Probability α Intformation

if Inin PE
I Iz
Logarithmic tune satisfies this conditions
an event with
di prob Pi
Ii log
avg.info in sourie produling random variable x is ENTROPY

of x
Hint PilogPi bits
PilogePi nats
Hb x Pilog ti
Hixt is nothing but no_of bits required to represent a sourie
41 1
Ex Platlog pint
0 0109101 0
log Pla
Avg Expectation
41 1 70
next
Elogy
05
p N P values.se

It all are equiprobable plai i

I
o HIXI I logl
I I L
pin 011 plat
H1 1 H FIM tested
I
amationifer by any
bijertion
Joint Entropy tonsidering 2 or more

P joint distribution
 y 9
y o i Augentropy Hery 4121 I21

9 1 111

if 1 141 Pt t distribution
then 41 14 Ply H 14 9
HIY Pla HIT A X

06108
Platyl Pla Ply when a y are independent

Y i it may Huta
septal
111412 410914
490914
12 12 1
Hly 82 12909
5
i HIT X i 10.51 0.75

similarity 111 141 0.815 17 410 51

HIXIT HIM HEY I

HIY.tt HIM H 1 171 HIX.tl
HIM HIT I HITI TH IT

Reduction in uncertainty of

am
 Aug 4919
Auguniertainty in after knowing

sat
Mutual information E x 71 41 1 H 14

conditioning redules entropy le x 41 141

then
441 41 are
independent
I x Y 70

HAITI 1114171

I
am_Hiiiii
i e I X Y

ix I
H 141 H Y I xY

07124 7 august
Discrete memoryless Source DMS op is discrete random variable
Current sourie outrome is independent of earlier

4 Iit's re t store 19 9fi's

Source Coding Theorem Average length of a source coding method
is bounded by the entropy of
source

Law His
L
li Ri Pi
l P here I is length req to store x
y Pa Lav Eg Pili HIS
12
L1 Ni FK
we not represent a source with a coding scheme
ran
having avg length less than entropy of source

Let p a 0.125 0.125 0.1 0 15,015

ii 111 1 E Pilogp

flog f flog f o.tlogso.it o.isloglo.it

flog t
0.7427 12 to 7427
83 2
i HIM 1.9927

Eiji.inydins n
1
Decreasing order
7
probabilities for 2 0.125 25 0.5 0 5

22 0.725 Re 0.15
43 0 I se _o.iq 101 0 15

24 0.15 0 21 0.125 i 0.125 7

 5 0.5 1114 0.1

0.5 0.5 01
101 0.275 0.5 11
11 0.225

High prob Lowest source code length

ie for 0 5 it is 0
Lowest prob Highest source code length
ie for 0.1 it is 111
Lav Pili
1 0.5 2 0.15 2 0.125 3 0.125 3 0.1

effiliency 0.995 99.51

1
7100
1.91

Apply Hoffmann source

coding

IE i
E EE
HIS Epilogp 2.12

EPili 01 4
t
i
n 2142 100 96.361

For unique Deludibility Pretin tree land must be satisfied

ie No code word should be prefix of other code word
Above example is also satisfying this land
 2 23 44 as no k 28
is femoryless
0.05 0.15 0.02 0.18 0.22 0.28 0 03 0.07

Find hottmann coding it the code alphabet is

An
find effilienly of it az

Section 9 3 4
may
idspeech
0.28 algorithm

Channel

find out mutual into bin to Rae

y
binary manner
I
Erroneous Results in error 0 1
 1 to 0

I x Y H X H Y

Capacity 1st man I a 4

HIM EI PlatlogP.int
H 14 9 1 P Ri Yr logPlaily
PLAIN
HIX YI
EI ply H LY Ya AB

It
Let H 1 Plait log EI Pth tail

li n.t
IIx.Y1
E PCailYa7logPfpfY

20ᵗʰ Aug
as I x.tl 41 1 H1 1
man of it is capacity

HIXIYI
EI ply H LY Ya

EE Ee Pla a
log.pt
asHIxI
ETP1ai7l09ef1ni
using Baye's theorem
41 7
EI EI Plai Pla di log
 Eg EI Pla 9117909
Finn
i I YI 41 1 H IT

EEPlai.tn log ftp EEPMiiynllogpf.gg

EEPlni.tn 109P1 if
Binary symmetric channel
Po o 9 1
Pp 0 Plan ol Po
P 22 11 1 Po
p
1P p 7 0 12 0 _p
l 12 07
1 p p Y I P

Pty 0 12 11 P
P 9 1 12 1 1
p

PCY oIa 17PM ll

t.pl Y
Ply 11 P Pol 11 pl 11 Po 1 P Po

We know PIY
to get Platyl
Play Platy P 41
p 4111 PIN
Plait PHI PITINI PIN
Pinly P1n
use also pl
i platory of PY O neolith o 1114 o ftp.P

olY ll PlY ilR p Pln

olilr ol ply il PPo I PP
P nil Y o Plt O kill Plaill ply of Ppo PtPo 2PP

P Yet Plt 11 n 1 Pla it p tell It Polli Pl I PP

Let P I Po 9 1 1
I 4 9 109 PP 909
qp Ipp pp qp
PP log t
qp pp 9Pilogefpp.at qp

PP PP
98 9P

if P O e e no uncertainty
change frrorfree
i I x Y H

if p L Full error

i I YI HIM
21stAug
Man of IfxY is achieved for P 0.5 for every p value
Capality is min for P O 5 for every value of P
Gaussian Distribution as pin maximizes the Mutual info given
the power of the symbols
Channel Capacity Theorem

channel

P op
bitsper a
symbol

put

let Rs Rate at which DMS emits the symbols

Rs Rate at which channel can be used

holds then there exists some channel coding

method whith provides transmission
RIEGLE
Pe Probability of error
Differential entropy

5
is ai i tneuous
ii iiiitysun

SI 4uat9 hsinnt pf i hgatat in S a

Fading
Central limit Theorem ICLTI

it t mean that Y
10
Distributed
 Y Y
sum of all
N.E.se Gaussian

For wired channel

Y
Gaussian distri CLT
poise
We can not predict nitt variations
Auto Correlation Function of noise n t nitt Dtl are

uncorrelated Dt extept for Dt 0

i 1 nitt is torrelated only with nett

Autonomelation funition of noise

power speltrum Density
ICAIF FT

Dt
st
ffis istalfomfiaif.ie nsstm8 tnotinAfreiq domain

i Power distribution is uniform in frequency domain freq

ie All frequencies are present in noise uniformly
ii Noise is termed as Additive White Gaussian Noise AWGN

Gaussian tune the exp l

y
Find Differential entropy him that log
as
In
2
him.fi y
e log ftp
ee doe
o

28 9 24

hint are 21
log
 as yet alt hit

Yn n t Mn Digital
Nic N 10 6m

Can an f a doe

n 1 First moment
re 1m doe mean

2ⁿᵈ moment x2 x f a da

Central Moments Ix v7 In wtf a da

second central moment n 2

in v1 In u7 f talk
v ante

a'filalda 12mn fx a doe

fu fxln7dx

1 4 2M F x m

x2 2m m

x2 m i e f x f x

Nii Yp Hic
N on 0

fy 9k Nic tn nk to Yi dm i Power
6m
Aug
txt 11.7 fy Y a 7 101

I NY h Y h Y l

y
Y 0,6m

i PY p 6m
 i I 4 h Yt h 41 1
ane on tpl site on
flog flog
109
III 109 t

Capality 111 man I mill

c
blog.C it signs

I SNR
flog
03 9 24

Sampling rate 2B
C I SNRI 2B
flog
B It bit see
log
PSD
of
f if B
ii 2B NoB
B Double of amp

i C
Blog it bits s

i 66 71000 Gbps

To capacity
increase the we tan not increase the
Bandwidth directly as it causes increment in
Noise Power
There are methods of channel loding to increase bits s
Energy
E
energy of 1 bit
Rb Rate
P Eb Rb

ideal case Rb C malling full use of channel

C B
log 1
437
i t In3iEss
log EF band
n
1 it
2 Bandwidth
I
EI Eg Eg Rf
effiliency
1
II 12

Rb C
Rb C
Rb CC
Eb No
gg

log Efi

4ᵗʰSept
Channel loding Redundant bits are added to into bits
at TRANSMITTER
 Error Detecting and correcting codes
correction
In ffifn.at which error onured

Block codes Convolution lodes

Fix the block code length
n bit block
n 3
0 0

coÉirds o o l collection of codewords is codebook

O l 0 C 110 111 001 010

o l l
I 1 0

1 0 0
valid codeword
1 0 1

I O

l l l

two lodewords resulting again in

Bit by Bit xoring of any
lode is talled Linear Code
valid lodeword then
otherwise non linear lode

Systematic Vs Nonsystematic
msg is part of lodeword
HI PEPSHI
Hard Vs Soft Deloding
Perfe it matthing of bits I TTI
 Linear Block Codes LBC
k bits of uniquely n bits of cw
msg
Ci Cj valid code
i j
In It lode
Representation
Redundant r n k
bits
code rate
m ian be anything
message m Mo Mi Mk
I
No Ri Rn 1 8
go Mo t 9 Mi t 9 Ma t LiangMin i
je lo n it
i MG

Xo 9 Mo t 9,0mi t 920 Ma t Ginno Min

Generator Matrix
Mo Mi My_ Goo go Gone
g
920

gi.io Thin exn

eg am BC

No Mo Ms
R Mo Me
22 M Mr
M Mg

gl 8hm 8En
Given
 i G em l I 0 0 1

mi o o l l l 0

M2 O I I O I I 1

Ms 1 0 0 1 0 I 1
In
Tfa Tmbination 24 16 625 128

4h
IT this two ports are called as DUAL if
Parity Matrix G HT 0

Systematic LBC
In RI code MG
bo b br i Mo my Min 7

i I TIC P I augmented

it G P In 14121
bo b Mom
bob mom
9
F
o 972 4
H for GMT O

1 L
L
 1 4 2 I PT

GHT HGT 0

Repetition lode
13
m

i t
job mo
o o o o

i Cmo

Deloding of 10th
LBC sept 24
weight 1W no of ones
110 1 W 3
w 3
Hamming style W 3

d ti 12 2
d 11 127 W 1 12

m nine in_
 W
message m

channel receives noise

to I olp
inie safe feed
s

191
e.g w
thannel
110 to received
signal
as It is not valid w
i some error is present
11 000 7 13 d y til 1

s are equidistant

equiptiible

Man likelihood deloding

Tx 11 0 1

Rx 7 10 01

i e Y 0 1 0 0

At reseiver you know only y but don't know the

exalt x codeword for it
G is known

y e

YHT el H

syndrome factor SI

one
 he

i e No error in system

Detection
C 11 12 Cm

2 2

2 left
Redundant
14,27LBC 9
e.g
QQ
g
i choose
any
dmin dmin t 1
In How much erronious fits should
be taken
I E't.at L 3
I taken
ie aw L
more

11th sept
Syndrome vector
Transmitter Receiver b m

M message
i
Parithmpupation
syndrome
V
Decoder
b m

bo b Transmit
 message bits k 2

CW n 2n
as n w are used from 2
Isible w

Remaining 2 20

Table
No X X2 22k_ 2 CW

e e x e x roset
e e2 x C 42

Ér
Y C

t
t E I'a Ifab in should
be unused
EI c

i total unused 1W 2
É Cj
Used W 2
Total 1W 2 t 2 Cj
as we know man Total w 2

2 2 2 Cj
2 it E Hamming Bound
 For eath i t radius sphere
IW is out of sphere
13ᵗʰ sept
16.31 lode BC
log
a
3 6
Find
it encoding table
21 Find dmin How many errors lodeword tan detect
How many errors lodeward Ian correct
diagram
it
M Draw hardware syndrome generator
61
suppose I 1 11 00 0 is sent Y 1 11 0017 is
received Show how Ian lode torneit this error

mm x ma
o i o o i o o I I
o I I o l l l 0
1 0 0 I 1 0
1 0 1 0 0 0
I 1 0 1 0 I
I 8 3 I 1 I 0
28 6

Pt I 9 98
1 1
3 6

From dmin 3

Errors lodeword Ian detent dmin I 2

lorrett
dminf.it
 y is 6 bit

Deloding
HT
I o
I
I
I

8 N K s 52153
Foror Pattern syndrome
000000 000 EAT
100000 1 1 0

0 I 1
1 01
1 0 0

1 0

0 0 I

Y Y Y 73 Ye Yt 96

syndryme 6 6 3
 17ᵗʰ sept

1
eg ed getinffneraf.gr polynomial ftp.ullyilic
Find encoding table dmin
 20ᵗʰsept
Convolution lodes
101
111
1,00
110011
itp FF F
111 no

111
I ool

101

initially Flipflops have value O

one bit of message converted to 2 bit of 1W

State diagram
9 90

Q input I hedline

be c

00
 Trellis Diagram
Gandhi sir wala book

Viterbi
Gandhi Sir
14 Relations
Raw uncoded em prob Binary Symm
Capacity
modulator it spectrum shifting is required force coding
Channel Codry LBC

modulator channel demodulator

Febant Communitation

Attenuation Amplifier
Noise Filters
Distortion Equalisers

non linear phase response

III I'abanditin
Ideal response

to
714 e
12
Hitl
th th l K
deviation
any of Hifi zittot linear
distortion
Frau as i t s as
8th oct
Rb N Rs
Rb
log no_of signals i
sec a

Th of signal
Plt l of t T
logic
o logis o

Rx signal
affilit
Pitt With 1
thermal noise
white noise resembles very
well with the
wit 0 white noise

Response of LTI
att hit Plt with I
hitt with 0

faith hits a
pill that with
514 n t

hitltwit
i
n t site nitt 1
i
waveform
nitt 0

It SITI MIT I

nitt to value
 maximize 15 Ttl
Entail
squaring
maximize 1517112

In til s th kt power
9ᵗʰ oct
i
51 1 s f e ᵗᵗdf
but Slt hitt pltl
i 5157 HIF p f

sit H F PIF e t
in game
ECMH sulfidt
9 9 4 sf f noise

system
W He
suit Ng 1415112

Parsevals
theorem
FINCH 1H41 Pdf
as

If
2
H f p f e
151 112
E m HI 1
1H14 dt
 Schware's inequality if F f Falf are complex

functions of f then

FIHEIHd.tl 1114145 IEA 1 dt

Equality holds for Filft KE ft where k is

real constant
t
Let Filt Hit F If pg let
use eq
t

1517112
I 1H41 Idt IP f e 2dg

E n 1 11
1 1H14 dt
8
ma nitude of
SNRoutput PHI If
something
e y
t
Equality gives p 1f e
my us rnogyheyHCf7 52 T
pf e

et of real iii with't

aptitude is egg
hitt Hitlet d
ft
pl the et of
ttl It th

net
I pl ti e of

tag du
 t
i hit pluse H.de
utet2autt tldu
JP
ii hit

i
h t p t t

if we have pitt then impulse response is ihosenb

Matthed Filter matthed with Pitt

S t h t A pit
PIT t pitt

jp u7plt
at
lt
t
ulldu
T

SITI p'luldu E energy of Plt

i SNR 2

5
 Torrelator 11th oct

pitt
Idt sampling
total

TX 51ft
N N
initiinia
filter
NX
C
Rx

s 1H plulp t t tulda silt

of p luldu
correlator is alternative to matthed filter

Multiple ilp
Pilt
P2 t

P.lt Palt orthogonal then more allurate o

p
Proper detection
 Data in MFI
Delision O
P
may
iii MF2
AWGN

P.lt
t.tl
n
ihannel
Decision
op

MF2 Decision 70
AWGN P
Pattl halt Patt t
CDMA
si9nats with spelific specs

gyd.tt
s

message
Initiates te si E th
mi 5
sin with

m s silt n messages then n signals

Energy E s It It For eath signal

silt
Ei Sisditt I
 d lth.at basis functions

setotcthongmal elementary

fdilttdnltldt.to of it at 1 j 1.2 n

Energy normalised to 1
for j K

For large
1 2526
this 255as gnatsepesjnfndbyaig.ge emnPinatilffiiifnts

797 tÉtho
also

Sig silt diktat operation of

29 9,1
i
loeffilient

sampling
fat
Tannel
f ve; torl
Zeleiver m

Response of correlators to
noisy input
R t Silt WCH Of t t

i IT FW GN PSD No 2 W HE
 Sample value at the output of 5th correlator
x
ofxitldi.lt
d Xj
SijtW
T cue
of noise

since with is Tero mean Wi

Expertation
x E Sig Efw
E Xj7 Sij Ma
T

10th oct
xp E a Maj
F x Sij
E w

on E Wi w
Ef tw tld ltidt fwiuldjlald

on 10 14 E w twin dtdu

dict d 1u Sit 4 didu

1
faith dims t 4 at
 on
I tilt at

on

org F For all

after transmission of
Quaternary Scheme many signals

i Constellation Diagram

constellation of the example quaternary scheme

5 511 512
 d If
Tx
ve Sit
message
Mi
or
whannel
i 11213.6
si Ew
Equiprobable

t
Rx
oh Xiesi tw
of dt
da t

of It 2 Sir We

19th oct
Generalised mF
No local generator here

mannel.fm f
from

Silt with
dilt tt
hLY tever in

mn dnlt tt.is
 When you sample of MF
sis w 99m relatied

Line loding Discrete PAM

Pulse Amplitude Modulation

simplex one
way communitation
Half duplex sending receiving only one at a time
Full duplex sending releiving simultaneously
Yesman data rate minimum Pe within bandwidth Power

t
it Eti fi i reseinawaresom
it centretapped transformer repeaters

Differential enloding decoding

iil
bit di bio 1

data bit Prev Encoded bit

when dir I bu 51
dir o bk bin t

odk
obk
I
Encode D Q f
can
 obk dk
I
Delode
F dii bk
Pan

b o i
bii.
i i
di i it
T T
b o to_on
t
dic 0 it t iii
Give same cloth to transmitter receiver

22nd Oct
0 0 o i
unappear A
Tb t
Unipolar A
RZ
Tb 2
a
like A
A
Polar A

BILE
A
Manihester
Tb bit duration

NRL tontinuous zeros present won't give proper o

p
if Aug amp Az when 0,1 equiprobable
then Ian't use transformer coupled lines

Non Return to zero Pulse duration bit duration

RZ long string of t's is avoided useful

Better transition effiliency
But affeits as larger bandwidth

as in Differential
bk dic bin t

1 A bit bini
dis 0 I 0 0 I 0 I l

bic 0 1 1 1 0 0 I 0

Ca O A O O A 0 A A

Bipolar NRZ or AMI alternate marli inversion

Pseudo ter
HDB 3 sulcessive 3 o's are replaced by some
value
B 82s g

To talkle the code violation

23ʳᵈ Oct
Mapping table for Quaternary format
Natural Quaternary Guatemary
binary 88 unipolar Polar
0 0 0 0 0 3A
0 1 0 1 A A
1 0 1 0 2A A
1 l l l 3A 3A

bandwidth
is decreased

M ary formats It is not a

power of 2

Ternary 13 level example

i o 1

Ternary Binary
1 1 000

1 0 0 0 I
I 1 0 I 0
0 1 0 1 1

0 0 Not used
0 1 I 0 0 Rs Rb
1 1 1 0 1
1 0 1 1 0

i ti l l l
How to Mbts to 100 Mb s

UTP unshielded twisted simple

pair
pc to hub
LANS on UTP
100 75 Mbaud spitter
Maps 8,1 1 0
HubTo Pc
Duple

8 binary to 6 ternary innammission

8Th 6 Ts at 25 Mbaud
on each pair
Ts Tb
Rs Rb

25ᵗʰ 0 It

Constellations of discrete PAM signals

Unipolar NRZ
s.lt A o Tb logic
Salt 01 t Tb
o
logic O
E spitt It Atb Eb

d ltl siltl generalised

Fb
d.tt t Tb
ate to

51 it so to sin E
 will mange
Gentia Stimma istbsaygl.ve

Polar NRZ
S ltl A
logi
521 1 A 0
logic

we
wit's fe Fb

Fb Fb
Bipolar NRZ

Eb 8 Feb

Quaternary format

Eb Ee EE 3 Feb
00 01 10 11

Siltl A 1 0

A ol 0 It 2Tb

3A 1 I

3A 00

E A 2Tb
Es 276
 Normalised energy per bit
25 Eb
no of it
gym

5ᵗʰ NOV
6ᵗʰ NOV
We are dislussing error Probability
waterfall lurve
torgup.ie 9wY
10 log Eb No
Pe
logscale Unipolar
polar

Et
I

an.tt neiiiIii

Inter symbol interference

Tse trise Bandwidth

Rx trio for BW as
 limited BW lausing imperfection
187 inbhasfiatharfi.iq strie'ssponse

8ᵗʰ NOV 2024

a t
Eg an ult ktb Tx

an loett depending on dic

plt litbl Rx
Mf
t are
y

up H

Iii
Jen'Eator Decision
way
stampling
tip
hi di

Let pitt be symmetrial around to normalised to

1
plot
To delode ith bit sample yet at t Itb
y litbl M
Eg 9k P itb l Tb

Ma plo arp itb litbl

Y
ufE.jp
Tersymbly interferente
Iii.is Egabiin
ith bit deloding
 but p 07 1

we have to do ISI 0

plits kibl 1
initial
O Ifk 1
The wegodif.tt gndtteq.taimmtaindo79i

Pgltl plt Sylt Eco pints sit mtb

petiodi train
impulse

Pslfl F Psltl Rb P nRb

S Pdf
But.BE
then
Eg pit nRb Tb

We will see some Plfl that are following ear

Let Bo Bo is half of bitrate
RI 21

Tb
Bo
Ideal sold for zero ISI
2Bop 51
I

normalised axes
 n i n o n I

8 t f Htlf Half Hr ft t
v up

Raised losine
spectrum

13ᵗʰ NOV 2024

Eye Diagram
Tapped delay line filter

Delay line

Yiglin

come
early goes after centre tap travel more
in delay line
home late goes before centre tap
 Clock Tb

ADS SR
1 Δ
coett multiplier

Binary adder Δ
Cross talk in UTP tables
NEXT near end moss talk
FEXT Far end tross talk

Time Division Duplex

using Hybrids interference is avoided here
Echo is present it yeath are
ings

Telephode Hybrid sirluitary

This all concepts are applied to Gigabit Ethernet

95mi