—=—=—— Ss Error Control Codin, es 45 Inside this Chapter 16.1, Introduction 16.2. Few Terms Related to Error 16.3. Parity Coding 16.4. Vertical Redundancy Check (VRC) 16.5. Linear Block Code: 16.6. Hamming Codes Encoder of (7, 4) Hamming Code Control Coding «. ~ in 16. 46.8. Syndrome Decoding: Method of Correct Errors aes 46.9. Error Correction using Syndrome Vector in ~ 818 16.10. Syndrome Decoder for (n, k) Block Code 16.11. Other Linear Block Code 16.12. . Cyclic Codes... 16.13. Generator and Parity Check Matrices of Cyclic Codes 16.14. Encoders for Cyclic Codes 16.15. BCH Codes (Bose-Chaudhri-Hocquenghem Codes) 16.16. Convolutional Codes. eee [1612 Introduction As discussed earlier that during the transmission process, the transmitted signal passes throug a noisy channel. Due to noise interference, some errors are introduced in the received dat ese errors can be detected and sometimes corrected using coding techniques. Generally, te error control methods are of two types: @ Error detection with retrans-nission; and (ii) Forward acting error correction, h ‘i i Praae . cA ae when an error is detected, the retransmission request (ARQ) is ending toctuiiquag'at thaan the second method, the errors are detected and corrected by Prom receivers, the forward actine sree nnneg penever a single source transmits data to a mum isienee beim inee ae rection is used. This is due to the fact that the retransmiss™ In this chapter, we shi » We shalll study some si : ity check Some of the coding techniques which denne ret cetecting techniques such as pavitY che Subsequent sections. Coding techniques add area @ pero erters Will also be discussed add some extra bits to the message bits and thes? 798 Scanned with CamScannerat the receiver end. Although these extra bit: . ., rors put the advantage is that error probability reduced, na mnctehe o'the block diagram of the channel encoder and decoder. 4 gh ou Discrete Input Channel shea sossaee encoder Modulate | eee 1? “+ bits Noise Noisy Channel signal 1 fate from channe’ bits ® Digital communication system with channel encoding/decoding operations coaer adds extra bits to the message signal, On the other h somes nder oCtdundant bits and uses them to detect and correct Oeeaa aia ere ahere is any: Let We consider that the ‘k’ message bits are combined in one block bit* gome bits are ‘added by the encoder to this block and total encoded bits become ‘n’. hk) block code. Then we can write, : 20g ge is known as (" syndant DitS ded to the block by encoder = (Bits at the output of encoder) ~ Bits at the of encoder) ue (n-B) redundant bits (Le. code bits) are g technique used. 4 ji Few Terms Related to Error Control Coding nder: Fig. (16.1) related to ‘k’ message bit depending on the of the Codeword ‘Theencoded block of ‘n’ bits is Jmown as a code word. It consi e related terms are described here.as w ists of message bits and redundant Block Length Thewunber of bits ‘n’ after coding is known as the block length of the code. Code Rate The ratio of message bits (k) and the encoder output bits (n) is known as co fs rate is denoted by ‘r’ i.e., de rate. Usually, E (16.2) Che amel data rate is the bit rate/at the output of en, R, then channel “Ifthe bi the bit rate at the input of encoder is Tate will be, Chan el data ah 06.3) = code Vector eos ye : righ 16.2! Code vectom Sepa ee owenting s-bit code se sit code word ean be visualized in on viedimension@l a ae word, Let us visualize Biter Ctr Whose elements oF co-ordinates ie ue aie Spit eode vector There will be It eo de words as it is quite simpler. Fig Scanned with CamScanner800 = COMMUNICATION SYSTEMS = =2', 1 i distinct 8 code words (since number of code words = 24. If we points 09002 and 0, on z-axis, then the following Table 16. dimensional space. (Table 167 Bits of code vector S.No = ez b=y [=e 0 0 0 a 0 0 1 0 1 oO : 0 1 1 i 1 0 oO. . 1 0 1 fi 1 1 0 4 1 1 1 x ing Distance F “ Mth tae distance between the two code vectors is equal to the number Of clemeas i iffer. re beara ae ceennle let X= 101 and Y= 110. Since the two code vectors differ in secon, ad bits, therefore, Hamming distance between X and Y is ‘two’. Hamming distance is denne (X,Y) or simply ‘d’. i.e, (XY) = d=2 id arnt Hence, it may be observed from figure 16.2 that the hamming distance betwe (011) is maximum, i.e. 3. This has been indicated by the vector diagram also, em (100) ag 7. Minimum Distance (d,,;,) Minimum distance (d,,;,) is the smallest Hamming distance between the valid code Veter, S.No._| Name of errors detected /eorrected Distance requirement 1 Detect upto ‘s’ errors per word dpin2 8 +1 2 Correct upto ‘t’ errors per word Gyig 2 2t +1 3 Correct upto ‘t errors and detect 5 > t » Aig 2tts+1 errors per word — For the (n, k) block code, the minimum distance is expressed as 164) Grin S R-R+L: i abevciee 1 8. Code Efficiency : The code efficiency is ex; Pressed as t] for that particular block by itted bits ‘he ratio of message bits in'a block to the transmit the encoder ive,, f Code efficiency = —Message bits in a block } Transmitted bits for the block Scanned with CamScannerince for an (Rf) block code, there are spp ince ill be Message hi 804 por ficiency WY Be bits ang ty» pie el TaNnsmi we . k Mitted hitg ae efficiency = bove e: spwocompare tho abOVE OXPFESION With the ade at ©) gi (16.5) ven in equation (16 2), code efficiency = code rate , we de f the co . (16 Oger of non-zero elements in the transmitted co : enti a(), where Xis the code vector. Av at ye ect will be @(X) = 5. - nb « parity Coding : f coding is a parity cod ; Jest method of is a parity coding. In this met cael «the time of transmission. At the reception en Te bits ‘are detected if the expected pattern of parity bits is not received. iecks these Parity bits, qpmay be noted that parity coding is used only to detect the errors. l. parity may be of two types: . ven Parity * the party ofa binary word is known as ‘even’ if it contains ev ; 0101 has even parity since there are four number of len, ave of's For example, odd Parity The parity of a binary word is known as ‘odd’ 110001 has odd parity because there are three de vector is k : own a: mple, if X= 1110107 ete Meieht, if it contains odd number of 1's. For example, number of 1s in this word, jA!| Vertical Redundancy Check (VRC) tis method, the parity check is done on each character separately. This method is always fo ASCII code. In ASCII code, each character has seven binary digits. Also, one redundant itis added to each character to make the parity odd or even. As an example, let us consider the Table 16:8)\. Vertical redundancy check, even parity ASCII code of the character ‘Additional parity bit by | be | by | by | bs | b6.| Or | be mber of 1s in ASCII 2] fo ft |o} oft | 0 © Been reetore pity biti zero. i IT \dd number of 1s in ASCI vfaifafafolfo fa °F cote, tre pai 10 | 0 | 1 oslo | oe lee Tyojo};ofo|{o [i]? Ny “1. Few Points ink 2 Table 16,3, let us note the following points: ") Asdepicted in the Table 16.2, the message Cm hits b, to sats of total 8 its ie. by 6,6, BS - a the bit be 2 Are used for ASCII code of the character an is used for parity bit. Scanned with CamScanner802 = COMMUNICATION SYSTEMS = (ii) For example, the ASCI code of Kis 1101001. This consists of even number of 1, the 8th parity bit has ‘0’ value. Hence, the transmitted message for K will be peetn, which has even parity. an 1togis ) The ASCII code for 0 is 1111001. This contains odd (i.e, five) number of 1, ; ) parity bit has value of 1. Hence, the transmitted message for 0 is 11110011 HP® ty even parity. , es 7 which At Gv) During the transmission, if single error is created, it will make number of 1 example, let us consider that the received message for character K is as < by by by by bs bg by» by Transmitted messagefor—e 1 1 0 1 0 01% character K by Even Parity(see Table 16.3) Received message —-1101 : 0 10m sale It contains odd number of 1s. It means that there is an error This digit is an error Hence, as shown above, bit b; is in error and it makes number of 1s odd (ic, five). The receiver detects this error. (v) However, it is not possible for the receiver to detect exactly which digit is in error. The receiver will only detect that there are odd number of 1s and thus received message contain, some error. (vi) Thus, the receiver has to send repeat request to the transmitter, to retransmit the mes. sage. dire (vii) If two digits are in error, then number of 1’s in the received message is even. In this cas, receiver will detect no error. In fact, this is the major drawback of Vertical Redundancy Check (VRC). 16.4.2. Longitudinal Redundancy Check (LRC) In the last subarticle, we have observed that VRC checks only in one direction. Hence, oly single errors can be detected. But in LRC, large message block is divided into several characte. Now, parity check is applied in both row and columns. Thus, it is possible to detect upto tripple errors. It is also possible to correct single errors with LRC. In LRC for checking the messts® block, a complete character known as ‘Block Check Character’ (BCC) is added at the end of bet of information. The block check character can be of even or odd parity. An even parity block chet character (BCC) may be obtained by performing an exclusive OR operation on all the Figure 16.3 illustrates the structure of information block transmitted with block check charat#™ ee Start of Many characters Text (one character) Scanned with CamScannerthe figure 16.3, start of \ 803 show ck follows this character, Atte 81%) is one character in Ag; tion smitted Block (ETB) ig transmitted, sifotmation bloc, one cha implies that TRC) is transmited at the eng ut It is alsy ASCII charactor, 4rit indicating te t i character. The Bi blocks as shown in Table 164 which oe x ns consider the ea m - aa LRO) is used .s method (i.e., is used with ASCr TMS age block. The frst row of thie te oan Tl. hm the transmitted jemay be observed that the first charactes transmitted is STX The aramid. is shown in its column. The ASCII value op STX is (615,646, b, 5, i ‘Ascit value of STX p The last row (Tor bg) shows the even parity bite tobe added for ms ei Aa = 1 for STX since ASCII code of STX contain only one the igh tataCter. Hence, by transmitted character for STX is (byby...bg) = (01000001), ae eerd 1. Hence, the transmitted character according to figuse 14 sree s shown in Table 16.4 (a), the sequence : eat STX. This is known ee atte ae N,A\1,T.....ete., is transmitted ) In Table 16.4(a) it may be noted that the second last chi laracter transmitted is ETB (End of Transmitted Block). The ASCII value of ETR (bysbyy..b;) = (1110100) is shown : its column. The parity bit bg is zero. Table 16.4; LRC Using Block Check Character Characters STX'N A IT 1 8/0 K ETB! BCC rare ohn) n bits ; ‘Fig 16.4.) Functional block diagram of a block coder In this section, we shall consider binary codes ie., in which all tranamitted digits are bi Scanned with CamScanner ted. 6.+ of any two code vectors produces anothe code vector. r combination of other code vectors. the sum’ ay be expressed as a lineat termination of all the Codewords for a Linear Block Code ce : : at the particular code vector consists of m,, my, mg, ... m, message bits ta Then this code vector may be written as under: ¢, check bi oe X= (amyy May + My Cy» Car CQ) as tinent’ it eode vector ™ q= are the number of re ye written as, x =(MIC) M= k-bit message vector and C = q-bit check vector he role of error detection and correction. The function of the ‘te these ‘check bits’. The code vector can be represented as, (16.10) dundant bits added by the encoder. The above ns that 7 _ Js be rector may # a (16.9) eck bits play is to general X=MG X = Code vector of 1x size or n bits ‘M = Message vector of 1k size or k bits G = Generator matrix of kx n size. ation (16.10) represents matrix form i.e., (16.11) [Xan = MD tx (Gliexn yr matrix depends upon the linear blo: e, the chi rear plock code i 1d fence, eau e generator ck code used. Normally, it is repre- ynted as, G= [nl Pxahecn (16.12) re I, = kx k identity matrix, and P =k xq submatrix Now, the check vector may be obtained as, Cc =MP (16.18) , in the expanded form, the above expression can be written as Pa Pa Par Poa =+(16.14) Scanned with CamScanner806 = COMMUNICATION SYSTEMS = Example 16.1. The generator matrix for a (6, words of this code. ay Solution: We know that generation matrix is given by G= [ht Pxaleen Given generator matrix is 100 G=|010 G ne 1 0 1 oO: Oo: 1: oo1 100 I, = Igx3 =|0 1 0 001 oid and Pyg = Paxg=|1 0 1 110 Here k= 3, q=3 andn=6. This means that the block size of the message vector is 3 bits. Thus, possible message vectors as shown in the Table 16.5. 3) block code is shown below, 0 Obtain ay N “dey “fi there will be total g Sr. No. ms 1 0 2 1 3 0 4 1 5 0 6 1 7 0 8 1 The P submatrix is given in the example which is reproduced here i.e. 0 =]. 1m o1 41° 0. For the check bit vector, there would be three bits. These may be obtained (16.13) as under: a C= MP according tooo? Scanned with CamScanner5 ERROR CONTROL CODING # 807 11 01 1% ore . 2 [mma ms] : texeaea) Im xc multiplication, we get matrix mul on the aoe © (gx m3) Om) © (ms) cy = om) © (0X mg) @ (ms) 22 cn) omg (0% MD ions, we get spree eatatiOD5> ast ey = OM m, @ Mg os m,@ my ations give check bits for each block of my, my m message bits. 6 equa Or plock of (rm, Mz, ms) = 000, we have ¢, 6 firs angle OF BN = 00=0 c, = 0@0=0 c, = 0@0=0 ,¢) 2000 eo of (my, May m,) = 001, we have co, = 0@1=1 c= 0@1=1 5 = 0@0=0 (cys €q) G3) = 110 qable 16,6 shows all the message bits, their check bits and code vectors calculated as ‘Table IGG! Code Vectors of (6, 3) Block Code. Bil of message ‘vector in Check bits Complete code vector or block Complete codeword qm = oy = My |_m, | mm, m,® my m,@m,|m, | my | mg Cc, Cy Cy BE} 0 0 o fo [o [ojo jo] o OH af 1 one 2s M0 Oa te oe eee ; : 1 0 1 Ot i Jet ]a0e [Et Ifo o 1 1 Ce Ee 1 pets g 1 Z 1 .[tor] o-f os] a ea in _ o 1 aero. s} 1: | a|fo 1 Tot 1 0 me [uo [ate | = ——__° 0 0 T2121 [° [o |] ° Cent of tter i ee ‘ity Check Matrix (H) for Linear Block Code as, fc Under,“ fF each block code, there is a q x n parity check matrix (H). This is Scanned with CamScannerH= [Prt], ot PTis the transpose of P submatix. The P submatrix is defined 7 Pr Ba Ba Py Pa Ps Pa Pa Big Here, OnE O35 Pa Fa As Paley Pay Poy Pa oe Fis Pag, Pag Ps Pas Pog f = e Le a0 Se Fa Pa By om Fadi 4 i coat A With the above equation, we can write equation (16.15) as under: Ril is acne 16.6. Hamming Codes Hamming codes are defined as the (n, These codes satisfy the following © Number of check bits q > 3 (i) Block length n = 20_ (i) Number of message bits & = ng Gv) Minimum distance d,.,, = 3 We know that the ¥) linear block codes. conditions; in code rate is expressed as, k Fi +1 101. (H) xtzieM dand), *10NY Loboa facta ae Scanned with CamScannere value of n = 24-1 in last equation, we get (16,20) } serve that r= 1ifq>>1. we may ob: yiscauator and Correction Capabilities of Hamming Codes ro (or Pete gum distance (din) of Hamming code is 3, it can be used to detect d 50° Peet dieteyis can also be obtained form the generalized Table 16.2. 4 errors => dyyig 2 24 LIC: dyin 2 8 im | sess a8 ‘upto one (1) errors =? pin 2 21) +1 be. dyin > 3 ‘i a arity enock matrix of a particular (7, 4) linear blok code is expressed as \ 62-7 1110100 sph Ca me|proro le ; an 170. 2 10h old enerator matrix (G) 8 distance between code vector? ibe detected? How many errors can be corrected? G7 To many errors Can ga How ven thatn=7and k= 4 tio! Fr check bits aren — = 7-4ie.g=3 amber of nem-1=2-157 ath «ates that the given code is a Hamming code. The parity check matrixis of qn size Lariat by equation (16.18). It can be written as (with q= 3 andn=7andk=4) dis 8 Ai. Pa Pat Put 09 (Hlgxa = [Piz Pee Pao. in Parnes 02 ALO. (ii) Pa Pas Psp Pa 0-0 7 e (ga = [P45] id) arity check matrices of equations (Hf) and (iii), we have Pa Poa Po Pa - pt=| Po Pao Poo Pa ; Ps Pos Pas Pas Nor, according to equations (i) and (iv), we have V1, 1-0 pr=|1 101 f ; ‘ , 1011 lence, the P submatrix may be obtained as, Pi Pe As Py Pe Pes| 2 Py Bo Pes Pa Pa Pas ‘On comparing P: wei) (0) orre nore heer Scanned with CamScannerFurther, we know that generator matrix G is expressed as under; =™ G= [hiPoaken nd n = 7, the above equation becomes, @= [hiPoalier Substituting the identity matrix I, of size 4 * 4 and parity submatrix P,,, of size i. in equation (v), in above equation, we get obta; 1000:111 0100:110 qg=|0010:1 o1 0001:011 —— Tyg Pars tax ‘This is the required generator matrix. To find all the Codewords ‘The check bits can be obtained using equation (16.13), i.e., c= MP In the more general form we can use equation (16.14) ic., (with q = 3, and k= 4), we have with k=4,q= 3a (vi) = [mama mace Plans eis as hae) Feere)= mmm mall 5 4 O21 aris [ey ¢2 eslixs Solving the above equation with mod-2 addition, we obtain, ¢, = (1X m,) @ (1x m,) @ (1X mg) @ (0X my) = (1% m,) @ (1 X m,) ® (0X ms) @ (1x my) and cg = (1% m,) ® (0X ms) @ (1X mg) @ (1 @ m,) Thus the above equation may be written as, ¢, = m, Om, Om, ©, sn(vii) cy = m, @ my @ my and ¢y = m, © my @ my For example, if (m, m m3 m4) =1011, we obtain, ce, = 1@0@1=0 ¢) = 1@0@1=0 at ¢3=1@1@1=1 bo he or wd teen gE mM (2m mz mg mg e, ¢9 ¢3) = (101 1:00) ; secasine.the same procedure as given, we may obtain the other eode sords or code veto ‘able 1e code vectors (code words). This table also lists the weight of each code wor® Scanned with CamScannerCode acer onde word Weight of code vector cfm Ma Ma Me 1 CaO oR) o}o o o oO Oo OO 0 il CE ee ee 7 7 Lo enle Oa 8D 3 Do en ae a 28 4 Drops ove OF toto 3 rr Ore ae ee em oc 4 Tle ma eo ot 4 rage eee 3 cla oO 0f-ed SF 4 Tee 3 Pi. 0 ak ee ous daa 3 fae ee ke A > 4 a 1 1 0 0 0 0 1 3 0 i 1 0 7 0 1 0 4 rarer ee 4 1 1 1 1 J: 1 1 1 7 hh their weights. The smallest weight of ay Hence we can write ‘The minimum distance of h TheTable 16.7 lists 2* on-zero code vector is =24= 16 code vectors along wit! 2 We know that the minimum distance is nin linear block code is equal to th tdevector. This means that d, (i Minimum distance betwee d, (iv) Now, since ia Fe [w @nin ¥= (0 0...0) zn code words is Weknow that ms dain 2 81 ie ie 32 st1 ence ss2 two errors will be ae e detected. ® pig 2 241 e 32 2+1 ts. Th “tefore one error will be corrected. ‘¢ minimum weigth of any non-zero Scanned with CamScannerHence, with the help of Hamming code (d,,;, = 3), two errors can be dete can be corrected by its property. [16:7:) Encoder of (7, 4) Hamming Code Figure 16.5 shows the encoder of (7, 4) Hamming code. This encoder is implemented fo, f matrix of the example 16.2. The lower register contains check bits c,,e,, ands Thee tty obtained from the message bits by mod-2 additions. These additions are perfonsiegnet® ite at nd sin leg To equation (vi). The mod-2 addition operation is nothing but exclusive-OR operatigg *ding i ~ Input bit sequence ‘Code words output Check bits register i Encoder for (7, 4) Hamming code or (7, 4) linear block code for example 16.2, The switch ‘S' is connected to message register first and all message bits are transmitted, This switch is then connected to the check bit register and check bits are transmitted, This forms a block of ‘7’ bits. After this input bits are taken for next block. : 3 116.8. - Syndrome Decoding: Method to Corect Errors In this article, we shall discuss the method to correct errors in linear block. coding. Let the transi code vector be “X’ and corresponding received code vector be represented by ‘Y’. Then, we can write, _ si X=Y- if there are ho transmission errors : and X+Y__ ifthere are errors produced during transmission ‘The decoder detects or corrects these errors in Y by using the stored bit pattern in the decoder about the code. For larger block lengths, more and more bits are needed to be stored in the decoder. This increases the memory requirement and adds to the complexity and cost of the system. ‘Therefore, to avoid these problems, synchrome decoding is used in linear block codes: This method is shown in the paragraphs to follow: () We know that with every (n, 2) linear block code, there exists a parity check matrix () of size q x n. It is defined as, mitted A= PTI), The transpose of the above matrix may be obtained by interchanging the rows and the columns. Thus, we have P HT = Wax q a Here P is the submatrix of size & x q and I, is the identity matrix of size q xq. We B® defined P submatrix earlier. 2 Scanned with CamScannerim ERROR CONTROL CODING = heck matrix (H®) has very important property as under: (16,21) (16.22) arity check matrix and code vectors obtained in _for al us consider the P' 1 sere : : Le et jcke matrix is given by equation (d. The transpose of this matrix can 1 110 101 qra|o1l 100 (16.23) 010 0 0 Lis (retanda=9) able 16.7 ists all the code vectors for this parity check matrix. Let us consider the third qa TP his table ie ier X=(0010101) (16.24) soe et 8 eY the property of equation (16.21) ie. £24 110 101 XH = [0010101], <7] 1 1 (16.28) 10.0 01.0 0,0; Ipggnocicw ( On solving the above two matrices with the rules of mod-2 addition (Exclusive - OR operation), -wecbtain, Hie g09160810000080808 080908008001 209 000000.) =(0 0 0) ‘This proves the property. # fe vectors also. Hence, X belongs to the valid code vector temals be proved for other cod eer end, At the receiver en‘ Ten, mean writ, Pai rare seOsikk= Vie, no erat or Yis valid coa ae eee errors. res, when some errors are present in receive i is non-zero, will not satisfy the property of equation (16.21). rome in Y. The non-zero output of tl in Y, Syndrome is represe’ .d the received code is n e vector .d vector Y, then it will not be from valid ‘This shows that whenever she product YHT is called nted by ‘S’ and may be represent error in the output, When all el Scanned with CamScanner814 ay i tant No error in the outpul o Vis ‘some other valid code vec! al re undetectable. it ector E. Let, this vector repres et us consider an r-bit error ent the posi n example, in Y. Asal =(1011 0 x=( tor other than X. This means that the transmis, (ix) Now, consider, be a transmitted vector transmission errors 1 0.01 2, . ‘be a received yorim to and E=(0010 1) _ represents the error vector Then : in. coro entries represent enrard in T+ easing the mod-2 addition rules, we can write: X@E y=(1@0 000 101 100 061) (162g - Bit by bit mod-2 addition or y=a001)0 write X=YeE ae (100000000110 01 1) or X= (10110) or we can write, X=Y@E (@0 060 0@1 160 181) =(10110) ation (16.26), we know that syndrome vector i $= YH" y= X @ E from equation (16.28) in this equation, we get (xo E) HT XH? @ EH (16.21) we know that XH? = 0, then above equation will be 6.29) mdrome depends on the error pattern only. It does not “g’ is of size 1 x q. Hence, q bits of syndrome ticular error or X From equ: s expressed as, Substituting the value of s From the property of equation S= EH or (zi) This relation indicates that sy) depend upon a particular message. Syndrome vector can only represent 24 syndrome vectors. Each syndrome vector corresponds to a pa pattern. Example 16.3, The parity check matrix of a (7, 4) Hamming code is expressed as under: Leer oer od. H=j]O111 010 1101:00 1);,, Evaluate the syndrome vector for single bit errors. Solution: This is a (7, 4) linear block code: This means n= and k= 4 We k q=n-k=3 ‘aten tsa eyndoeme vector is a q bit vector. For this example, syndrom ; Teoria ail bo cae be 2°—1= 7 non-zero ayndromes. ‘This shows that ‘7 single bit deceit presented by these ‘7’ non-zero syndromes. Error vector E is a” bit vert singe errr ptiern in a7 For this example, £ is ‘7 bit vector. Following Table 168 shows a a shown). it ertor vector (Note that only single bit error patterns have Je will be a 3H Scanned with CamScanner