inearBlockCodes

CapedmainFocus Nocintroduction Taansmonratehchannffffo error

toLinearBlockcode probability
Howtodesign a Error correctingcodes
nodeBlockeachblock are independentof the
100 Block
informationbit IRedudantbitI
befination in the note Isethisdetectand
a n Length ofthecode correctErrors
b k itsdimension q Exampleof error correctingcodes
c nk its redundancy a Repetitioncode
d M q its iardenality Iso a linearBlockcode
Rate Rate
geodecoder134Miggy
e R kin rate

ow Largecan wechoosethe Rate LinearBlock BinaryRepetition o ooo

codes O 7 00
Eachchannelhas a certain capacity depends Efedudan
only on physicalproperty ofchannel SEEdudant bit ht
ate e Examplenow its usefulinadeo
7Encoding candetectsingle Error notcorrectonlydetect

É gn Biassystematicencoding Errorsdetected d 2 I

Fete
mim
communication a Encoding
themat
information
me
and hence Errordetected code
in repetition
Hammingdistance
n
1
Is
b Transmission over a communication Error corrected
channel
c Decodetheinformation 21
hat is a transmissionmedium
Kam
a Binarysymmetricchannel y
g

e
e
g
1EÉE cypertent
IM digital
encoder
mad
I l e
2
this the
chap is subject
b BinaryErasurechannelmainlyused
Ir channel
Basicallythe reverse transmuter
g e
O therearealso
of
e other
channels
Istchap EYE Reduceunnecessary redudanct
E bits
e Huffmancodingetc
informationtheory

g ChannelEncoder Addadditionalredeedanctinto
2 the
l e sourceand later detectand correct error
 to the information and hencedemodulation is y uz
required U
4
L
Harddemolition 43 40 tow thecodedbits
soptDemodulation
we loose a lot of into 42 40 41 are mapped forabove
y not uz cast
ed.gg
e are not Iaton.as
quantising
matron as we ate
the
this mnapkfmagfno.se
yo UgtUz
u u decodingerrors
Bit Error Rate FrameErrorRate Convolutionalcode
apectednumber of intobit in dataFrames s processinformation sequencecontinuosly
info bit decodinger's Error
Alsohas memory n bitencoder
dependson
a Blockades
kinformationbits and also m previousinput
wa correctingcodes blocks l e a convolutionalencoder has a
memory of order mi
Pass
Blockofdata
b convolutional
BlockCodes codes moreinformationfrom convolutionalcode
et K bit information needsto be transm chapter
saythe it's coded or addedredundant
ated Decodingstrategies
bits such that the mainlythese are introductiontopics
thechapter
which I
gothroughagain whenstudying
FEC error correctingcode for oneway
transmission Forward Error correction
lockCode Output AutomaticRepeatrequest ARQ Both dire
onlydepend oncurrentbit
memory lessLess ton notone sided
no Information bitCK

É
R
no ofcodedbits n
In no ofredudantbits Paritybits
untilNow mainlytalkingover introduction and
sample LinearBlockCode hence Fromhere is more important
3 and n 6 hence R 22

Message codewords
O O O O O O O O O codewords
0 22 2 o o giveshow
o 0 I the

jp
0 I 0 2 0 g o g o messages
are
1 I 0 2 10 22 0

I 0 I 2 0 I 2 0 I
0 2 I 0 I 1 0 9 2
I 1 I 0 0 0 2 21
LinearBlockCodes 2 En n 1,23 Binarycheck singleparitycheck
Linear n k d En k blockcodeis a k 4 Generatormatrix isalwayssuchthat
dimensional linearsubspace of the vector space T o o 2
Gspe
I with minimum hamming distance d 0 2 0 I
Length ofcode 0 0 1
a Itsdimension o o o 2
24
2
K redundancy redudantbits o o 0 2
go
qialphabetsing whatis alphgbmetangeuniting
9 qk cardenatelity
as a paritybit
3 Klm Rate
sample
Basically its Citi xn matrix
o o a a
Iq k dimension
I 2 and 22 iii

minimumdistance d Fewimportantemma

de min utca gKeng

b
and d
maid a
R
G I 93 kin
this condition is important 4 2 n 3
for
and 3 2,2
axing a
noiselesschannel

Oconsiderthefollowingcode
GeneratorMatrix
K MateenG C 000007 10209 012223 2210233
hug a iscodelinear
orbinaryLinearBlockCode additionmods
Conditionsto be a linearBlockcode
a y v2 codeword thenwitVa
b Co o o codeword
Bec ÉÉE 221022
Hencenot linear
Shithtsystematic b
generatormatrix mindistance 7
example
n s n q Repetitioncode III dcab III
4Ii
Binary 2 1,23 I
U UG
9 3 pays
mm ee 122027
ML decoder actas
when 9 497 P nearestcodeworddecoding
 Generatormatrix andparitycheck arerelated
and hence
Gt1
G P Ik 3H In K PT
and also downstate PC00008 04 Note herethegeneratormature mustbe in
Hencethat
systematicorder or its useless
ien G 2 0 2 2 2 22
0 I 0 0 0 22 campo given

É
T 0 1 1 01 o o
0 2 2 0 2 0 I k 4n 7 1
I i o o 1 0
n k 3
and hence
a n t
3
R Kn 3
7 Y o 1 0 I 1 I 0

b u 210
v How to usesyndrome error correction andeero
and UG detection
n Ybor Lets y n k Linearcode
Go y un s
2
Y mods
u Tus 2 ee ro eh an rte

v 200 Cyo tea rn a ten D

Vy
V5
2
I 22222 e Ceo e
v t e
en D error pattern

V ryO a's in e representtransmission error

v8 O
2 f ri y

if
e
Generator matrix G fP In
systematic o Cf Mi Yi
coutycheckmateux
he has an error
ForNikianearblockcan bespecifiedbyan
n Syndrome for error detection
S HHT syndrome r
Yo y vn i r codeword then
y
a Ia codeword only if 5 0
nggdonbqdIIII.is n eeaed7 s to then there is an error expections
uundetected Error
gic of the proof Basicallyexpress VI Y but e was 42and
deword in term of messagebits and later saytransmitted
NAT O Hence undetected errorhappens
write equation's in term of v and
only 4 V2
pantytranspose
then the matrix is
0 ye codeword
when e is a validnon zero
notdependontransmutedcogewordand
syndrome does
Verifywhetherits a valid codeword
a S UHT Cut E Ht VAT e t g
 ample consider act 4 anearblockcodewith the all combinations of solutions are
duty checkmature
CHEZ

III
0 1 0 2 2 I 0
map thefts'tandard
simian
Array
9 has 2 possibilities Decoding

et ee Co 1 0000 9
r is
Thesyndromeof a
iii
an n K Linearblockcode
be

S so Si Sa r Ht and from 2k we choosethe most

probable error pattern minizethe hamming
nd HT weight
0 00002
6 8 masamamarachooddecoding all

I 2 O eg
it
5 6 2 2

I
2 I 2
O 2 0 2 2 I 0
syndrome to Hence the code word has
errors and a i o o oo
Error Correction IÉdmeofr
the a Find
eo e
Estimatedcodeword
em
Hence Error has happended
2k possible solutions
to the syndrome
d only 2 iscorrect Chow 247 solution and among that which
ConceptRecalisation has minimum number ofone's
n Bsc the maximumlikelihooddecoderchoose e ee t y
as the codeword I that minizes hamming
eightof error pattern E o o I 0 023
I0 0 I 0 I 1
0 2 o n I I

i o o o 11 is transmit StandardArray Decoding SyndromeDecoding

appose v
and 9 I 0 O O 2 is revered tuples bit and
one Transmitting a codewode n
Bsc received vector has 5 possibilities
syndrome so s sp GH Ci I 1 There are n code a
et e c e o e er ez e4 es e 6 be words and
eachcan be ol's
error pattern
do t est est es
and a partition I possiblecodewords
e t est ext es into 2kdisjointsubset
a nooverlapping
a eaten testes
ggg
 uncorseet v3 ve
et ya o v2 ranbe the at codewords in 000000 022200
v2
120120
Chik anearblock code 202020
I 0.0000 222200 002020 020220
Foam an array as follow's co 20000002200 222020 200220
002000010100 200020 222220
Arrange a'codeword as the topofrow 000200022000 202220 220020
with y o as the sst element 22020
I Suppose I s rows ofarray havebeen
000010022220 202000
22022
000002022202202021
Formed choose a vector ej from y which 000021022222202001 22020
is not in the previousj i rows
hisFormthejth row byadding ej to each uncorsett
v5 V6
codeword y in the top row and placing 000000 022200 202020 2255
egitvi under Yi 20.0000 222200 002020 02022
line continue untill all the vectors from 020000 002200 222010 10022
002000020200 200020 22222
In appear in the array 000200022000 202220 21002
000010022220 202000 22020
41 0 v2 yo Wak 00000202220920202222022
000021022222 202009 22020
ez
ez ez v32 ezthe estrak F
I Basically tell the table
How to pickchoose one entuple
i s au the element's in the standardarraydecoding
are distinct Notepoint
m Caneatv2 CanatVi EanTusk
heftmostentry corsetLeaderandelsecorset

StandardArray S CejtVi HI treat

Thisarray is called standardArray
Syndromeonlydepends
samples on error
6 3 Linear codegeneratedbythefollow pattern
ng mateux Cn a 3 723 817 y 48 hassame syndrome
1 2 0 0
G 19g asyndrometo 2Row so each ecorseleaderhas
10
2 0 2 0 2
0 0 2
0J one syndrome

tandard array
Decoding will be correct if andonly if our
error pattern is a corset
ghqÉÉÉÉI total 8codeword
castename If g correctableerrorpattern
distance b w a and Yi is
42 000000 888
u
00
since u e typ
ca 005
i acee.us wcrtvis wce.tv tri
348,3
65 220001 a 001 L
u 101 g so basicallyweightof errorrector
98196 we 110
coderector
say vj
In dpt
day warty wceitvity
WCe 4s where is Vity
o 9 8 s o s
Since el e tvs are in samecoset o
andsince
wee wee try is as follows o u 42 ro V Vallo4 b
that

ve not us
V2 no tu
summary
steps compute thesyndrome strut steps creation of standardArraycolumn
step2 Find thecosetLeader é whose correctable error patterns Moro corsetleader
expect for two
syndrome is equal to s Hammingwag
3 Decode r into theestimated
codeword
hence 10 a
I a te Detectable error a chatroom
2 weight
o s o e so that entire now element Hamming
all the otherelements in the standard min
the above two
undetectable error pattern
San k eenk
which syndrome corresponds to whichcorset Distance Properties
Leader of a LinearBlockCode
a no
used to perform combination error
Hammingweight ofnon zerocomponents
o of of
correction and error detection
eg u c o o lo i D
o cosets leaders corresponding to the lowest Hammingweight 3
weight error pattern are used for error
correction These are the most
likely D
error pattern Hammingdifference Hammingdistance
b w rand w do co number of places
o to
syndrome corresponding higherweight where
less likely error patterns are used to
declare a
dig I 8go if
der W 2
example consider a 6 3 Linear systematic
codegenerated by c Let w w and x be threebinaryn tap
Then
2 2 2 0 O dew x 3 dad
Go do w
fo g
2
g
0 I
o g gg
D CTraingle inequality
roop traingularInequality Example 2.1 Let K 3 and n o the tablegives
consider a CG 3 uneaseblock code
binary n tuples
message codewords
WCVT
dlr W
MI ya godown
no us us Cvo Vi Va Bivens
dew X WCWTD O O O 0 0 0 0 0 0

WCVTX I 0 0 0 I 2 2 00 3
d VX
as onlytheplacesthey o I 0 I 0 2 0 20 3
dipperhavea
2 I 0 z g o 22 0 4
or anycode a and b
0 0 1 2 I 0 0 0 2 3
Wca w Cb7 w Catbcondition 2 0 2 2 02 4
Hammingcode
0 1 2 0 I 2 0 22
Let a rt w and b wax weget
0 0 0 212 3
WA
WCVTW WcwTX Y WAT WT From the code we can write
W rtx
Mus
no VI 4 42
V4 UI Vg not U2
fd.vn dcw x yda I y Ia
v2 40 41
d The minimum distance d min of G
linear blockcode c is defined as

n
and the hammingweight
The minimum weight inmin of c isdefined minimum
as minimumdistance 3
hammingweight
lamin Emin wer V EC V o
Ao no ofcodeword with namingweigh0 1
s n 2 0
Mote Ag
Az o

as twJ.r.wec.r
Idmm.FI
w
as
At
wagner

min WCutw V WEC V W g g

min W x TEEC X 0

way wtygg
acode In A0 2 and
É Ai 2kif
property ofLbt
Let c be a unear Blockcode withparity E the probability of undetected Error on a
eck mature it If no d g or fewercolumns Bsc is given by
of H add to o the code has minimum
the
weight at d
g Let c be a linear blockcodewithparity
pace pi p IgA
check mature H The minimum weight of C cog s
amen is equal to the fewest number of
m pace
gypsy
p Ifp'd p5ti's
 Bsc p is the crossoverprobability
i p is theprobabilityofsucesstall
ddd
There exist Ch k linearblockcodes for dmin all the n tuples ofweight t comin271
or less can beused as corsetLeader
of standardarray c
Puce I II an p g ng weight tta cannot beused as corset Leader
on a BSC
Dual Code
D The aboveboundshows that the undetec two n u and u are orthogonal it there
ed error can be made to tuples
probability innerproduct u r is zero l e
decrease exponentially with thenumber
of paritycheckbets n k in a linearcode
Ég
U V Cui ri
ForBinarylinearCn 14blockcode
c the
o

4 For a codeword with minimumdistance

cm n ga and code Cd defined as set of
is
min no error pattern with weightdmin 2 all codewords r that are to
or less can change a transmuted codeword all the codewords u EC orthogonal
into another codeword

M Therefore all error pattern with dm 2

n n k
Eddi Cd is also aunearcode
or fewer errors are detectable and dmin 2
s called the random error detecting
capability of a blockcode mature G Then se e ed if and
XGt o
n Theorem
A Block Code c with minimum distancedmin
d Generator mateuse G of a linear n k
s capable of correcting an error patterns
blockcode is the parity check matrix I of
of weight t or less where t is an its dual code and vice
versa
ntegersuch that
It 2
e when c ca then its called selfdualcode
It 2 Edmin code Rate ofself dualcoder y
Codelength is evenCD
Proofixnot necessary K Ma
eg 24,127Golaycodeisself
gg dual
altern of weight that
may not be cornea scalarproduct La b
Éaibi
Ty decoded Ml decoder
by Intutions at ate if at c o tee c
ng Also not necessary ofDual
Hamming dew n s t ta f et cetera C't o
hence

Gangwonincorrectly wheredecode
oh d d
CtLm Kt It
and q
J t
drew 1kt n K

tweight
an defect ly decode
q
or
 condition
a RepetitionCode
Hummingcode
C
parameter n 2 LinearBlockCode
Length n Hamming codes are single error correcting
n times codes
Repeatinformation
iii
eg o Co o 0 m 3 there exist ahammingcodewithfollow
g Ca 2 2 conditions

Generatormature codelength n 2m I
4Ponts informationbits k
G I I 23 21 m s

hmm
Paritybits n k m
Error correcting capability t I dnt
mm
g e.g
and Errors detected n 1
Errors corrected the paritycheck mature inSfmform
nog
H Im PT
b SingleparitycheckCode
where2m m s columns oftoebasicallyPT
it is a linear Ckti k blockcode with consistof an m tuples of
weight more
singleparitybit than s 2 or more
YfarhEEI a n
u Cao a ua n m t3
tuk i K 23 3 2 4 that H p k
p not uit
Evenweightcodewords
to go 2 2 20
4 a u and maggi
Generatormature for spa code in Slmform you

qnyjyq.gg How to use hammingcode to correct

such that
single
Error
Arrange such a way
IT ithinteger

EDIT
Entomumn
H CI i Pty ay Binary
from aboveexample repof
II dimension
802
Minimumdistance 2

can'detected
anyodd error pattern
Let a receivedvector
Note TheCn n D Spccode and Cn 2
repetitioncode are dual toeachother G compute the syndrome tofind
occurred or not
error IET
 If shortenedHamming Code
single
Then it wouldthe column of paritycheck Ip from theparity
matrix
we delete any e columns code we
check mature of a Hamming get
5
01 no error shortened Hamming code with the following
parameter
Kampf 0202020 codeword
010 Code Length 2m I I n
f 43
Hampe a
Ogg Informationbits 2m m l l K
a H Pt
4 Igel Party bits
minimum distance
n k em
dmin 3

1 No ofRow's m are still same

Hamming codeshould be 3 and deleting

ang
p
kaz's same as now will not
maywecan
Butitbe
change the minimum
i ncrease
atleast
previousone
the distance

1 Egg example 7 4 Hammingcode

given by
Hammingcode H 8 0 2 I 2 2
computesyndrometh
and a HI o o 01 0 1
ft0
O
I
O
2
E
remove
0
0
this
0
2
2 2
0 2

YgYgg
column
shortenedCG 3
Hammingcodehasparityched
mature
A d 53

fi
2 2 2
backtoffrey'check 0 2 0 0 2 I
Go mature and see
0 0 0 2 0 2
which column is this
Findminimumdistance
through howmanycolumns
will add to
ITifencepipmeyyzy up o

Hence G o 2 0 I 0 I 0 445ft
ExpurgatedHamming code
Let c be Ch k Hamming codewith parity
nHammingcode one m isfreed all the checkmateux It Let us define a new code4
arameters are also Freed withparity check mateuse Ha Call one rec
eg 81 4 tore as the last row
d Gk
n

f
s
N K m 1
m Since the paritycheck mature ofHamming
de doesn't havean all one rector in any
of the rows any anear combination include
 e last row of it will neveryield a zero
vector
Thus all the rows of it are unearyinde
endent Hence the row spaceof H has
I g
dimension n k t 2 Boundson the Carden
ability andMinimum
Dimension Distance
ofcodesequence
dimcc n n Kt 1 the basic idea is to find the largestcodeof
a given length n and minimum distance d

tence c'gCn ÉÉTs ExpurgatedHamming

code
n
addition ofoddvaluesof v 2
Hence
any oddweightrector v
o

and Hence its a aleeven'Edetord

all odd codewords are expurgated
This make the minimum
hammingdistancet
toy
CExpugartedparitycheckmateux

atendedHammingCode
htt k c
Let c be a Cn k Hammingcode with
arity check mature H Let us define
new code c with parity checkmateux
s all one vector as last row
Rank N K 2
Tus all rows are unearly independant

Eti
it
Ha

paramer Cnt
k

codegiven Hg ExtendedHamming
by code
Same evenweight due to Last
7 dmin 411