications is
not possible.
z “Arbitrarily reliable” means the BER can be made
arbitrarily small through sufficiently complex coding.
z C can be used as a measure of how close a system
design is to the best achievable performance.
z Bandwidth Wc & SNR determine C
Example
z Find the Shannon channel capacity for a
telephone channel with Wc = 3400 Hz and
SNR = 10000

C = 3400 log2 (1 + 10000)

= 3400 log10 (10001)/log102 = 45200 bps

Note that SNR = 10000 corresponds to

SNR (dB) = 10 log10(10001) = 40 dB
 Bit Rates of Digital Transmission
Systems

System Bit Rate Observations

Telephone 33.6-56 kbps 4 kHz telephone channel
twisted pair
Ethernet 10 Mbps, 100 Mbps 100 meters of unshielded
twisted pair twisted copper wire pair
Cable modem 500 kbps-4 Mbps Shared CATV return channel
ADSL twisted 64-640 kbps in, 1.536- Coexists with analog
pair 6.144 Mbps out telephone signal
2.4 GHz radio 2-11 Mbps IEEE 802.11 wireless LAN
28 GHz radio 1.5-45 Mbps 5 km multipoint radio
Optical fiber 2.5-10 Gbps 1 wavelength
Optical fiber >1600 Gbps Many wavelengths
Examples of Channels
Channel Bandwidth Bit Rates

Telephone 3 kHz 33 kbps

voice channel
Copper pair 1 MHz 1-6 Mbps

Coaxial cable 500 MHz 30 Mbps/

(6 MHz channels) channel
5 GHz radio 300 MHz 54 Mbps /
(IEEE 802.11) (11 channels) channel
Optical fiber Many TeraHertz 40 Gbps /
wavelength
 Chapter 3
Digital Transmission
Fundamentals
Digital Representation of
Analog Signals
Digitization of Analog Signals
1. Sampling: obtain samples of x(t) at uniformly
spaced time intervals
2. Quantization: map each sample into an
approximation value of finite precision
z Pulse Code Modulation: telephone speech
z CD audio
3. Compression: to lower bit rate further, apply
additional compression method
z Differential coding: cellular telephone speech
z Subband coding: MP3 audio
z Compression discussed in Chapter 12
 Sampling Rate and Bandwidth
z A signal that varies faster needs to be sampled
more frequently
z Bandwidth measures how fast a signal varies
10 10 1 0 1 0 11 1 1 0 000
x1(t) x2(t)
... ... ... ...

t t

1 ms 1 ms

z What is the bandwidth of a signal?

z How is bandwidth related to sampling rate?
Periodic Signals
z A periodic signal with period T can be represented
as sum of sinusoids using Fourier Series:
x(t) = a0 + a1cos(2πf0t + φ1) + a2cos(2π2f0t + φ2) + …
+ akcos(2πkf0t + φk) + …

“DC” fundamental
long-term frequency f0=1/T kth harmonic
average first harmonic

•|ak| determines amount of power in kth harmonic

•Amplitude specturm |a0|, |a1|, |a2|, …

 Example Fourier Series
10 10 1 0 1 0 11 1 1 0 000
x1(t) x2(t)
... ... ... ...

t t

T2 =0.25 ms T1 = 1 ms

4 4
x1(t) = 0 + cos(2π4000t) x2(t) = 0 + cos(2π1000t)
π π
4 4
+ cos(2π3(4000)t) + cos(2π3(1000)t)
3π 3π
4 4
+ cos(2π5(4000)t) + … + cos(2π5(1000)t) + …
5π 5π

Only odd harmonics have power

Spectra & Bandwidth
Spectrum of x1(t)
z Spectrum of a signal: 1.2

magnitude of amplitudes as 1

0.8

a function of frequency 0.6

0.4

z x1(t) varies faster in time & 0.2

has more high frequency

0

9
12

15

18

21

24

27

30

33

36

39

42
content than x2(t) frequency (kHz)

z Bandwidth Ws is defined as Spectrum of x2(t)

range of frequencies where 1.2

a signal has non-negligible

1

0.8

power, e.g. range of band 0.6

0.4

that contains 99% of total 0.2

signal power
0
0

9
12

15

18

21

24

27

30

33

36

39

42
frequency (kHz)
Bandwidth of General Signals
“speech”

s (noisy ) |p (air stopped) | ee (periodic) | t (stopped) | sh (noisy)

z Not all signals are periodic

X(f)
z E.g. voice signals varies
according to sound
z Vowels are periodic, “s” is
noiselike
z Spectrum of long-term signal
z Averages over many sounds,
many speakers f
z Involves Fourier transform 0 Ws
z Telephone speech: 4 kHz
z CD Audio: 22 kHz
Sampling Theorem

Nyquist: Perfect reconstruction if sampling rate 1/T > 2Ws

(a) x(t) x(nT)

t Sampler t

(b)

x(nT) x(t)

t Interpolation t
filter
Digital Transmission of Analog
Information
2W samples / sec m bits / sample
Analog Sampling Quantization
source (A/D)

Original x(t) 2W m bits/sec

Bandwidth W

Transmission
or storage
Approximation y(t)

Display Interpolation Pulse

or filter generator
playout

2W samples / sec
Quantization of Analog Samples
3.5Δ
Quantizer maps input
output y(nT) into closest of 2m
2.5Δ
1.5Δ representation values
0.5Δ
−4Δ −3Δ −2Δ −Δ
−0.5Δ Δ 2Δ 3Δ 4Δ
−1.5Δ input x(nT) Quantization error:
−2.5Δ “noise” = x(nT) – y(nT)
−3.5Δ

Original signal
Sample value
7Δ/2 Approximation
3 bits / sample

5Δ/2
3Δ/2
Δ/2
-Δ/2
-3Δ/2
-5Δ/2
-7Δ/2
Quantizer Performance
M = 2m levels, Dynamic range( -V, V) ∆ = 2V/M
error = y(nT)-x(nT)=e(nT)
Δ
2
... −2Δ Δ Δ 2Δ 3Δ ... input

V x(nT)
-V −
Δ
2

If the number of levels M is large, then the error is

approximately uniformly distributed between (-∆/2, ∆2)
Average Noise Power = Mean Square Error:
∆
1 ∆2
∫
2
σe =2 x2 dx =
∆ ∆ 12
2
Quantizer Performance
Figure of Merit:
Signal-to-Noise Ratio = Avg signal power / Avg noise power
Let σx2 be the signal power, then
σx2 12 σ 2 σx σx
SNR = = x
= 3( )2 M2
Δ2/12 4V2/M2 = 3( )2 22m
V V
The ratio V/σx ≈ 4

The SNR is usually stated in decibels:

SNR db = 10 log10 σx2/σe2 = 6 + 10 log10
3σx2/V2
SNR db = 6m - 7.27 dB for V/σx = 4.
Example: Telephone Speech
W = 4KHz, so Nyquist sampling theorem
⇒ 2W = 8000 samples/second
Suppose error requirement = 1% error
SNR = 10 log(1/.01)2 = 40 dB
Assume V/σx =4, then
40 dB = 6m – 7
⇒ m = 8 bits/sample

PCM (“Pulse Code Modulation”) Telephone

Speech:
Bit rate= 8000 x 8 bits/sec= 64 kbps
 Chapter 3
Digital Transmission
Fundamentals
Characterization of
Communication Channels
Communications Channels
z A physical medium is an inherent part of a
communications system
z Copper wires, radio medium, or optical fiber
z Communications system includes electronic or
optical devices that are part of the path followed by
a signal
z Equalizers, amplifiers, signal conditioners
z By communication channel we refer to the combined
end-to-end physical medium and attached devices
z Sometimes we use the term filter to refer to a
channel especially in the context of a specific
mathematical model for the channel
How good is a channel?
z Performance: What is the maximum reliable
transmission speed?
z Speed: Bit rate, R bps
z Reliability: Bit error rate, BER=10-k
z Focus of this section
z Cost: What is the cost of alternatives at a
given level of performance?
z Wired vs. wireless?
z Electronic vs. optical?
z Standard A vs. standard B?
Communications Channel
Transmitter Transmitted Received
Receiver
Signal Signal

Communication channel

Signal Bandwidth Transmission Impairments

z In order to transfer data z Signal attenuation
faster, a signal has to vary z Signal distortion
more quickly.
z Spurious noise
Channel Bandwidth
z Interference from other
z A channel or medium has
signals
an inherent limit on how fast
z Limits accuracy of
the signals it passes can
vary measurements on received
signal
z Limits how tightly input
pulses can be packed
Frequency Domain Channel
Characterization
x(t)= Aincos 2πft y(t)=Aoutcos (2πft + ϕ(f))

Channel
t t
Aout
A(f) = Ain

z Apply sinusoidal input at frequency f

z Output is sinusoid at same frequency, but attenuated & phase-shifted
z Measure amplitude of output sinusoid (of same frequency f)
z Calculate amplitude response
z A(f) = ratio of output amplitude to input amplitude
z If A(f) ≈ 1, then input signal passes readily
z If A(f) ≈ 0, then input signal is blocked
z Bandwidth Wc is range of frequencies passed by channel
Ideal Low-Pass Filter
z Ideal filter: all sinusoids with frequency f<Wc are
passed without attenuation and delayed by τ seconds;
sinusoids at other frequencies are blocked

y(t)=Aincos (2πft - 2πfτ )= Aincos (2πf(t - τ )) = x(t-τ)

Amplitude Response Phase Response

1 ϕ(f) = -2πft
1/ 2π
0
f

Wc f
Example: Low-Pass Filter
z Simplest non-ideal circuit that provides low-pass filtering
z Inputs at different frequencies are attenuated by different amounts
z Inputs at different frequencies are delayed by different amounts

Amplitude Response Phase Response

A(f) = 1 ϕ(f) = tan-1 2πf
1
(1+4π2f2)1/2 1/ 2π
0
f

-45o

-90o
f
Example: Bandpass Channel
Amplitude Response

A(f)

Wc f

z Some channels pass signals within a band that

excludes low frequencies
z Telephone modems, radio systems, …
z Channel bandwidth is the width of the frequency band
that passes non-negligible signal power
Channel Distortion
x(t) = Σ ak cos (2πfkt + θk) Channel y(t)

z Let x(t) corresponds to a digital signal bearing data

information
z How well does y(t) follow x(t)?

y(t) = ΣA(fk) ak cos (2πfkt + θk + Φ(fk ))

z Channel has two effects:

z If amplitude response is not flat, then different frequency
components of x(t) will be transferred by different amounts
z If phase response is not flat, then different frequency
components of x(t) will be delayed by different amounts
z In either case, the shape of x(t) is altered
Example: Amplitude Distortion
x(t) 1 0 0 0 0 0 0 1
... ...

1 ms
t

z Let x(t) input to ideal lowpassπfilter that has zero delay and
Wc = 1.5 kHz, 2.5 kHz, or 4.5 kHz

sin( π )cos(2π1000t)
4
x(t) = -0.5 +
π 4
sin( 2π )cos(2π2000t) + sin(3π )cos(2π3000t) + …
4 4
+
π 4 π 4
z Wc = 1.5 kHz passes only the first two terms
z Wc = 2.5 kHz passes the first three terms
z Wc = 4.5 kHz passes the first five terms
Amplitude Distortion
1.5 (a) 1 Harmonic
1
0.5 z As the channel
0
bandwidth
0

0.125

0.25

0.375

0.625

0.75

0.875

1
0.5
-0.5
-1
-1.5
increases, the
output of the
1.5 (b) 2 Harmonics channel
1
0.5
resembles the
0
input more
0.125

0.375

0.625

0.875
0

0.25

0.5

0.75

1
-0.5
-1
-1.5
closely

(c) 4 Harmonics
1.5
1
0.5
0
0.125

0.375

0.625

0.875
0.25

0.75
0

0.5

-0.5
-1
-1.5
Time-domain Characterization
h(t)

Channel
t
0 t
td

z Time-domain characterization of a channel requires

finding the impulse response h(t)
z Apply a very narrow pulse to a channel and observe
the channel output
z h(t) typically a delayed pulse with ringing
z Interested in system designs with h(t) that can be
packed closely without interfering with each other
 Nyquist Pulse with Zero
Intersymbol Interference
z For channel with ideal lowpass amplitude response of
bandwidth Wc, the impulse response is a Nyquist pulse
h(t)=s(t – τ), where T = 1/2 Wc, and
s(t) = sin(2πWc t)/ 2πWct
1.2
1
0.8
0.6
0.4
0.2
0 t
-7T -6T -5T -4T -3T -2T T
-1-0.2 0 1T 2T 3T 4T 5T 6T 7T

-0.4

z s(t) has zero crossings at t = kT, k = +1, +2, …

z Pulses can be packed every T seconds with zero interference
 Example of composite waveform
Three Nyquist pulses +s(t) +s(t-T)
1
shown separately
z + s(t)

z + s(t-T) 0 t
-2 T -1T 0 1T 2T 3T 4T
z - s(t-2T)

Composite waveform
-1
r(t) = s(t)+s(t-T)-s(t-2T) -s(t-2T)
Samples at kT r(t) 2

r(0)=s(0)+s(-T)-s(-2T)=+1 1

r(T)=s(T)+s(0)-s(-T)=+1
0 t
r(2T)=s(2T)+s(T)-s(0)=-1 -2T -1T 0 1T 2T 3T 4T
-1
Zero ISI at sampling
times kT -2
Nyquist pulse shapes
z If channel is ideal low pass with Wc, then pulses maximum
rate pulses can be transmitted without ISI is T = 1/2Wc sec.
z s(t) is one example of class of Nyquist pulses with zero ISI
z Problem: sidelobes in s(t) decay as 1/t which add up quickly
when there are slight errors in timing
z Raised cosine pulse below has zero ISI
z Requires slightly more bandwidth than Wc

z Sidelobes decay as 1/t3, so more robust to timing errors

A(f) 1
sin(πt/T) cos(παt/T)
πt/T 1 – (2αt/T)2

(1 – α)Wc Wc (1 + α)Wc f
0
 Chapter 3
Digital Transmission
Fundamentals
Fundamental Limits in Digital
Transmission
 Signaling with Nyquist Pulses
z p(t) pulse at receiver in response to a single input pulse (takes
into account pulse shape at input, transmitter & receiver filters,
and communications medium)
z r(t) waveform that appears in response to sequence of pulses
z If s(t) is a Nyquist pulse, then r(t) has zero intersymbol
interference (ISI) when sampled at multiples of T

1 0 1 1 0 1
+A
0 T 2T 3T 4T 5T t
-A

Transmitter Communication Receiver r(t)

Filter Medium Filter Receiver

Received signal
Multilevel Signaling
z Nyquist pulses achieve the maximum signalling rate with zero
ISI,
2Wc pulses per second or
2Wc pulses / Wc Hz = 2 pulses / Hz

z With two signal levels, each pulse carries one bit of

information
Bit rate = 2Wc bits/second

z With M = 2m signal levels, each pulse carries m bits

Bit rate = 2Wc pulses/sec. * m bits/pulse = 2Wc m bps

z Bit rate can be increased by increasing number of levels

z r(t) includes additive noise, that limits number of levels that
can be used reliably.
Example of Multilevel Signaling
z Four levels {-1, -1/3, 1/3, +1} for {00,01,10,11}
z Waveform for 11,10,01 sends +1, +1/3, -1/3
z Zero ISI at sampling instants

1.2

0.8
Composite waveform
0.6

0.4

0.2

0
-1 0 1 2 3
-0.2

-0.4

-0.6
 Noise Limits Accuracy
z Receiver makes decision based on transmitted pulse level + noise
z Error rate depends on relative value of noise amplitude and spacing
between signal levels
z Large (positive or negative) noise values can cause wrong decision
z Noise level below impacts 8-level signaling more than 4-level signaling

+A +A
+5A/7
+A/3 +3A/7
+A/7
-A/7
-A/3
-3A/7
Typical noise
-5A/7
-A -A

Four signal levels Eight signal levels

 Noise distribution
z Noise is characterized by probability density of amplitude samples
z Likelihood that certain amplitude occurs
z Thermal electronic noise is inevitable (due to vibrations of electrons)
z Noise distribution is Gaussian (bell-shaped) as below

σ2 = Avg Noise Power

x
x0

Pr[X(t)>x0 ] = ? t

Pr[X(t)>x0 ] =
1 − x 2 2σ 2
e Area under
2π σ graph

0 x0 x
 Probability of Error
z Error occurs if noise value exceeds certain magnitude
z Prob. of large values drops quickly with Gaussian noise
z Target probability of error achieved by designing system so
separation between signal levels is appropriate relative to
average noise power

0 2 4 6 8
1.00E+00 δ/2σ
1.00E-01
1.00E-02
1.00E-03
1.00E-04
Pr[X(t)>δ ] 1.00E-05
1.00E-06
1.00E-07
1.00E-08
1.00E-09
1.00E-10
1.00E-11
1.00E-12
Channel Noise affects Reliability

signal noise signal + noise

High
SNR

virtually error-free
signal noise signal + noise
Low
SNR
error-prone
Average Signal Power
SNR =
Average Noise Power

SNR (dB) = 10 log10 SNR

 Shannon Channel Capacity
z If transmitted power is limited, then as M increases spacing
between levels decreases
z Presence of noise at receiver causes more frequent errors
to occur as M is increased

Shannon Channel Capacity:

The maximum reliable transmission rate over an ideal channel
with bandwidth W Hz, with Gaussian distributed noise, and
with SNR S/N is
C = W log2 ( 1 + S/N ) bits per second

z Reliable means error rate can be made arbitrarily small by

proper coding
Example
z Consider a 3 kHz channel with 8-level signaling.
Compare bit rate to channel capacity at 20 dB SNR

z 3KHz telephone channel with 8 level signaling

Bit rate = 2*3000 pulses/sec * 3 bits/pulse = 18 kbps

z 20 dB SNR means 10 log10 S/N = 20

Implies S/N = 100
z Shannon Channel Capacity is then
C = 3000 log ( 1 + 100) = 19, 963 bits/second
 Chapter 3
Digital Transmission
Fundamentals
Line Coding
What is Line Coding?
z Mapping of binary information sequence into the
digital signal that enters the channel
z Ex. “1” maps to +A square pulse; “0” to –A pulse
z Line code selected to meet system requirements:
z Transmitted power: Power consumption = $
z Bit timing: Transitions in signal help timing recovery
z Bandwidth efficiency: Excessive transitions wastes bw
z Low frequency content: Some channels block low
frequencies
z long periods of +A or of –A causes signal to “droop”
z Waveform should not have low-frequency content
z Error detection: Ability to detect errors helps
z Complexity/cost: Is code implementable in chip at high
speed?
Line coding examples
1 0 1 0 1 1 1 0 0
Unipolar
NRZ

Polar NRZ

NRZ-inverted
(differential
encoding)

Bipolar
encoding

Manchester
encoding

Differential
Manchester
encoding
Spectrum of Line codes
z Assume 1s & 0s independent & equiprobable

z NRZ has high

content at low
1.2 frequencies
NRZ z Bipolar tightly
1
packed around T/2
0.8 Bipolar
z Manchester wasteful
pow er density

0.6
of bandwidth
0.4
Manchester
0.2

0
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

2
-0.2
fT
Unipolar & Polar
Non-Return-to-Zero (NRZ)
1 0 1 0 1 1 1 0 0

Unipolar NRZ

Polar NRZ

Unipolar NRZ Polar NRZ

z “1” maps to +A pulse z “1” maps to +A/2 pulse
z “0” maps to no pulse z “0” maps to –A/2 pulse
z High Average Power z Better Average Power
0.5*A2 +0.5*02=A2/2 0.5*(A/2)2 +0.5*(-A/2)2=A2/4
z Long strings of A or 0 z Long strings of +A/2 or –A/2
z Poor timing z Poor timing
z Low-frequency content
z Low-frequency content
z Simple z Simple
Bipolar Code
1 0 1 0 1 1 1 0 0

Bipolar
Encoding

z Three signal levels: {-A, 0, +A}

z “1” maps to +A or –A in alternation
z “0” maps to no pulse
z Every +pulse matched by –pulse so little content at low
frequencies
z String of 1s produces a square wave
z Spectrum centered at T/2
z Long string of 0s causes receiver to lose synch
z Zero-substitution codes
Manchester code & mBnB codes
1 0 1 0 1 1 1 0 0
Manchester
Encoding

z “1” maps into A/2 first T/2, -A/2 z mBnB line code
last T/2 z Maps block of m bits into n
z “0” maps into -A/2 first T/2, A/2 bits
last T/2 z Manchester code is 1B2B
z Every interval has transition in code
middle z 4B5B code used in FDDI
z Timing recovery easy LAN
z Uses double the z 8B10b code used in Gigabit
minimum bandwidth Ethernet
z Simple to implement z 64B66B code used in 10G
z Used in 10-Mbps Ethernet & Ethernet
other LAN standards
Differential Coding
1 0 1 0 1 1 1 0 0
NRZ-inverted
(differential
encoding)

Differential
Manchester
encoding

z Errors in some systems cause transposition in polarity, +A become

–A and vice versa
z All subsequent bits in Polar NRZ coding would be in error
z Differential line coding provides robustness to this type of error
z “1” mapped into transition in signal level
z “0” mapped into no transition in signal level
z Same spectrum as NRZ
z Errors occur in pairs
z Also used with Manchester coding
 Chapter 3
Digital Transmission
Fundamentals
Modems and Digital
Modulation
Bandpass Channels

0 fc – Wc/2 fc fc + Wc/2

z Bandpass channels pass a range of frequencies

around some center frequency fc
z Radio channels, telephone & DSL modems
z Digital modulators embed information into waveform
with frequencies passed by bandpass channel
z Sinusoid of frequency fc is centered in middle of
bandpass channel
z Modulators embed information into a sinusoid
 Amplitude Modulation and
Frequency Modulation
Information 1 0 1 1 0 1

+1
Amplitude
Shift t
Keying 0 T 2T 3T 4T 5T 6T
-1
Map bits into amplitude of sinusoid: “1” send sinusoid; “0” no sinusoid
Demodulator looks for signal vs. no signal

+1
Frequency
Shift t
0 T 2T 3T 4T 5T 6T
Keying
-1
Map bits into frequency: “1” send frequency fc + δ ; “0” send frequency fc - δ
Demodulator looks for power around fc + δ or fc - δ
 Phase Modulation
Information 1 0 1 1 0 1
+1
Phase
Shift
t
Keying 0 T 2T 3T 4T 5T 6T
-1

z Map bits into phase of sinusoid:

z “1” send A cos(2πft) , i.e. phase is 0
z “0” send A cos(2πft+π) , i.e. phase is π

z Equivalent to multiplying cos(2πft) by +A or -A

z “1” send A cos(2πft) , i.e. multiply by 1
z “0” send A cos(2πft+π) = - A cos(2πft) , i.e. multiply by -1

z We will focus on phase modulation

Modulator & Demodulator
Modulate cos(2πfct) by multiplying by Ak for T seconds:

Ak x Yi(t) = Ak cos(2πfct)

cos(2πfct) Transmitted signal

during kth interval

Demodulate (recover Ak) by multiplying by 2cos(2πfct)

for T seconds and lowpass filtering (smoothing):

Lowpass
Yi(t) = Akcos(2πfct) x Filter Xi(t)
(Smoother)
Received signal
during kth interval 2cos(2πfct)
2Ak cos2(2πfct) = Ak {1 + cos(2π2fct)}
Example of Modulation

Information 1 0 1 1 0 1
+A
Baseband
Signal
0 T 2T 3T 4T 5T 6T
-A

+A
Modulated
Signal T 2T 4T 5T
0 3T 6T
x(t)
-A

A cos(2πft) -A cos(2πft)
 Example of Demodulation
A {1 + cos(4πft)} -A {1 + cos(4πft)}

After multiplication
+A
at receiver
0 T 2T 3T 4T 5T 6T
x(t) cos(2πfct)
-A

+A
Baseband
signal discernable
after smoothing 0 T 2T 3T 4T 5T 6T
-A

Recovered
Information 1 0 1 1 0 1
 Signaling rate and
Transmission Bandwidth
z Fact from modulation theory:
If
Baseband signal x(t)
with bandwidth B Hz
then f
B
Modulated signal
x(t)cos(2πfct) has
bandwidth 2B Hz f
fc-B fc fc+B

z If bandpass channel has bandwidth Wc Hz,

z Then baseband channel has Wc/2 Hz available, so
z modulation system supports Wc/2 x 2 = Wc pulses/second
z That is, Wc pulses/second per Wc Hz = 1 pulse/Hz
z Recall baseband transmission system supports 2 pulses/Hz
Quadrature Amplitude Modulation
(QAM)
z QAM uses two-dimensional signaling
z Ak modulates in-phase cos(2πfct)
z Bk modulates quadrature phase cos(2πfct + π/4) = sin(2πfct)
z Transmit sum of inphase & quadrature phase components

Ak x Yi(t) = Ak cos(2πfct)

cos(2πfct) + Y(t)
Transmitted
Bk x Yq(t) = Bk sin(2πfct) Signal

sin(2πfct)

z Yi(t) and Yq(t) both occupy the bandpass channel

z QAM sends 2 pulses/Hz
QAM Demodulation

Lowpass
Y(t) x filter Ak
(smoother)
2cos(2πfct)
2cos2(2πfct)+2Bk cos(2πfct)sin(2πfct)
= Ak {1 + cos(4πfct)}+Bk {0 + sin(4πfct)}
smoothed to zero
Lowpass
x filter Bk
(smoother)
2sin(2πfct)
2Bk sin2(2πfct)+2Ak cos(2πfct)sin(2πfct)
= Bk {1 - cos(4πfct)}+Ak {0 + sin(4πfct)}

smoothed to zero
Signal Constellations
z Each pair (Ak, Bk) defines a point in the plane
z Signal constellation set of signaling points
Bk
Bk
(-A,A) (A, A)

Ak Ak

(-A,-A) (A,-A)

4 possible points per T sec. 16 possible points per T sec.

2 bits / pulse 4 bits / pulse
Other Signal Constellations
z Point selected by amplitude & phase
Ak cos(2πfct) + Bk sin(2πfct) = √Ak2 + Bk2 cos(2πfct + tan-1(Bk/Ak))

Bk Bk

Ak Ak

4 possible points per T sec. 16 possible points per T sec.

 Telephone Modem Standards
Telephone Channel for modulation purposes has
Wc = 2400 Hz → 2400 pulses per second

Modem Standard V.32bis

z Trellis modulation maps m bits into one of 2m+1 constellation
points
z 14,400 bps Trellis 128 2400x6
z 9600 bps Trellis 32 2400x4
z 4800 bps QAM 4 2400x2

Modem Standard V.34 adjusts pulse rate to channel

z 2400-33600 bps Trellis 960 2400-3429 pulses/sec
 Chapter 3
Digital Transmission
Fundamentals
Properties of Media and Digital
Transmission Systems
Fundamental Issues in
Transmission Media
d meters

Communication channel

t = d/c
t=0
z Information bearing capacity
z Amplitude response & bandwidth
z dependence on distance
z Susceptibility to noise & interference
z Error rates & SNRs

z Propagation speed of signal

z c = 3 x 108 meters/second in vacuum
z ν = c/√ε speed of light in medium where ε>1 is the
dielectric constant of the medium
z ν = 2.3 x 108 m/sec in copper wire; ν = 2.0 x 108 m/sec in
optical fiber
Communications systems &
Electromagnetic Spectrum
z Frequency of communications signals
Analog Optical
DSL Cell WiFi
telephone fiber
phone Frequency (Hz)

102 104 106 108 1010 1012 1014 1016 1018 1020 1022 1024

Ultraviolet light

Gamma rays
Infrared light
Power and
telephone

Microwave
Broadcast

Visible light
radio

radio

X-rays
106 104 102 10 10-2 10-4 10-6 10-8 10-10 10-12 10-14
Wavelength (meters)
Wireless & Wired Media
Wireless Media Wired Media
z Signal energy propagates in z Signal energy contained &
space, limited directionality guided within medium
z Interference possible, so z Spectrum can be re-used in
spectrum regulated separate media (wires or
z Limited bandwidth cables), more scalable
z Simple infrastructure: z Extremely high bandwidth

antennas & transmitters z Complex infrastructure:

z No physical connection ducts, conduits, poles, right-

between network & user of-way
z Users can move
Attenuation
z Attenuation varies with media
z Dependence on distance of central importance
z Wired media has exponential dependence
z Received power at d meters proportional to 10-kd
z Attenuation in dB = k d, where k is dB/meter
z Wireless media has logarithmic dependence
z Received power at d meters proportional to d-n
z Attenuation in dB = n log d, where n is path loss exponent;
n=2 in free space
z Signal level maintained for much longer distances
z Space communications possible
 Twisted Pair
Twisted pair
z Two insulated copper wires 30 26 gauge
arranged in a regular spiral 24 gauge
pattern to minimize
interference 24

Attenuation (dB/mi)
z Various thicknesses, e.g. 22 gauge
0.016 inch (24 gauge)
18
z Low cost
19 gauge
z Telephone subscriber loop
from customer to CO 12
z Old trunk plant connecting
telephone COs
z Intra-building telephone
6
from wiring closet to
desktop f (kHz)
z In old installations, loading 1
10 100 1000
coils added to improve
quality in 3 kHz band, but Lower
more attenuation at higher Higher
frequencies attenuation rate attenuation rate
analog telephone for DSL
 Twisted Pair Bit Rates
Table 3.5 Data rates of 24-gauge twisted pair z Twisted pairs can provide
high bit rates at short
distances
Standard Data Rate Distance z Asymmetric Digital
Subscriber Loop (ADSL)
T-1 1.544 Mbps 18,000 feet, 5.5 km z High-speed Internet Access
DS2 6.312 Mbps 12,000 feet, 3.7 km
z Lower 3 kHz for voice
z Upper band for data
1/4 STS-1 12.960 4500 feet, 1.4 km z 64 kbps inbound
Mbps
z 640 kbps outbound
1/2 STS-1 25.920 3000 feet, 0.9 km z Much higher rates possible at
Mbps shorter distances
z Strategy for telephone
STS-1 51.840 1000 feet, 300 m companies is to bring fiber
Mbps close to home & then twisted
pair
z Higher-speed access + video
Ethernet LANs
z Category 3 unshielded twisted pair
(UTP): ordinary telephone wires
z Category 5 UTP: tighter twisting to
z z z z z z
improve signal quality
z Shielded twisted pair (STP): to
minimize interference; costly
z 10BASE-T Ethernet
z 10 Mbps, Baseband, Twisted pair
z Two Cat3 pairs
z Manchester coding, 100 meters
z 100BASE-T4 Fast Ethernet
z 100 Mbps, Baseband, Twisted pair
z Four Cat3 pairs
z Three pairs for one direction at-a-time
z 100/3 Mbps per pair;
z 3B6T line code, 100 meters
z Cat5 & STP provide other options
 Coaxial Cable
Twisted pair
z Cylindrical braided outer 35
conductor surrounds
0.7/2.9 mm
insulated inner wire 30
conductor

Attenuation (dB/km)
z High interference immunity 25 1.2/4.4 mm
z Higher bandwidth than
twisted pair 20
z Hundreds of MHz
15
z Cable TV distribution
z Long distance telephone 10
transmission 2.6/9.5 mm
z Original Ethernet LAN 5
medium
0.1 1.0 10 100
f (MHz)
Cable Modem & TV Spectrum
Downstream Downstream
Upstream

42 MHz

500 MHz

MHz
750
550 MHz
5 MHz

54 MHz

z Cable TV network originally unidirectional

z Cable plant needs upgrade to bidirectional
z 1 analog TV channel is 6 MHz, can support very high data rates
z Cable Modem: shared upstream & downstream
z 5-42 MHz upstream into network; 2 MHz channels; 500 kbps to 4
Mbps
z >550 MHz downstream from network; 6 MHz channels; 36 Mbps
Cable Network Topology

Upstream fiber Fiber Fiber

Head Fiber Fiber
end node node
Downstream fiber
Coaxial
distribution
plant

= Bidirectional
split-band
amplifier
 Optical Fiber
Electrical Modulator Optical fiber Electrical
Receiver
signal signal

Optical
source

z Light sources (lasers, LEDs) generate pulses of light that are

transmitted on optical fiber
z Very long distances (>1000 km)
z Very high speeds (>40 Gbps/wavelength)
z Nearly error-free (BER of 10-15)
z Profound influence on network architecture
z Dominates long distance transmission
z Distance less of a cost factor in communications
z Plentiful bandwidth for new services
Transmission in Optical Fiber
Geometry of optical fiber

Light
Cladding Jacket
Core
Total Internal Reflection in optical fiber

θc

z Very fine glass cylindrical core surrounded by concentric layer of glass

(cladding)
z Core has higher index of refraction than cladding
z Light rays incident at less than critical angle θc is completely reflected
back into the core
Multimode & Single-mode Fiber
Multimode fiber: multiple rays follow different paths

Reflected path

Direct path

Single-mode fiber: only direct path propagates in fiber

z Multimode: Thicker core, shorter reach

z Rays on different paths interfere causing dispersion & limiting bit rate
z Single mode: Very thin core supports only one mode (path)
z More expensive lasers, but achieves very high speeds
Optical Fiber Properties
Advantages Disadvantages
z Very low attenuation z New types of optical signal
impairments & dispersion
z Noise immunity
z Polarization dependence
z Extremely high
z Wavelength dependence
bandwidth
z Limited bend radius
z Security: Very difficult to
z If physical arc of cable
tap without breaking
too high, light lost or
z No corrosion won’t reflect
z More compact & lighter z Will break
than copper wire z Difficult to splice
z Mechanical vibration
becomes signal noise
Very Low Attenuation
100
Water Vapor Absorption
50 (removed in new fiber
designs)

10
5
Loss (dB/km)

1 Infrared absorption
0.5
Rayleigh scattering

0.1
0.05

0.01 Wavelength (μm)

0.8 1.0 1.2 1.4 1.6 1.8

850 nm 1300 nm
Low-cost LEDs Metropolitan Area 1550 nm
LANs Networks Long Distance Networks
“Short Haul” “Long Haul
 Huge Available Bandwidth
z Optical range from λ1 to 100
λ1 +∆λ contains bandwidth 50

v v
B = f1 – f 2 = – 10
λ1 λ1 + ∆λ
5

Loss (dB/km)
v ∆λ / λ1 v ∆λ
= ≈
λ1 1 + ∆λ / λ1 λ 12 1
0.5

z Example: λ1 = 1450 nm 0.1

λ1 +∆λ =1650 nm:

2(108)m/s 200nm 0.8 1.0 1.2 1.4 1.6 1.8
B= ≈ 19 THz
(1450 nm)2
Wavelength-Division Multiplexing
z Different wavelengths carry separate signals
z Multiplex into shared optical fiber
z Each wavelength like a separate circuit
z A single fiber can carry 160 wavelengths, 10 Gbps
per wavelength: 1.6 Tbps!

λ1 λ1

λ2 λ1 λ2 . λ2
λm

optical
optical fiber
λm optical
mux λm
demux
Coarse & Dense WDM
Coarse WDM
z Few wavelengths 4-8
with very wide spacing
z Low-cost, simple

Dense WDM
z Many tightly-packed
wavelengths
z ITU Grid: 0.8 nm
separation for 10Gbps
1540

1550

1560

signals
z 0.4 nm for 2.5 Gbps
Regenerators & Optical Amplifiers
z The maximum span of an optical signal is determined by the
available power & the attenuation:
z Ex. If 30 dB power available,
z then at 1550 nm, optical signal attenuates at 0.25 dB/km,
z so max span = 30 dB/0.25 km/dB = 120 km
z Optical amplifiers amplify optical signal (no equalization, no
regeneration)
z Impairments in optical amplification limit maximum number of
optical amplifiers in a path
z Optical signal must be regenerated when this limit is reached
z Requires optical-to-electrical (O-to-E) signal conversion,
equalization, detection and retransmission (E-to-O)
z Expensive
z Severe problem with WDM systems
 DWDM & Regeneration
z Single signal per fiber requires 1 regenerator per span

R R R R R R R R
Regenerator

z DWDM system carries many signals in one fiber

z At each span, a separate regenerator required per signal
z Very expensive

R R R R

… …
R R
… … …
R
…
R

R R R R

R R R R
DWDM
multiplexer
Optical Amplifiers
z Optical amplifiers can amplify the composite DWDM signal
without demuxing or O-to-E conversion
z Erbium Doped Fiber Amplifiers (EDFAs) boost DWDM signals
within 1530 to 1620 range
z Spans between regeneration points >1000 km
z Number of regenerators can be reduced dramatically
z Dramatic reduction in cost of long-distance communications

R R

… OA … OA … R
OA … OA …
R

R R

Optical R R
amplifier
Radio Transmission
z Radio signals: antenna transmits sinusoidal signal
(“carrier”) that radiates in air/space
z Information embedded in carrier signal using
modulation, e.g. QAM
z Communications without tethering
z Cellular phones, satellite transmissions, Wireless LANs
z Multipath propagation causes fading
z Interference from other users
z Spectrum regulated by national & international
regulatory organizations
Radio Spectrum
Frequency (Hz)

104 105 106 107 108 109 1010 1011 1012

FM radio and TV
Wireless cable
AM radio Cellular
and PCS
Satellite and terrestrial
microwave
LF MF HF VHF UHF SHF EHF
10 103 102 101 1 10-1 10-2 10-3
4
Wavelength (meters)

Omni-directional applications Point-to-Point applications

Examples
Cellular Phone Point-to-Multipoint Systems
z Allocated spectrum z Directional antennas at
z First generation: microwave frequencies
z 800, 900 MHz z High-speed digital

z Initially analog voice

communications between sites
z High-speed Internet Access
z Second generation:
Radio backbone links for rural
z 1800-1900 MHz areas
z Digital voice, messaging Satellite Communications
Wireless LAN z Geostationary satellite @ 36000
z Unlicenced ISM spectrum km above equator
z Industrial, Scientific, Medical z Relays microwave signals from
z 902-928 MHz, 2.400-2.4835 uplink frequency to downlink
GHz, 5.725-5.850 GHz frequency
z IEEE 802.11 LAN standard z Long distance telephone

z 11-54 Mbps z Satellite TV broadcast

 Chapter 3
Digital Transmission
Fundamentals
Error Detection and Correction
Error Control
z Digital transmission systems introduce errors
z Applications require certain reliability level
z Data applications require error-free transfer
z Voice & video applications tolerate some errors
z Error control used when transmission system does
not meet application requirement
z Error control ensures a data stream is transmitted to
a certain level of accuracy despite errors
z Two basic approaches:
z Error detection & retransmission (ARQ)
z Forward error correction (FEC)
Key Idea
z All transmitted data blocks (“codewords”) satisfy a
pattern
z If received block doesn’t satisfy pattern, it is in error
z Redundancy: Only a subset of all possible blocks
can be codewords
z Blindspot: when channel transforms a codeword
into another codeword
All inputs to channel Channel
satisfy pattern or condition output
Deliver user
User Encoder Pattern
Channel information or
information checking
set error alarm
Single Parity Check
z Append an overall parity check to k information bits
Info Bits: b1, b2, b3, …, bk
Check Bit: bk+1= b1+ b2+ b3+ …+ bk modulo 2
Codeword: (b1, b2, b3, …, bk,, bk+!)

z All codewords have even # of 1s

z Receiver checks to see if # of 1s is even
z All error patterns that change an odd # of bits are
detectable
z All even-numbered patterns are undetectable
z Parity bit used in ASCII code
Example of Single Parity Code
z Information (7 bits): (0, 1, 0, 1, 1, 0, 0)
z Parity Bit: b8 = 0 + 1 +0 + 1 +1 + 0 = 1
z Codeword (8 bits): (0, 1, 0, 1, 1, 0, 0, 1)

z If single error in bit 3 : (0, 1, 1, 1, 1, 0, 0, 1)

z # of 1’s =5, odd
z Error detected

z If errors in bits 3 and 5: (0, 1, 1, 1, 0, 0, 0, 1)

z # of 1’s =4, even
z Error not detected
Checkbits & Error Detection

Information bits Received information bits

Recalculate
check bits
k bits
Channel
Calculate
check bits Compare
Sent Received
Information
check check bits
accepted if
bits check bits
match
n – k bits
How good is the single parity
check code?
z Redundancy: Single parity check code adds 1
redundant bit per k information bits:
overhead = 1/(k + 1)
z Coverage: all error patterns with odd # of errors can
be detected
z An error patten is a binary (k + 1)-tuple with 1s where
errors occur and 0’s elsewhere
z Of 2k+1 binary (k + 1)-tuples, ½ are odd, so 50% of error
patterns can be detected
z Is it possible to detect more errors if we add more
check bits?
z Yes, with the right codes
What if bit errors are random?
z Many transmission channels introduce bit errors at random,
independently of each other, and with probability p
z Some error patterns are more probable than others:

p
P[10000000] = p(1 – p)7 = (1 – p)8 and
1–p
P[11000000] = p (1 – p) = (1 – p)
2 6 8 p 2

1–p

z In any worthwhile channel p < 0.5, and so (p/(1 – p) < 1

z It follows that patterns with 1 error are more likely than patterns
with 2 errors and so forth
z What is the probability that an undetectable error pattern
occurs?
Single parity check code with
random bit errors
z Undetectable error pattern if even # of bit errors:
P[error detection failure] = P[undetectable error pattern]
= P[error patterns with even number of 1s]
n 2 n
= p (1 – p)n-2 + p4(1 – p)n-4 + …
2 4

z Example: Evaluate above for n = 32, p = 10-3

32 32
P[undetectable error] = (10-3)2 (1 – 10-3)30 + (10-3)4 (1 – 10-3)28
2 4
≈ 496 (10-6) + 35960 (10-12) ≈ 4.96 (10-4)

z For this example, roughly 1 in 2000 error patterns is

undetectable
What is a good code?
z Many channels have o o
o o
preference for error patterns x x Poor
that have fewer # of errors x x x o o distance
o
x x properties
z These error patterns map o o
transmitted codeword to o o o
nearby n-tuple
z If codewords close to each x = codewords
other then detection failures o = noncodewords
will occur
z Good codes should x
o x
maximize separation o o Good
o o
between codewords x o x o x distance
properties
o o
o o
x o x
Two-Dimensional Parity Check
z More parity bits to improve coverage
z Arrange information as columns
z Add single parity bit to each column
z Add a final “parity” column
z Used in early error control systems
1 0 0 1 0 0
0 1 0 0 0 1
Last column consists
1 0 0 1 0 0 of check bits for each
1 1 0 1 1 0 row
1 0 0 1 1 1
Bottom row consists of
check bit for each column
Error-detecting capability
1 0 0 1 0 0 1 0 0 1 0 0
0 0 0 0 0 1 0 0 0 0 0 1
One error Two
1 0 0 1 0 0 1 0 0 1 0 0 errors
1 1 0 1 1 0 1 0 0 1 1 0
1, 2, or 3 errors
1 0 0 1 1 1 1 0 0 1 1 1
can always be
detected; Not all
patterns >4 errors
1 0 0 1 0 0 1 0 0 1 0 0
can be detected
0 0 0 1 0 1 0 0 0 1 0 1
1 0 0 1 0 0 Three 1 0 0 1 0 0
errors Four errors
1 0 0 1 1 0 1 0 0 0 1 0 (undetectable)

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

Arrows indicate failed check bits

Other Error Detection Codes
z Many applications require very low error rate
z Need codes that detect the vast majority of errors
z Single parity check codes do not detect enough
errors
z Two-dimensional codes require too many check bits
z The following error detecting codes used in practice:
z Internet Check Sums

z CRC Polynomial Codes

Internet Checksum
z Several Internet protocols (e.g. IP, TCP, UDP) use
check bits to detect errors in the IP header (or in the
header and data for TCP/UDP)
z A checksum is calculated for header contents and
included in a special field.
z Checksum recalculated at every router, so algorithm
selected for ease of implementation in software
z Let header consist of L, 16-bit words,
b0, b1, b2, ..., bL-1
z The algorithm appends a 16-bit checksum bL
Checksum Calculation
The checksum bL is calculated as follows:
z Treating each 16-bit word as an integer, find
x = b0 + b1 + b2+ ...+ bL-1 modulo 216-1
z The checksum is then given by:
bL = - x modulo 216-1
Thus, the headers must satisfy the following pattern:
0 = b0 + b1 + b2+ ...+ bL-1 + bL modulo 216-1
z The checksum calculation is carried out in software
using one’s complement arithmetic
Internet Checksum Example
Use Modulo Arithmetic Use Binary Arithmetic
z Assume 4-bit words z Note 16 =1 mod15

z Use mod 24-1 arithmetic z So: 10000 = 0001 mod15

z b0=1100 = 12 z leading bit wraps around

z b1=1010 = 10 b0 + b1 = 1100+1010
z b0+b1=12+10=7 mod15 =10110
z b2 = -7 = 8 mod15
=10000+0110
=0001+0110
z Therefore
=0111
z b2=1000 =7
Take 1s complement
b2 = -0111 =1000
Polynomial Codes
z Polynomials instead of vectors for codewords
z Polynomial arithmetic instead of check sums
z Implemented using shift-register circuits
z Also called cyclic redundancy check (CRC)
codes
z Most data communications standards use
polynomial codes for error detection
z Polynomial codes also basis for powerful
error-correction methods
Binary Polynomial Arithmetic
z Binary vectors map to polynomials
(ik-1 , ik-2 ,…, i2 , i1 , i0) Æ ik-1xk-1 + ik-2xk-2 + … + i2x2 + i1x + i0

Addition:
(x7 + x6 + 1) + (x6 + x5) = x7 + x6 + x6 + x5 + 1
= x7 +(1+1)x6 + x5 + 1
= x7 +x5 + 1 since 1+1=0 mod2

Multiplication:
(x + 1) (x2 + x + 1) = x(x2 + x + 1) + 1(x2 + x + 1)
= x3 + x2 + x) + (x2 + x + 1)
= x3 + 1
Binary Polynomial Division
z Division with Decimal Numbers
34 quotient dividend = quotient x divisor +remainder
35 ) 1222 dividend
105 1222 = 34 x 35 + 32
divisor 17 2
140
32 remainder

x3 + x2 + x = q(x) quotient
z Polynomial
Division x3 + x + 1 ) x6 + x5
x6 + x4 + x3 dividend
divisor
x5 + x4 + x3
x5 + x3 + x2
Note: Degree of r(x) is less than x4 + x2
degree of divisor
x4 + x2 + x
x = r(x) remainder
Polynomial Coding
z Code has binary generating polynomial of degree n–k
g(x) = xn-k + gn-k-1xn-k-1 + … + g2x2 + g1x + 1
z k information bits define polynomial of degree k – 1
i(x) = ik-1xk-1 + ik-2xk-2 + … + i2x2 + i1x + i0
z Find remainder polynomial of at most degree n – k – 1
q(x)
g(x) ) xn-k i(x) xn-ki(x) = q(x)g(x) + r(x)
r(x)
z Define the codeword polynomial of degree n – 1
b(x) = xn-ki(x) + r(x)
n bits k bits n-k bits
Polynomial example: k = 4, n–k = 3
Generator polynomial: g(x)= x3 + x + 1
Information: (1,1,0,0) i(x) = x3 + x2
Encoding: x3i(x) = x6 + x5
x3 + x2 + x
1110
x3 + x + 1 ) x6 + x5 1011 ) 1100000
x6 + x4 + x3 1011
x5 + x4 + x3 1110
x5 + x3 + x2 1011
x4 + x2 1010
x4 + x2 + x 1011
x 010
Transmitted codeword:
b(x) = x6 + x5 + x
b = (1,1,0,0,0,1,0)
The Pattern in Polynomial Coding

z All codewords satisfy the following pattern:

b(x) = xn-ki(x) + r(x) = q(x)g(x) + r(x) + r(x) = q(x)g(x)

z All codewords are a multiple of g(x)!

z Receiver should divide received n-tuple by g(x) and
check if remainder is zero
z If remainder is nonzero, then received n-tuple is not a
codeword
Shift-Register Implementation
1. Accept information bits ik-1,ik-2,…,i2,i1,i0
2. Append n – k zeros to information bits
3. Feed sequence to shift-register circuit that
performs polynomial division
4. After n shifts, the shift register contains the
remainder
 Division Circuit
Encoder for g(x) = x3 + x + 1
g0 = 1 g1 = 1 g3 = 1
0,0,0,i0,i1,i2,i3
+ Reg 0 + Reg 1 Reg 2

Clock Input Reg 0 Reg 1 Reg 2

0 - 0 0 0
1 1 = i3 1 0 0
2 1 = i2 1 1 0
3 0 = i1 0 1 1
4 0 = i0 1 1 1
5 0 1 0 1
6 0 1 0 0
7 0 0 1 0
Check bits: r0 = 0 r1 = 1 r2 = 0
r(x) = x
Undetectable error patterns
(Transmitter) (Receiver)
b(x) + R(x)=b(x)+e(x)

(Channel) e(x) Error polynomial

z e(x) has 1s in error locations & 0s elsewhere

z Receiver divides the received polynomial R(x) by g(x)
z Blindspot: If e(x) is a multiple of g(x), that is, e(x) is a
nonzero codeword, then
R(x) = b(x) + e(x) = q(x)g(x) + q’(x)g(x)
z The set of undetectable error polynomials is the set of
nonzero code polynomials
z Choose the generator polynomial so that selected
error patterns can be detected.
Designing good polynomial codes
z Select generator polynomial so that likely error
patterns are not multiples of g(x)
z Detecting Single Errors
z e(x) = xi for error in location i + 1
z If g(x) has more than 1 term, it cannot divide xi
z Detecting Double Errors
z e(x) = xi + xj = xi(xj-i+1) where j>i
z If g(x) has more than 1 term, it cannot divide xi
z If g(x) is a primitive polynomial, it cannot divide xm+1 for all
m<2n-k-1 (Need to keep codeword length less than 2n-k-1)
z Primitive polynomials can be found by consulting coding
theory books
Designing good polynomial codes
z Detecting Odd Numbers of Errors
z Suppose all codeword polynomials have an even
# of 1s, then all odd numbers of errors can be
detected
z As well, b(x) evaluated at x = 1 is zero because
b(x) has an even number of 1s
z This implies x + 1 must be a factor of all b(x)
z Pick g(x) = (x + 1) p(x) where p(x) is primitive
Standard Generator Polynomials
CRC = cyclic redundancy check

z CRC-8:
= x8 + x2 + x + 1 ATM

z CRC-16:
= x16 + x15 + x2 + 1 Bisync
= (x + 1)(x15 + x + 1)

z CCITT-16:
= x16 + x12 + x5 + 1 HDLC, XMODEM, V.41

z CCITT-32: IEEE 802, DoD, V.42

= x32 + x26 + x23 + x22 + x16 + x12 + x11 + x10 + x8 + x7 + x5 + x4 + x2 + x + 1
Hamming Codes
z Class of error-correcting codes
z Capable of correcting all single-error patterns
z For each m > 2, there is a Hamming code of length
n = 2m – 1 with n – k = m parity check bits
Redundancy

m n = 2m–1 k = n–m m/n

3 7 4 3/7
4 15 11 4/15
5 31 26 5/31
6 63 57 6/63
m = 3 Hamming Code
z Information bits are b1, b2, b3, b4
z Equations for parity checks b5, b6, b7
b5 = b1 + b3 + b4
b6 = b1 + b2 + b4
b7 = + b2 + b3 + b4

z There are 24 = 16 codewords

z (0,0,0,0,0,0,0) is a codeword
Hamming (7,4) code
Information Codeword Weight
b1 b2 b3 b4 b1 b2 b3 b4 b5 b6 b7 w(b)

0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 1 1 1 1 4
0 0 1 0 0 0 1 0 1 0 1 3
0 0 1 1 0 0 1 1 0 1 0 3
0 1 0 0 0 1 0 0 0 1 1 3
0 1 0 1 0 1 0 1 1 0 0 3
0 1 1 0 0 1 1 0 1 1 0 4
0 1 1 1 0 1 1 1 0 0 1 4
1 0 0 0 1 0 0 0 1 1 0 3
1 0 0 1 1 0 0 1 0 0 1 3
1 0 1 0 1 0 1 0 0 1 1 4
1 0 1 1 1 0 1 1 1 0 0 4
1 1 0 0 1 1 0 0 1 0 1 4
1 1 0 1 1 1 0 1 0 1 0 4
1 1 1 0 1 1 1 0 0 0 0 3
1 1 1 1 1 1 1 1 1 1 1 7
Parity Check Equations
z Rearrange parity check equations:
0 = b5 + b5 = b1 + b3 + b4 + b5
0 = b6 + b6 = b1 + b2 + b4 + b6
0 = b7 + b7 = + b2 + b3 + b4 + b7

z In matrix form: b1
b2 z All codewords must
satisfy these
0 = 1011100 b3 equations
0 = 1101010 b4 = H bt = 0 z Note: each nonzero
0 = 0111001 b5 3-tuple appears once
as a column in check
b6
matrix H
b7
Error Detection with Hamming
Code 0
0
1011100 1 1 Single error detected
s=He= 1101010 0 = 0
0111001 0 1
0
0
0
1
1011100 0 0 1 1
s=He= 1101010 0 = 1 + 0 = Double error detected
1
0111001 1 1 0 1
0
0
1
1
1011100 1 1 0 1
s=He= 1101010 0 = 1 + Triple error not
1 + 0 = 0
0 detected
0111001 0 1 1
0
0
Minimum distance of Hamming
Code
z Previous slide shows that undetectable error pattern
must have 3 or more bits
z At least 3 bits must be changed to convert one codeword
into another codeword

Set of n- o Set of n-
o
tuples Distance 3 o tuples
within within
distance 1 o
b1 o o b2 distance 1
of b1 of b2
o o

z Spheres of distance 1 around each codeword do not

overlap
z If a single error occurs, the resulting n-tuple will be in a
unique sphere around the original codeword
General Hamming Codes
z For m > 2, the Hamming code is obtained
through the check matrix H:
z Each nonzero m-tuple appears once as a column
of H
z The resulting code corrects all single errors
z For each value of m, there is a polynomial
code with g(x) of degree m that is equivalent
to a Hamming code and corrects all single
errors
z For m = 3, g(x) = x3+x+1
Error-correction using Hamming
Codes
(Transmitter) b + R (Receiver)

e Error pattern

z The receiver first calculates the syndrome:

s = HR = H (b + e) = Hb + He = He
z If s = 0, then the receiver accepts R as the transmitted
codeword
z If s is nonzero, then an error is detected
z Hamming decoder assumes a single error has occurred
z Each single-bit error pattern has a unique syndrome
z The receiver matches the syndrome to a single-bit error
pattern and corrects the appropriate bit
Performance of Hamming Error-
Correcting Code
z Assume bit errors occur independent of each other
and with probability p

s = H R = He
7p

s=0 s=0
1–3p 3p

No errors in Undetectable Correctable Uncorrectable

transmission errors errors errors
(1–p)7 7p3 7p(1–3p) 21p2
 Chapter 3
Digital Transmission
Fundamentals
RS-232 Asynchronous Data
Transmission
Recommended Standard (RS) 232
z Serial line interface between computer and
modem or similar device
z Data Terminal Equipment (DTE): computer
z Data Communications Equipment (DCE):
modem
z Mechanical and Electrical specification
Pins in RS-232 connector
1 13

(a)
• • • • • • • • • • • • •
• • • • • • • • • • • •

14 25

(b) 1 Protective Ground (PGND) 1

2 Transmit Data (TXD) 2
3 Receive Data (RXD) 3
4 Request to Send (RTS) 4
5 Clear to Send (CTS) 5
DTE 6 Data Set Ready (DSR) 6 DCE
7 Ground (G) 7
8 Carrier Detect (CD) 8
20 Data Terminal Ready (DTR) 20
22 Ring Indicator (RI) 22
Synchronization
z Synchronization of
clocks in transmitters 1 0 1 1 0 1 0 0 1 0 0

and receivers. Data

z clock drift causes a loss
of synchronization T
z Example: assume ‘1’
S Clock
and ‘0’ are represented
by V volts and 0 volts
respectively 1 0 1 1 1 0 0 1 0 0 0

z Correct reception Data

z Incorrect reception due
to incorrect clock (slower T
clock)
S’ Clock
Synchronization (cont’d)
z Incorrect reception (faster clock)
z How to avoid a loss of synchronization?
z Asynchronous transmission
z Synchronous transmission

1 0 1 1 1 0 0 1 0 0 0
Data

S’ Clock
Asynchronous Transmission
z Avoids synchronization loss by specifying a short maximum
length for the bit sequences and resetting the clock in the
beginning of each bit sequence.
z Accuracy of the clock?

Data bits

Line Start Stop

idle 1 2 3 4 5 6 7 8
bit bit

3T/2 T T T T T T T

Receiver samples the bits

Synchronous Transmission
z Sequence contains data + clock information (line coding)
z i.e. Manchester encoding, self-synchronizing codes, is used.
z R transition for R bits per second transmission
z R transition contains a sine wave with R Hz.
z R Hz sine wave is used to synch receiver clock to the
transmitter’s clock using PLL (phase-lock loop)

1 0 0 0 1 1 0 1 0
Voltage

time