Int. J. Contemp. Math. Sciences, Vol. 5, 2010, no.
33, 1627 - 1635
On the Combination of Five Cyclic Code
Manish Gupta
Department of Mathematics D.A.V. College, Bathinda (Pb.), India manish_guptabti@[Link]
J. S. Bhullar
Department of Applied Sciences MIMIT, Malout (Pb.), India
O. P. Vinocha
Department of Applied Sciences Principal, Ferozpur College of Engineering Ferozshah (Pb.), India
Abstract In this paper, we shall construct a long cyclic binary code from small codes. The generator polynomial of the cyclic combination of sub codes is derived and shown to be a simple function of the generator polynomials of the subcodes. We will show that binary cyclic codes of length 5n can be obtained from 5 cyclic codes of length n by two methods. For both the methods the length of codes must be even. We conclude with detailed examples of binary codes of even length and their generators for which exact results are obtained.
Keywords: Cyclic Codes, Generator Polynomial, Product Codes, |u|u + v|, Minimum Distance, Concatenated Codes
1628
M. Gupta, J. S. Bhullar and O. P. Vinocha
I.
Introduction
Concatenated codes have been evolved as a standard for those applications where very high coding gains are needed, such as deep-space applications. Concatenation of codes is the combination of two or more codes so as to obtain a more powerful code. We will discuss the construction of long powerful codes based on the concatenation of simple component codes. The first published concatenated codes were product code by Elias [16]. Concatenated coding schemes have been studied by Forney [7]. They consist of the flow of an inner code and an outer code, which, in Forney's approach, would be a relatively short inner block code, or a convolutional code with maximum-likelihood Viterbi decoding, and a long high-rate nonbinary Reed-Solomon outer code decoded by an algebraic error-correction algorithm. Alternative solutions for concatenation have also been studied in [17], [12], and [18]. Tanner [18] proposed a method for construction of long error-correcting codes from shorter ones based on bipartite graphs and an iterative decoding. The recent research activities has been largely surrounded by parallel concatenated convolutional codes (PCCCs) also known as turbo codes [1][2] and iterative soft decoding. Turbo codes [1] or PCCCs are codes which are two constituent codes. Parallel concatenation was extended to more than two codes by Divsalar and Pollara [3] (see also [10]). These codes were analyzed in [19, 4, 21, 13]. Using convolutional encoders and interleaves, Serially Concatenated Convolutional Codes (SCCCs) yields performances comparable and, in some cases, superior to turbo codes [20]. SERIAL concatenation of interleaved codes, introduced by Forney in [2], has been in use for a while in many diverse applications from deep space to compact discs for the music industry. Cyclic codes have been used in most error-control systems constructed to date. Generally speaking, these systems are of the following two fundamentally different types: those which correct errors by detection and retransmission and those which correct using forward-acting error-correcting codes. Cyclic codes possess significant advantages in both situations. Cyclic codes were first defined by Prange [5]. Since then many families of cyclic codes have been found. These codes are usually studied under the assumption that the code length is relatively prime to the field size. A linear code C is cyclic if it is closed under all cyclic shift c0 c1 c2 cn-1 ck..cn-1 c0 c1 c2.. ck-1 An efficient definition is established by identifying each vector C = (c0 c1 c2 cn-1) with polynomial c(x) = c0 + c1x + c2x2 cn-1xn-1 The fact that C is invariant under a cyclic shift if then expressed as follows: c( x) C xc( x) mod (x n 1) C
On the combination of five cyclic code
1629
Thus the proper content for shifting cyclic codes of length n over Fq is the residue Fq [ x] . Here Rn is principal ideal domain. This means that any ideal in class ring Rn = n x 1 Rn is generated by a single element g in ideal. So there is a unique monic polynomial g(x) of minimum degree in C. This polynomial generates C, i.e. C = g ( x) , and it is
called the generator polynomial for C. The generator polynomial g(x) divides x n 1 . During recent years, a great deal of research has been devoted to the concatenation of cyclic codes. This research was initiated by the invention of turbo codes by Berrou et all [1]. There are several ways to obtain a code from several short codes or old codes, in Section II we had given some of them. In section III we constructed the cyclic code of length 5n from five cyclic codes of length n is two ways.
II. Background
Many good codes have been constructed by modifying or accumulating previously constructed codes. Coset Codes Coset Codes were first studied by Forney [8]. Let C0 be a binary linear [n, k0 ] block
with generator G0 and let C1 C0 be a [ n, k1 ] -sub code of C0 . A coset of C1 is a set
of the form cl + C1 = {cl + c : c C1} . where cl C0 is a coset leader. We will take that non zero coset leaders in C0 \ C1 . C0 \ C1 Forms a factor group, partitioning C0 into 2k0 k1 disjoint subsets each containing 2 k1 code words. Each of these subsets can be represented by a coset leader. The set of coset leaders is called the coset representative space. We denote this coset representative space by [C0 / C1 ]. The code C1 and the set [C0 / C1 ] share only the zero vector in common C1 [C0 / C1 ] = 0 . Every codeword in C0 can be expressed as the sum of a codeword in C1 and a vector in [C0 / C1 ]. We denote this as C0 = C1 [C0 / C1 ] = {u + v : u C1 , v [C0 / C1 ]} The set operand sum is called the direct sum. Forney [9] showed that how codes such as the Reed-Muller and Golay codes can be built up from short codes in this way. | u | u + v | Construction | u | u + v | construction was first given by Plotkin [14], and was rediscovered by Sloane and Whitehead [15]. A construction similar to our construction of the extended Binary Golay Construction is known as the ( u , u + v ) - construction. Let C1
1630
M. Gupta, J. S. Bhullar and O. P. Vinocha
be an [n1, k1, d1] code and C2 an [n2, k2, d2] code C := {(u , u + v ) / u C1 , v C2 } . Then C is a ( 2n, M 1M 2 , d ) code, where d = min {2d1 , d 2 } . To show this we consider two code words ( u1 , v1 + v2 ) and ( u2 , v2 + v2 ) . If v1 = v2 and v1 v2 their distance is at least 2d1. If v1 v2 the distance is w ( v1 v2 ) + w ( v1 v2 + v1 v2 ) which clearly exceeds w ( v1 v2 ) ,i.e. it is at least d 2 . Product codes Product Codes were initially studied by Elias [16]. Let C1 be an [n1, k1, d1] code and C2 an [n2, k2, d2] code. The product code is the [n1n2, k1k2, d1d2] code C whose code words are usually written as an n1 n2 array, C is the set of all such arrays where all rows belong to C1 and all columns to C2. Gluing Codes Gluing has been most effective in constructing self-dual of minimum weight 4. If G be generator matrix, we let {G} denote the code it generates. If C1 is a code of length n1 with generator matrix A and C2 is a code of length n2 with generator matrix B, then a code C of length n1 + n2 is obtained from gluing C1 to C2 or C2 to C1 if C has a A 0 generator matrix G= 0 B D E
where D C1 and E C2 . Tensor Product Codes Let C1 be an [n1, k1, d1] code with parity check matrix H1 = [hij[1]] and C2 an [n2, k2, d2] code with parity check matrix H2 = [hij[2]] then the tensor product is the [n1n2, n1k2 + n2k1 - k1k2, d] code with parity check matrix H = [hij] which is the tensor product of H1 and H2, that is hi1 ( n2 k2 )+i2 , j1n2 + j2 = hi1 , j1[1]hi2 , j2 [2] |a + x|b + x|a + b +x| construction Let C1 be an [n1, k1, d1] code and C2 an [n2, k2, d2] code, The new C will be |a + x|b + x|a + b +x|, a, b C1 , x C2. Here C is a code of length 3n containing k12 k2 code words. Where as lower bound on minimum distance is defined as wt|a + x|b + x|a + b +x| 2wt(a 'or' b) - wt(x) Combination of Repeated Root Cyclic Codes.
On the combination of five cyclic code
1631
All previous studies of cyclic codes assume at the outset that gcd (n,p) = 1 where p is the characteristic of GF(q). This is equivalent to assuming that g(x) has no repeated irreducible factors, as follows from the fact that g(x) divides xn-1 but not its formal derivative nxn-1 unless and only unless the latter is 0, where as Van Lint [11] and Castangoli [6] hdiscarded this property and defined repeated root cyclic codes. Definition: Repeated-root cyclic code Cyclic code for which g(x) has at least one irreducible factor of multiplicity at least 2. The conventional cyclic codes where gcd (n,p) = 1 will, for contrast, be called simple-root cyclic codes. Van Lint [11] constructed a binary cyclic code of length 2n (n odd) from two cyclic codes of length n by introducing the generator g ( x ) = g12 ( x ) g 2 ( x ) and has the dimension 2n r1 r2 . Furthermore Van Lint [11]extended his research by constructing the cyclic code of length 3n and 4n. Cyclic code of length 3n can be obtained from three cyclic codes of n ( q r . p, ( p, q ) = 1) by introducing the generator length
length 4n can be obtained from four cyclic codes of length n = ( 2a 1) with generator matrix g ( x ) = ( x 1) m1 ( x ) . But
3
2 g ( x ) = g13 ( x ) g 2 ( x ) g3 ( x ) and dimension 3n r1 r2 r3 and the cyclic codes of the
here
he
lied
the
condition
that
g1 ( x ) = 1, g 2 ( x ) = ( x 1) , g3 ( x ) = 1, g 4 ( x ) = m1 ( x ) .
III. Combination of five cyclic codes.
A) Cyclic Code where xn 1 has factors more than five Let us consider the factorization of x n 1 into irreducible factor in F2 [ x]
x n 1 = f1 ( x) f 2 ( x).......... fi ( x) . Now take g1 ( x) = f1 ( x).......... f m ( x), m<n<s g 2 ( x) = f m +1 ( x).......... f n ( x), n<s<t g3 ( x) = f n +1 ( x).......... f s ( x), s<t<u g 4 ( x) = f s +1 ( x).......... f t ( x), t<u<i g5 ( x) = f t +1 ( x).......... f u ( x). We had consider five binary cyclic codes Ci (i = 1, 2,3, 4,5) of length n with generator polynomial g1 ( x), g1 ( x) g 2 ( x) , g1 ( x) g 2 ( x) g3 ( x), g1 ( x) g 2 ( x) g3 ( x) g 4 ( x), g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) g5 ( x). and let di be the minimum distance of Ci (i = 1, 2,3, 4,5) . Let r1 = deg (g1(x)), r2 = deg (g1(x)g2(x)), r3 = deg ( g1 ( x) g 2 ( x) g3 ( x) ),
1632
M. Gupta, J. S. Bhullar and O. P. Vinocha
r4 = deg ( g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) ) r5 = deg ( g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) g5 ( x). ) The dimension of Ci be n ri and the minimum distance of C is d1 + d 2 + d3 + d 4 + d5 We are interested in the cyclic code C of length 5n and dimension 4 3 2 5n r1 r2 r3 r4 and g ( x) = g15 ( x) g 2 ( x) g3 ( x) g 4 ( x) g5 ( x) be the generator polynomial of a cyclic code of length 5n. A polynomial a ( x) = a0 + a1 ( x) + ..........an 1 x n 1. This polynomial can be written as
a( x) = ( a0 + a5 x5 + ......) + x(a1 + a6 x5 + .......) + x 2 ( a2 + a7 x5 + .....) + x3 (a3 + a8 x5 + .....) + x 4 (a4 + a9 x 5 + .......)
= a (0) ( x3 ) + xa (1) ( x3 ) + x 2 a (2) ( x3 ) + x3 a (3) ( x3 ) + x 4 a 4 ( x3 ) . We have following representation for polynomial w(x) corresponding to codeword w : w( x) = ( x n 1) 4 a1 ( x) + x n ( x n 1)3 a2 ( x) + x 2 n ( x n 1) 2 a3 ( x) + x3n ( x n 1)a4 ( x) + x 4 n a5 ( x). Example 1 : Take n = 12 12 x 1 = ( x 1)( x 2 + x + 1)( x + 1)( x 2 + 1)( x 2 x + 1)( x 4 x 2 + x 2 + 1) g1 ( x) = ( x 1) , g 2 ( x) = x 2 + x + 1, g3 ( x) = ( x 1)( x + 1), g 4 ( x) = x 2 x + 1,
g5 ( x) = x 2 + 1 C1 has generator g1 ( x) C1 is [12,11,2]-code C2 has generator g1 ( x) g 2 ( x) C2 is [12,9,2]-code C3 has generator g1 ( x) g 2 ( x) g3 ( x) C3 is [12,7,4]-code C4 has generator g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) C4 is [12,5,4]-code C5 has generator g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) g5 ( x) C5 is [12,3,8]-code The code C of length 5n is [60,35,20] code. Where distance d of C can be obtained by adding distance of Ci , i =1,2,3,4,5. 4 3 Now if we take generator of C as g(x) = g15 ( x) g 2 ( x) g3 ( x) g12 ( x) g5 ( x) we will obtain the cyclic code C [60,35,20]-code
B) Cyclic code of length 5n where n is even. We consider factorization of x n 1 into irreducible factors of F2[x]. Here we take g1 ( x ) = 1, g 2 ( x ) = ( x 1), g 3 ( x ) = 1, g 4 ( x ) = ( x + 1) , g5 ( x ) = x 2 1 . Now using the
same procedure as in I we will consider the five different codes by taking the generator polynomials g1 ( x), g1 ( x) g 2 ( x) , g1 ( x) g 2 ( x) g3 ( x), g1 ( x) g 2 ( x) g3 ( x) g 4 ( x), g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) g5 ( x) and let di be the minimum distance of Ci (i = 1, 2,3, 4,5) . Using the same notation as in previous section, we denote the dimension of Ci by n ri and the minimum distance by di. Let g ( x ) = ( x 1)
4
(x
+ 1) be the generator of
2
Cyclic Code of length 5n. The minimum distance of C is 2.
On the combination of five cyclic code
1633
Example 2 Take n = 6. g1 ( x ) = 1, g 2 ( x ) = ( x 1) , g 3 ( x ) = 1, g 4 ( x ) = ( x + 1) , g5 ( x ) = ( x 2 1) C1 has generator g1 ( x ) C2 has generator g1 ( x) g 2 ( x) C3 has generator g1 ( x) g 2 ( x) g3 ( x) C4 has generator g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) C5 has generator g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) g5 ( x) The Code C of length 5n is C is [30, 22, 2] Now if we take generator of C as g ( x ) = ( x 1)
4
C1 is [6,6,1]-code C2 is [6,5,2]-code C3 is [6,5,2]-code C4 is [6,4,2]-code C5 is [6,2,2]-code
2
(x
+ 1) we will obtain the cyclic
2
code C [30, 22, 2]. Example 3. Take n = 8. g1 ( x ) = 1, g 2 ( x ) = x 1, g3 ( x ) = 1, g 4 ( x ) = ( x + 1) , g5 ( x ) = x 2 + 1 C1 has generator g1 ( x ) C2 has generator g1 ( x) g 2 ( x) C3 has generator g1 ( x) g 2 ( x) g3 ( x) C4 has generator g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) C5 has generator g1 ( x) g 2 ( x) g3 ( x) g 4 ( x) g5 ( x) The Code C of length 5n is C is [40,32, 2] Now if we take generator of C as g ( x ) = ( x 1)
4
C1 is [8,8,1]-code C2 is [8,7,2]-code C3 is [8,7,2]-code C4 is [8,6,2]-code C5 is [8,4,2]-code
(x
+ 1) we will obtain the
2
cyclic code C [40,32, 2]. Example 4. Take n = 12. Using the terminology of previous example we take g1 ( x ) = 1, g 2 ( x ) = ( x 1),
g3 ( x ) = 1, g 4 ( x ) = ( x + 1) , g5 ( x ) = ( x 2 + 1) .
Now ingredients are five cyclic codes of length 1, 2, 2, 2, 2. We obtain the cyclic code C [60, 52, 2] The same code is formed if generator polynomial is taken g ( x ) = ( x 1) ( x 2 + 1) .
4 2
The minimum distance is 2
Now as we have seen that by taking generator g ( x ) = ( x 1) ( x 2 + 1) , the dimension
4 2
of the codes and distance is matched with the code formed by five cyclic code of length n.
1634
M. Gupta, J. S. Bhullar and O. P. Vinocha
References
[1] C. Berrou, A. Glavieux, and P. Thitimajshima, Near shannon limit errorcorrecting coding and decoding Turbo-codes, IEEE Int. Conf. Communications, Geneva, Switzerland, (1993), 10641070. [2] D. Divsalar and F. Pollara, Turbo codes for PCS applications, IEEE Int. Conf. Communications, Seattle, WA, (1995), 5459. [3] D. Divsalar and F. Pollara, Turbo Codes for PCS Applications," Proceedings of IEEE ICC'95, Seattle, Washington, June 1995. [4] D. Divsalar, S. Dolinar, R. J. McEliece, and F. Pollara, Transfer Function Bounds on the Performance of Turbo Codes, The Telecommunications and Data Acquisition Progress Report, Jet Propulsion Laboratory, Pasadena, California (1995), 44-55. [Link] report/42-122/[Link] [5] E. Prange , Cyclic error-correcting codes in two symbols, Air Force Cambridge Research Center, Cambridge, MA (1957), 57-103. [6] G. Castagnoli, J. L. Massey, P. A. Shoeller and N. V. Seemann, On Repeated Root Cyclic Codes, IEEE Trans. Inf. Theory, 37 (1991), 337-342. [7] G. D. Forney Jr., Concatenated Codes. Cambridge, MA: MIT Press, 1966. [8] G. D. Forney, Jr., Coset Codes-Part I, IEEE Trans. Inform. Theory, 34 (1988), 1123-1151. [9] G. D. Forney, Jr., Coset Codes-Part II, IEEE Trans. Inform. Theory, 34 (1988), 1152-1187. [10] J. H. Lodge, R. J. Young, P. Hoeher, and J. Hagenauer, Separable MAP `Filters' for the Decoding of Product Codes and Concatenated Codes, Proceedings of ICC'93, Geneva, Switzerland, (1993) 1740-1745. [11] J. H. V. Lint , Repeated-Root Cyclic Codes, IEEE Trans. Inf. Theory, 37 (1991), 343-345. [12] J. Hagenauer and P. Hoeher, Concatenated Viterbi Decoding, Proceedings of Fourth Joint Swedish-Soviet Int. Workshop on Information Theory, Gotland, Sweden, Studenlitteratur, Lund, (1989), 29-33.
On the combination of five cyclic code
1635
[13] L. C. Perez, J. Seghers, and D. J. Costello, Jr., A Distance Spectrum Interpretation of Turbo Codes," IEEE Transactions on Information Theory, 42, (1996), 1698-1709. [14] M. Plotkin, Binary codes with specified minimum distances, IEEE Trans. Inf. Theory, 6 (1960), 445-450. [15] N. J. A. Sloane and D.S. Whitehead, A new family of single-error-correcting codes, IEEE Trans. Inf. Theory, 16 (1970), 717-719. [16] P. Elias, : Error-Free Coding. IRE Trans., 4 (1954), 29-37. [17] R. H. Deng and D. J. Costello, High Rate Concatenated Coding Systems Using Bandwidth Efficient Trellis Inner Codes, IEEE Transactions on Communications, 37 (1989), 420-427. [18] R. M. Tanner, Error-Correcting Coding System, U.S. Patent 4,295,218, October 13, 1981. [19] S. Benedetto and G. Montorsi, \Unveiling Turbo-Codes: Some Results on Parallel Concatenated Coding Schemes," IEEE Transactions on Information Theory, 43 (1996). [20] S. Benedetto, G. Montorsi, D. Divsalar, and F. Pollara, Serial Concatenation of Interleaved Codes: Performance Analysis, Design, and Iterative Decoding, The Telecommunications and Data Acquisition Progress Report 42-126, Jet Propulsion Laboratory, Pasadena, California (1996), 1-26. [Link] report/42-126/[Link] [21] S. Dolinar and D. Divsalar, Weight Distributions for Turbo Codes Using Random and Nonrandom Permutations, The Telecommunications and Data Acquisition Progress Report, Jet Propulsion Laboratory, Pasadena, California, (1995), 56-65. [Link] report/42-122/[Link]
Received: January, 2010
