ArticlePDF Available

Abstract

Se definen nuevos códigos convolucionales (MDS) generalizando los códigos Goppa. A new kind of Convolutional Codes generalizing Goppa Codes is proposed. This provides a systematic method for constructing convolutional codes with prefixed properties. In particular, examples of Maximum-Distance Separable (MDS) convolutional codes are obtained.
arXiv:math/0310148v1 [math.OC] 10 Oct 2003
CONVOLUTIONAL CODES OF GOPPA TYPE
J.A DOM´
INGUEZ P´
EREZ, J.M MU ˜
NOZ PORRAS, AND G. SERRANO SOTELO
Abstract. A new kind of Convolutional Codes generalizing Goppa Codes is pro-
posed. This provides a systematic method for constructing convolutional codes with
prefixed properties. In particular, examples of Maximum-Distance Separable (MDS)
convolutional codes are obtained.
1. Introduction
The aim of this paper is to propose a definition of Convolutional Goppa Codes
(CGC). This definition will provide an algebraic method for constructing Convolutional
Codes with prescribed invariants.
We propose a definition of CGC in terms of families of curves XA1parametrized
by the affine line A1= Spec Fq[z] over a finite field Fq. In this setting, the usual
definition of a Goppa Code as the code obtained by evaluation of sections at several
rational points, is translated as a code obtained by evaluation (of sections of some
invertible sheaf over X) along several sections of the fibration XA1.
The paper is organized as follows.
In §2 we offer a summary on Goppa Codes following [3], [6], and using the standard
notations of Algebraic Geometry [2].
§3 is devoted to giving the general definition of CGC and gives some general results.
In §4 we study the case of a trivial fibration of projective lines over A1and we
conclude giving some explicit examples of MDS convolutional codes.
We freely use the standard notations of abstract Algebraic Geometry as can be found
in [2]. After the works of V. Lomadze [4], J. Rosenthal and R. Smarandache [7], [8],
there is evidence that the use of methods of Algebraic Geometry can be relevant to the
study of Convolutional Codes. This paper is a step in favor of that evidence.
2. Background on Algebraic Geometry and Goppa Codes
In this Section we summarize the basic definitions about Goppa Codes, constructed
using methods of Algebraic Geometry (see [3], [6]).
Let Xbe a geometrically irreducible, smooth and projective curve over the finite
field Fq. Let p1,...,pnbe ndifferent Fq-rational points of X, and Dthe divisor
D=p1+···+pn. Let Gbe another effective divisor with support disjoint from D.
The Goppa code C(G, D) defined by (G, D) is the linear code of length nover Fq
defined as the image of the linear map
α:L(G)Fn
q
f7→ (f(p1),...,f(pn)) ,
Key words and phrases. Convolutional Codes, Goppa Codes, MDS Codes, Algebraic Curves, Co-
herent Sheaves, Finite Fields.
This research was partially supported by the Spanish DGESYC through research project BMF2000-
1327 and by the “Junta de Castilla y Le´on” through research projects SA009/01 and SA032/02.
1
2 J.A DOM´
INGUEZ P´
EREZ, J.M MU ˜
NOZ PORRAS, AND G. SERRANO SOTELO
where L(G) is the complete linear series defined by G. That is, let Fq(X) be the field
of rational functions over the curve X,
L(G) = {fFq(X) such that Div(f) + G0}.
The Goppa code has dimension
k= dim C(G, D) = dim L(G)dim L(GD).
Let gbe the genus of X; if we assume the inequality 2g2<deg(G)< n, then one
has
k= deg(G)g+ 1 ,
and the minimum distance dof C(G, D) satisfies the inequality
dndeg(G).
Let OX(D) be the invertible sheaf on Xdefined by the divisor D. One has the following
exact sequence of sheaves
0 OX(D) OX OD0,
where OD Op1/mp1× · · · × Opn/mpnFq×n)
...×Fq. Tensoring the above exact
sequence by OX(G), one obtains
0 OX(GD) OX(G) OD0.
By taking global sections, we obtain an exact sequence of cohomology
0H0(X, OX(GD)) H0(X, OX(G)) α
ODH1(X, OX(GD))
H1(X, OX(G)) 0,
where L(G) = H0(X, OX(G)) and αis the evaluation map defined above.
In the case 2g2<deg(G)< n, one has the exact sequence
(2.1) 0 H0(X, OX(G)) α
ODH1(X, OX(GD)) 0.
Let ωXbe the dualizing sheaf of X, which is isomorphic to the sheaf of regular
1-forms over X;H0(X, ωX) is the Fq-vector space of global regular 1-forms over X,
which is of dimension g= genus of X.
By Serre’s duality ([2]), there exist canonical isomorphisms of Fq-vector spaces
H1(X, L)H0(X, ωX L1)
for every invertible sheaf Lon X. Given a divisor Dover X, we shall denote by Ω(D)
the vector space H0(X, ωX OX(D)).
The dual Goppa code, C(G, D), associated with the Goppa code C(G, D) is defined
as the linear code of length nover Fqgiven by the image of the linear map
α: Ω(GD)Fn
q
η7→ (Resp1(η),...,Respn(η)) ,
Let us take duals in the exact sequence (2.1):
0H1(X, OX(GD))β
O
D
αt
H0(X, OX(G))0.
By Serre’s duality, one has isomorphisms
H1(X, OX(GD))Ω(GD),
H0(X, OX(G))H1(X, ωX OX(G)) ,
CONVOLUTIONAL CODES OF GOPPA TYPE 3
and the above sequence is the cohomology sequence induced by the exact sequence of
sheaves
0ωX(G)ωX(DG)ωX(DG)OXOD0,
where we denote ωX(G) = ωX OX(G), and βis precisely the map αdefining
C(G, D).
Given a linear series Γ H0(X, OX(G)), that is, a vector subspace defining a family
of divisors linearly equivalent to G, we define the Goppa code C, D) associated whit
Γ and Das the image of the homomorphism α|Γ:
H0(X, OX(G)) α//OD
S|
Γ
α|Γ
77
p
p
p
p
p
p
p
p
p
p
p
p
p
p
When Γ &H0(X, OX(G)), we shall say that C, D) is a non-complete Goppa code.
3. Convolutional Goppa Codes
We shall contruct a kind of convolutional code that generalizes the notion of Goppa
codes. These codes will be associated with families of algebraic curves.
Given an algebraic variety Sover Fq, a family of projective algebraic curves parametrized
by Sis a morphism of algebraic varieties π:XS, such that πis a projective and flat
morphism whose fibres Xs=π1(s) are smooth and geometrically irreducible curves
over Fq(s) (the residue field of sS).
Let us consider a family of curves Xπ
Uparametrized by U= Spec Fq[z] = A1.
Given a closed point uUwith residue field Fq(u), the fibre Xu=π1(u) is a curve
over the finite field Fq(u).
Let pi, 1 in, be ndifferent sections, pi:UX, of the projection π. These
sections define a Cartier divisor on X:
D=p1(U) + ···+pn(U),
which is flat of degree nover the base U([2]).
Note that given a coherent sheaf Fon X, the cohomology groups Hi(X, F) are finite
Fq[z]-modules and Hi(X, F) = 0 for i0 (see [2] III).
Let Lbe an invertible sheaf over X. One has an exact sequence of sheaves on X
(3.1) 0 L(D) L OD0,
which induces a long exact cohomology sequence
(3.2)
0H0(X, L(D)) H0(X, L)α
H0(X, OD)H1(X, L(D)) H1(X, L)0.
Let rbe the degree of Lin each fibre of π(which is independent of the fibre) and
let gbe the genus of any fibre of π(also independent of the fibres).
Proposition 3.1. Let us assume that 2g2< r. Then, one has that H1(X, L) = 0
and H0(X, L)is a free Fq[z]-module of rank rg+ 1
Proof. Under the condition 2g2< r, one has that H1(Xu,L|Xu) = 0 for every point
uU. Note that Hi(X, F)e=RiπFfor every coherent sheaf Fon X([2] III), and
applying ([2] III Corollary 12.9) one concludes the proof.
4 J.A DOM´
INGUEZ P´
EREZ, J.M MU ˜
NOZ PORRAS, AND G. SERRANO SOTELO
Under the hypothesis of Proposition 3.1, there exists an exact sequence of Fq[z]-
modules
(3.3) 0 H0(X, L(D)) H0(X, L)α
H0(X, OD)H1(X, L(D)) 0.
where H0(X, OD) is a free Fq[z]-module of rank n.
Remark 3.2.Let ηUbe the generic point of U, whose residue field is Fq(z); the fibre
Xη=π1(η) is a smooth, irreducible curve over Fq(z). Note that p1(η),...,pn(η) are n
different Fq(z)-rational points of the curve Xη. One then has a canonical decomposition
of H0(X, OD)ηas a Fq(z)-algebra
H0(X, OD)η=Fq(z)×n)
...×Fq(z).
Given a Fq[z]-module M, let us denote by Mηthe Fq(z)-vector space
Mη=MFq[z]Fq(z).
The sequence (3.3) induces an exact sequence of Fq(z)-vector spaces
(3.4) 0 H0(X, L(D))ηH0(X, L)η
αη
H0(X, OD)ηH1(X, L(D))η0.
Definition 3.3. The complete convolutional Goppa code associated with Land Dis
the image of the homomorphism αη
C(L, D) = ImH0(X, L)η
αη
H0(X, OD)ηFq(z)n.
Given a free submodule Γ H0(X, L), the convolutional Goppa code associated with
Γ and Dis the image of αη|Γη
C, D) = ImΓη
αη
Fq(z)n.
Remark 3.4.We use definition 2.4 of [5] as definition of convolutional codes. Any
matrix defining αη(respectively αη|Γη) is a generator matrix of rational functions for
the code C(L, D) (resp. C, D)).
The canonical decomposition H0(X, OD)ηFq(z)nas Fq(z)-algebras does not ex-
tend (in general) to a decomposition H0(X, OD)Fq[z]nas rings. In fact, one has a
canonical isomorphism of rings H0(X, OD)φ
Fq[z]nonly when p1(U),...,pn(U) are
disjoint sections. However, H0(X, OD) is a free Fq[z]-module; then, there exist (non-
canonical) isomorphisms of Fq[z]-modules:
H0(X, OD)φ
Fq[z]n)
...Fq[z],
which are not (in general) isomorphism of rings.
This allows us to give another definition of convolutional Goppa codes.
Definition 3.5. Given a trivialization φ:H0(X, OD)
Fq[z]nas Fq[z]-modules, one
defines the convolutional Goppa code C(L, D, φ) as the image of φα
H0(X, L)α
H0(X, OD)φ
Fq[z]n.
Anagously, one defines the convolutional Goppa code C, D, φ).
CONVOLUTIONAL CODES OF GOPPA TYPE 5
Let us assume (for the rest of the paper) that the invariants (r, n, g) satisfy the
inequality
2g2< r < n .
Proposition 3.6. Under the above conditions on (r, n, g),H0(X, L(D)) = 0 and
H1(X, L(D)) is a free Fq[z]-module. The following exact sequence is exact
(3.5) 0 H0(X, L)α
H0(X, OD)H1(X, L(D)) 0.
and remains exact when we take fibres over every point uU.
Proof. If 2g2< r < n,H0(Xu,L(D)|Xu) = 0 for every point uU; and applying
([2] III Corollary 12.9) one concludes.
Corollary 3.7. The convolutional code C(L, D, φ)has dimension k=rg+ 1 and
length n. Every matrix defining φαis a basic generator matrix [5] for C(L, D, φ).
Proof. This is a direct consecuence of the last statement of Proposition 3.6 and the
characterization of basic generator matrices of [5].
Let us consider the convolutional Goppa code C, D, φ) defined by a submodule
ΓH0(X, L) and a trivilization φ. With the above restrictions, one has:
Proposition 3.8. Every matrix defining φα|Γis a basic generator matrix for the
code C, D, φ)if and only if H0(X, L)/Γis a torsion-free Fq[z]-module.
Proof. The sequence (3.5) induces a diagram
0
0
0//Γ
α|Γ//H0(X, OD)//H1(X, Γ)
//0
0//H0(X, L)
//H0(X, OD)//H1(X, L(D))
//0
H0(X, L)/Γ0
Then, the kernel of H1(X, Γ) H1(X, L(D)) is isomorphic to H0(X, L)/Γ and
H1(X, L(D)) is free. This implies that the torsion elements of H1(X, Γ) are contained
in H0(X, L)/Γ, from which one concludes the proof.
The above results allow us to construct basic generator matrices for the codes
C, D, φ). If p1(U),...,pn(U) are disjoint sections and φthe canonical trivializa-
tion, this gives us a basic generator matrix for C, D). However, in general the codes
C, D) and C, D, φ) are different.
Let us describe a geometric way to obtain a basic generator matrix for C(L, D) and
C, D).
Assume that the curves p1(U),...,pn(U) meet transversally at some points, and let
¯
Xbe the blowing-up [2] of Xat these points. One has morphisms
¯
Xβ//
¯π=πβ
@
@
@
@
@
@
@X
π
U
6 J.A DOM´
INGUEZ P´
EREZ, J.M MU ˜
NOZ PORRAS, AND G. SERRANO SOTELO
such that the proper transform of Dunder πis a divisor ¯
D¯
Xsatisfying
¯
D=p1(U) · · · pn(U)β
D ,
and one has a canonical homomorphism of rings
0 ODβO¯
D
which induces
0πOD
β
¯πO¯
D
Fq[z]n,
where ¯πO¯
D
Fq[z]nis the canonical isomorphism of sheaves of rings.
βLis an invertible sheaf on ¯
Xand there exists a canonical homomorphism
βL O ¯
D0,
whose kernel is (βL)(¯
D). This induces
0 L ββL βO¯
D,
and taking global sections one obtains
0H0(X, L)γ
H0(X, ββL)µ
Fq[z]n.
The image of µis precisely a free submodule of Fq[z]nthat defines a basic generator
matrix for C(L, D).
Let us consider the sequence of homomorphisms
0H0(X, L)α
H0(X, OD)β
֒H0(X, O¯
D) = Fq[z]n.
βαis not in general a basic matrix, since H0(X, O¯
D)/H0(X, OD) has torsion. Let
us define
¯
H0(X, L) = {pFq[z]nsuch that λp H0(X, L) for some λFq[z]}.
¯
H0(X, L)/H0(X, L) is a torsion module and Fq[z]n/¯
H0(X, L) is torsion-free. Then,
every matrix defining the homomorphism ¯
H0(X, L)֒Fq[z]nis a basic generator
matrix for C(L, D).
This is an algebraic-geometric interpretation of Forney’s construction of the basic
matrices of a convolutional code [1].
4. Convolutional Goppa Codes associated with the projective line
Let P1= Proj Fq[x0, x1] be the projective line over Fq, and
X=P1×Uπ
U= Spec Fq[z]
the trivial fibration. Let us denote by t=x1/x0the affine coordinate in P1, and by p
its infinity point. Let us consider the following ndifferent sections of π
pi:UP1×U
defined in the coordinates (t, z) by
pi(z) = (αiz+βi, z), αi, βiFq.
Let D=p1(U) + ···+pn(U) and let Lbe the invertible sheaf on X
L=π
1OP1(rp)FqOU, r < n ,
CONVOLUTIONAL CODES OF GOPPA TYPE 7
The exact sequence (3.5) is in this case:
0H0(X, L)α//H0(X, OD)//H1(X, L(D)) //0.
|| ||
H0(P1,OP1(rp)) Fq[z]α//Fq[z]n
Taking the fibres over the generic point η, and the canonical trivialization (πOD)η
Fq(z)n, the homomorphism αηis the evaluation map at the points p1(η),...,pn(η)
αη:H0(P1,OP1(rp)) FqFq(z)Fq(z)n
αη(tj) = tj(p1(η)),...,tj(pn(η))=(α1z+β1)j,...,(αnz+βn)j,
where {1,t,...,tr}is the “canonical” basis of H0(P1,OP1(rp)) in the affine coordinate
t. The convolutional code C(L, D) is a kind of generalized Reed-Solomon (RS) code (for
z= 0 we obtain a classical RS-code).
Let Γ H0(P1,OP1(rp)) be the linear subspace generated by {ts,...,tr}. The
convolutional Goppa code C, D) is the image of the homomorphism
αη: ΓFqFq(z)Fq(z)n
tj7− αη(tj),for sjr .
In this case H0(X, L)/Γ(H0(P1,OP1(rp))/Γ) FqFq[z] is torsion-free. Then, by
Proposition 3.8 every matrix defining
α: Γ FqFq[z]H0(X, OD)
is a basic generator matrix. To compute a matrix for αexplicitly, we need to fix an
isomorphism of Fq[z]-modules
H0(X, OD)φ
Fq[z]n,
and this gives a generator matrix for C, D, φ). However, it would be desirable to com-
pute basic matrices for the codes C, D). We shall do this in general in a forthcoming
paper. Here we shall offer some explicit examples.
Example 4.1.Let a, b Fqbe two different non-zero elements, and
pi(z) = (ai1z+bi1, z), i = 1,...,n, with n < q .
The evaluation map αηover Γ is defined by the matrix
(4.1)
(z+ 1)s(az +b)s(a2z+b2)s... (an1z+bn1)s
(z+ 1)s+1 (az +b)s+1 (a2z+b2)s+1 ... (an1z+bn1)s+1
.
.
..
.
..
.
.....
.
.
(z+ 1)r(az +b)r(a2z+b2)r... (an1z+bn1)r
.
This matrix is a generator matrix for the code C, D). Using this construction we can
give concrete examples of CGC of dimension k=rs+ 1 that are Maximum-Distance
Separable (MDS) convolutional codes, i.e., whose free distance attains the generalized
Singleton bound [7].
8 J.A DOM´
INGUEZ P´
EREZ, J.M MU ˜
NOZ PORRAS, AND G. SERRANO SOTELO
If s=r, the convolutional Goppa code C, D) has dimension 1, degree r, and
(4.1) is a canonical (reduced and basic [5]) generator matrix. We can list a few
examples, where k/n,δand dare respectively the rate, the degree and the free
distance of the code.
field canonical generator matrix k/n δ d
F3={0,1,2}z+ 1 z+ 21/2 1 4
F4={0,1, α, α2}z+ 1 z+α z +α21/3 1 6
where α2+α+ 1 = 0
F5={0,1,2,3,4}(z+ 1)2(z+ 2)2(z+ 4)21/3 2 9
In these examples the sections p1,...,pnare disjoint, such that one has C, D) =
C, D, φ), where φ:H0(X, OD)
Fq[z]nis the corresponding canonical trivi-
alization.
If s < r, let us take aFqas a primitive element.
Now, the matrix (4.1) is reduced, since the matrix of highest-degree terms in
each row is a Vandermonde matrix of rank k. The sections p1,...,pnare not
disjoint, but in some cases the matrix (4.1) is actually basic and we do not have
to find an isomorphism of Fq[z]-modules, φ:H0(X, OD)
Fq[z]n, in order to
compute a basic generator matrix for the code C, D).
We present two examples of this situation.
field canonical generator matrix k/n δ d
F41 1 1
z+ 1 αz +α2α2z+α2/3 1 3
F5z+ 1 2z+ 3 4z+ 4 3z+ 2
(z+ 1)2(2z+ 3)2(4z+ 4)2(3z+ 2)21/2 3 8
Acknowledgments. We thank F.J. Plaza Mart´ın and E. omez Gonz´alez for many
enlightening comments, and J. Prada Blanco and J.I. Iglesias Curto for helpful ques-
tions that helped us to improve this paper.
CONVOLUTIONAL CODES OF GOPPA TYPE 9
References
[1] G.D. Forney Jr., Convolutional Codes I: Algebraic Structure, IEEE Trans. Inform. Theory 16
(1970) 720–738.
[2] R. Hartshorne, Algebraic Geometry Grad. Texts in Math., vol. 52, (Springer-Verlag, New York,
1977).
[3] T. Høholdt, J.H. van Lint and R. Pellikaan, Algebraic Geometric Codes, in: Handbook of Coding
theory, Ed. by V.S. Pless and W.C. Huffman (Elsevier, Amsterdam, 1998) 871–962.
[4] V. Lomadze, Convolutional Codes and Coherent Sheaves, AAECC 12 (2001) 273–326.
[5] R.J. McEliece, The Algebraic Theory of Convolutional Codes, in: Handbook of Coding theory,
Ed. by V.S. Pless and W.C. Huffman (Elsevier, Amsterdam, 1998) 1065–1138.
[6] J.H. van Lint and G. van der Geer, Introduction to Coding Theory and Algebraic Geometry DMV
Seminar, vol. 12, (Birkh¨auser, Basel, 1998).
[7] J. Rosenthal and R. Smarandache, Maximum Distance Separable Convolutional Codes, AAECC
10 (1999) 15–32.
[8] R. Smarandache and J. Rosenthal, Constructions of MDS-Convolutional Codes, IEEE Trans.
Inform. Theory 47 (2001) 2045–2049.
E-mail address :jadoming@usal.es, jmp@usal.es and laina@usal.es
Departamento de Matem´
aticas, Universidad de Salamanca, Plaza de la Merced 1-4,
37008 Salamanca, Spain
... A first attempt to define convolutional Goppa codes was made in [MDS04], where instead of a curve, a family of curves parameterized by the affine line was considered. Instead of points, disjoint sections of the projection of this family over the affine line were taken, and instead of divisors on a curve, a Cartier divisor and an invertible sheaf. ...
... Note however that in order to achieve simplicity the resulting objects defined as convolutional codes are F q (z)-vector spaces. For the construction of convolutional Goppa codes as submodules we refer to [MDS04]. ...
... Dicha estrategia ha dado como resultado los denominados códigos de Goppa convolucionales presentados primero en [MDS04] y después con herramientas más simples en [MDIS06], cuya construcción exponemos brevemente. ...
... One of the main problems in coding theory is the construction of codes with a large distance, such as so-called MDS codes. The aim of this paper is to give a very general method to construct onedimensional MDS convolutional codes using the techniques developed in our previous papers [1, 2, 3]. In Section 1 we give a general introduction to convolutional codes, reformulated in terms that enables a good understanding of the choices of the generator matrices and the submodules generated by them. ...
... Moreover, a characterization of one-dimensional MDS convolutional codes in terms of their associated block linear codes is given (Theorem 1.11). In Section 2 we describe the notion of Convolutional Goppa Code, introduced in [1] and [2] and we recall the construction of convolutional Goppa codes over the projective line. We use this construction in Section 3 to give families of examples of one-dimensional convolutional Goppa codes; moreover, we prove, using Theorem 1.11, that they are MDS. ...
... Definition 2.1. ([1],[2])The convolutional Goppa code C(D, G) associated with the pair (D, G) is the image of the F q (z)-linear map α : L(G) → F q (z) n Analogously, given a subspace Γ ⊆ L(G), one defines the convolutional Goppa code C(D, Γ) as the image of α | Γ . ...
Article
Full-text available
We give a general method to construct MDS one-dimensional convolutional codes. Our method generalizes previous constructions of H. Gluesing-Luerssen and B. Langfeld. Moreover we give a classification of one-dimensional Convolutional Goppa Codes and propose a characterization of MDS codes of this type.
... More precisely, we provide an algorithm for computing the free distance, a proof of its lower semicontinuity and its preservation by enlargements of the alphabet. As applications, we offer examples of families of MDS convolutional Goppa codes (CGC, see [3,14]), where the size of the base field has been kept as small as possible, as well as a method to produce MDS CGC of greater length. ...
... For simplicity, we shall restrict ourselves to the case of 1-dimensional codes constructed on the projective line ( [2]). Interested readers can check the general construction of higher-dimensional CGC on arbitrary curves in [3,14]. Observe that if a 1-dimensional convolutional code is MDS, then its column indices are all equal to the degree. ...
Article
Full-text available
We show that the free distance, as a function on a space parameterizing a family of convolutional codes, is a lower-semicontinuous function and that, therefore, the property of being Maximum Distance Separable (MDS) is an open condition. For a class of convolutional codes, an algorithm is offered to compute the free distance. The behaviour of the free distance by enlargements of the alphabet and by increasing the length is also studied. As an application, the algebraic equations characterizing the subfamily of MDS codes is explicitly computed for families of 1-dimensional convolutional Goppa codes (CGC).
... Methods from algebraic geometry have already been used by the authors in the context of convolutional codes. In particular, they made use of them to construct convolutional codes of Goppa type [10,2]. The characterization of this particular class of convolutional codes inside the moduli space is part of a work in progress. ...
... Let us use G 1 , G 2 to refer to the homogeneous polynomial matrices representing the morphisms of graded modules corresponding to φ 1 , φ 2 , and let g (j) i be the vector on the ith row of G j . g (1) i , g (2) i are homogeneous polynomial vectors of degree ν i . Therefore, Im( φ i ) is generated by ...
Article
Convolutional codes have the structure of an F[z]-module. To study their properties it is desirable to classify them as the points of a certain algebraic variety. By considering the correspondence of submodules and the points of certain quotient schemes, and the inclusion of these as subvarieties of certain Grassmannians, one has a one-to-one correspondence of convolutional codes and the points of these subvarieties. This classification of convolutional codes sheds light on their structure and proves to be helpful to give bounds on their free distance and to define convolutional codes with good parameters.
... Esses códigos possuem propriedades interessantes para diversas aplicações, além de serem assintoticamente bons [3,5,15]. Entretanto, apenas poucos deles [4,18,20] lidam com a construção de códigos convolucionais a partir da utilização de códigos AG como sua contrapartida fundamental. ...
Article
The performance of Rosenthal’s decoding algorithm is measured when it is applied to certain family of 1-dimensional MDS convolutional codes. It is shown that the performance of this algorithm depends on the algebraic structure of the convolutional code we are working with.
Article
Full-text available
In this paper, we construct new families of classical convolutional codes (CCC’s) as well as new families of quantum convolutional codes (QCC’s). The CCC’s are derived from (block) algebraic geometry (AG) codes. Furthermore, new families of CCC’s are constructed by applying the techniques of puncturing, extending, expanding and by the direct product code construction applied to AG codes. Additionally, utilizing the new CCC’s constructed here, we obtain new families of QCC’s. The parameters of these new codes are good. More precisely, in the classical case, a family of almost near maximum distance separable (MDS) codes are presented; in the quantum case, we construct a family of MDS (optimal) quantum convolutional codes.
Article
We present the family of generalized AG convolutional codes, constructed by using algebraic geometric tools. This construction extends block generalized AG codes on the one hand and several algebraic constructions of convolutional codes on the other. The tools employed to define these codes are also used to obtain information about their parameters and to determine conditions such that the resulting codes have optimal free distance.
Article
Convolutional Goppa codes (CGC) were defined in Appl. Algebra Eng. Comm. Comput., vol. 15, pp. 51-61, 2004 and IEEE Trans. Inf. Theory, vol. 52, 340-344, 2006. In this paper, we prove that every convolutional code is a CGC defined over a smooth curve over BBF q(z) and we give an explicit construction of convolutional codes as CGC over the projective line BBP BBF q(z)1. We characterize which convolutional codes are defined by a complete linear system over curves of genus 0, 1, and over hyperelliptic curves. We apply these results to provide detailed constructions of some linear block codes as Goppa codes.
Conference Paper
The algebraic geometric tools used by Goppa to construct block codes with good properties have been also used successfully in the setting of convolutional codes. We present here this construction carried out over elliptic curves, yielding a variety of codes which are optimal with respect to different bounds. We provide a number of examples for different values of their parameters, including some explicit strongly MDS convolutional codes. We also introduce some conditions for certain codes of this class to be MDS.
Article
Full-text available
Maximum-distance separable (MDS) convolutional codes are characterized through the property that the free distance attains the generalized singleton bound. The existence of MDS convolutional codes was established by two of the authors by using methods from algebraic geometry. This correspondence provides an elementary construction of MDS convolutional codes for each rate k/n and each degree δ. The construction is based on a well-known connection between quasi-cyclic codes and convolutional codes
Article
Full-text available
It is well known that a convolutional code is essentially a linear system defined over a finite field. In this paper we elaborate on this connection. We will define convolutional codes as the dual of a complete linear behavior in the sense of Willems. Using ideas from systems theory we describe a set of generalized first order descriptions for convolutional codes. As an application of these ideas, we present a new algebraic construction for convolutional codes. Index-Terms: Convolutional codes, behaviors, duality, first order representations, code constructions. I. Introduction In this paper we take a detailed look at convolutional codes from the perspective of linear systems theory with an emphasis on duality relations and on the different representations of these codes. Using these representations, we present a construction of convolutional codes with distance lower bounded by the complexity of the encoder. Throughout the relatively short history of the theory of convolutional codes...
Article
It is shown that the Pellikaan’s decoding algorithm of some families of Goppa codes C Ω (D,G) needs at most (g+1) effective divisors if the degree of G is odd and at most ⌊g/2⌋+1 effective divisors if the degree is even, where g is the genus of the curve used.
Article
We investigate the notion of cyclicity for convolutional codes as it has been introduced by Piret and Roos. Codes of this type are described as submodules of F[z] n with some additional generalized cyclic structure but also as specific left ideals in a skew polynomial ring. Extending a result of Piret, we show in a purely algebraic setting that these ideals are always principal. This leads to the notion of a generator polynomial just like for cyclic block codes. Similarly a parity check polynomial can be introduced by considering the right annihilator ideal. An algorithmic procedure is developed which produces unique reduced generator and parity check polynomials. We also show how basic code properties and a minimal generator matrix can be read off from these objects. A close link between polynomial and vector description of the codes is provided by certain generalized circulant matrices.
Article
. We present a sheaf-theoretic setting for convolutional codes in that coherent sheaves (over a projective line) play the same role as finite-dimensional linear spaces in the theory of block codes. Using coherent sheaves (and their cohomologies), we give natural straightforward proofs of the main results on the structural properties of convolutional codes.
Article
. A maximum distance separable (MDS) block code is a linear code whose distance is maximal among all linear block codes of rate k=n. It is well known that MDS block codes do exist if the field size is more than n. In this paper we generalize this concept to the class of convolutional codes of a fixed rate k=n and a fixed code degree ffi. In order to achieve this result we will introduce a natural upper bound for the free distance generalizing the Singleton bound. The main result of the paper shows that this upper bound can be achieved in all cases if one allows sufficiently many field elements. 1. Introduction Let F be a finite field and let C ae F n be an [n; k] linear block code. Let d(C) be the distance of C, i.e. d(C) is equal to the minimum Hamming distance between any two different code words x; y 2 C. The main linear coding problem asks for the construction of linear [n; k] codes whose distance d(C) is maximal among all linear [n; k] codes. The distance d(C) is always upper...
Article
A convolutional encoder is defined as any constant linear sequential circuit. The associated code is the set of all output sequences resulting from any set of input sequences beginning at any time. Encoders are called equivalent if they generate the same code. The invariant factor theorem is used to determine when a convolutional encoder has a feedback-free inverse, and the minimum delay of any inverse. All encoders are shown to be equivalent to minimal encoders, which are feedback-free encoders with feedback-free delay-free inverses, and which can be realized in the conventional manner with as few memory elements as any equivalent encoder, Minimal encoders are shown to be immune to catastrophic error propagation and, in fact, to lead in a certain sense to the shortest decoded error sequences possible per error event. In two appendices, we introduce dual codes and syndromes, and show that a minimal encoder for a dual code has exactly the complexity of the original encoder; we show that systematic encoders with feedback form a canonical class, and compare this class to the minimal class.