* 3.1, Consider a systematic (8, 4) code whose parity-check equations are v9 = My ta + us, Dy My ty + tay Dy Mo ts + us, U3 My + a tay where to, 1, 2, and 3 are message digits and vp, v1, v3, and v3 are parity-check digits. Find the generator and parity-check matrices for this code. Show analytically that the minimum distance of this code is 4, 3.1 The generator and parity-check matrices are: [orstt000] [reev0111] 11100100 01001110 G H 11010010 oo10ii04 From the parity-check matrix we see that each column contains odd number of ones, and no two columns are alike, Thus no two columns sum to zero and any three columns sum to a 4- tuple with odd number of ones. However, the first, the second, the third and the sixth columns sum to zero, Therefore, the minimum distance of the code is 4. 3.6. Consider an (n, k) linear code C whose generator matrix G contains no zero column, Arrange all the code vectors of C as rows of a 2*-by-n array. (@) Show that no column of the array contains only zeros. (b) Show that each column of the array consists of 2*~1 zeros and 24-1 ones. (© Show that the set of all code vectors with zeros in a particular component forms ‘a subspace of C. What is the dimension of this subspace? 3.6 (a) From the given condition on G, we see that, for any digit position, there is a row in G with a nonzero component at that position. This row is a code word in C’. Hence in the code ‘array, each column contains at least one nonzero entry. Therefore no column in the code array contains only zeros.(c) Let Si be the set of code words with a "0" at the ¢-th position. From part (b), we see that So consists of 2"! code words. Let x and y be any two code words in Sy. The sum x + y also has a zero at the /-th location and hence is code word in Sy. Therefore Sp is a subspace of the vector space of all n-tuples over GF(2). Since Sp is a subset of C, itis a subspace of C. The dimension of Sp is k — 1. 4.8. Prove that a linear code is capable of correcting 4 or fewer errors and simultaneously detecting [(!> A) or fewer errors if its minimum distance daia > 4 + 1+ 1. 3.8 From the given condition, we see that \ < [== |. It follows from the theorem 3.5 that all the error patterns of A or fewer errors can be used as coset leaders in a standard array. Hence, they are correctable. In order to show that any error pattern of ¢ or fewer errors is detectable, we need to show that no error pattern x of ¢ or fewer errors can be in the same coset as an error pattem y of A or fewer errors. Suppose that x and y are in the same coset. Then x + y isa nonzero code word. The weight of this code word is w(x +y) S w(x) + wly) SE+A < dinine This is impossible since the minimum weight of the code is dain. Hence x and y are in different cosets, As a result, when x occurs, it will not be mistaken as y. Therefore x is detectable. 3.13, Let C; be an (t;, k) linear systematic code with minimum distance d, and generator matrix G, =[P; Ix]. Let C, be an ('tz,k) linear systematic code with minimum distance d, and generator matrix Gz = (P Ix}. Consider an (1 + m2, k) linear code with the following parity-check matri will ‘Show that this code has minimum distance at least d; + di.3.13 The generator matrix of the code is Gs (Pk Pe hj = [G@ G) Hence a nonzero codeword in C’ is simply a cascade of a nonzero codeword v; in C; and a nonzero codeword v2 in C, (iva) Since w(v,) > dy and w(v2) > dp, hence w{(v1, v2)] > di + da. 3,45. Form a parity-check matrix for a (15, 11) Hamming code. Devise a decoder for this code. 3.15 It follows from Theorem 3.5 that all the vectors of weight ¢ or less can be used as coset leaders. ‘There are (+> 2 ‘such vectors. Since there are 2""* cosets, we must have ‘Taking logarithm on both sides of the above inequality, we obtain the Hamming bound on t, nob atontrs (1) +: +())3.17. Show that the Hamming codes achieve the Hamming bound. 3.17 The number of nonzero vectors of length n and weight d— 1 or less is s X() From the result of problem 3.11, each of these vectors is contained in at most 24° finear systematic codes. Therefore there are at most os ara geimn§(") lincar systematic codes contain nonzero codewords of weight d— 1 or less. The total number ‘of linear systematic codes is N= 2-8) If M < N, there exists at least one code with minimum weight at least d. M < N implies 13 that “a 2h) MY ght) 2 x (") <2 a4 Si (") < 20-9 i i=