IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO.

5, MAY 2001 793

Interleaver Properties and Their Applications to the

Trellis Complexity Analysis of Turbo Codes
Roberto Garello, Member, IEEE, Guido Montorsi, Member, IEEE, Sergio Benedetto, Fellow, IEEE, and
Giovanni Cancellieri

Abstract—In the first part of this paper, the basic theory of in- quantities are then explored, especially those concerning
terleavers is revisited in a semi-tutorial manner, and extended to latency, a key parameter for applications.
encompass noncausal interleavers. The parameters that charac- Finally, the class of convolutional interleavers is considered,
terize the interleaver behavior (like delay, latency, and period) are
clearly defined. The input–output interleaver code is introduced and their practical implementation discussed. Block inter-
and its complexity studied. Connections among various interleaver leavers, which are the basis of most turbo code schemes, are
parameters are explored. The classes of convolutional and block in- introduced as a special case.
terleavers are considered, and their practical implementation dis- A turbo code is obtained by the parallel concatenation of
cussed. In the second part, the trellis complexity of turbo codes is simple constituent convolutional codes connected through an
tied to the complexity of the constituent interleaver. A procedure
of complexity reduction by coordinate permutation is presented, interleaver. Turbo codes, although decoded through an iterative
together with some examples of its application. suboptimum decoding algorithm, yield performance extremely
close to the Shannon limit [3]. In spite of the fact that several
Index Terms—Concatenated codes, interleavers, trellis com-
plexity, turbo codes. research issues are still open, turbo codes are obtaining a large
success and their introduction in many international standards
is in progress. For decoding purposes, the trellises of the con-
I. INTRODUCTION stituent convolutional encoders are almost always terminated,
so that the turbo code can be considered as a block code.
I NTERLEAVERS are simple devices that permute (usually
binary) sequences: they are widely used for improving error
correction capabilities of coding schemes over bursty channels.
Every linear block code can be represented by a minimal
trellis, which, in turn, can be used for soft-decision decoding
Their basic theory has received a relatively limited attention in with the Viterbi or the BCJR [4] algorithms. For decoding pur-
the past, apart from some classical papers [1], [2]. Since the in- poses, the complexity of a given trellis is usually expressed in
troduction of turbo codes [3], where interleavers play a funda- terms of complexity parameters like the maximum state com-
mental role, researchers have dedicated many efforts to inter- plexity (logarithm of the maximum number of states), the max-
leaver design. However, misunderstanding of basic interleaver imum branch complexity (logarithm of the maximum number
theory often causes confusion in the turbo code literature. of branches), and the average branch symbol complexity (av-
In this paper, interleaver theory is revisited. Our intent is erage number of branch symbols per information bit). More-
twofold: first, we want to establish a clear mathematical frame- over, for a given block code, it is well known that all com-
work which encompasses old definitions and results on causal plexity measures strongly depend on the ordering of the time
interleavers. Second, we want to extend this theory to noncausal axis, which in turn involves in principle all possible permuta-
interleavers, which can be useful for some application like, for tions of its segments.
example, turbo codes. It is a folk theorem among researchers in the field that
We begin by a proper definition of the key quantities Maximum Likelihood decoding of turbo codes presents a
that characterize an interleaver, like its minimum/maximum computational complexity increasing exponentially with the
delay, its characteristic latency, and its period. Then, interleaver interleaver length, so that it becomes prohibitively complex
equivalence and deinterleavers are carefully studied. Interleaver when the turbo code employs a large interleaver, because of the
trellis complexity is analyzed by the introduction of a proper huge number of states involved. Thus, simpler, suboptimum
input/output interleaver code. Connections between interleaver iterative decoding strategies have been proposed and suc-
cessfully applied. The previous, qualitative folk theorem may
contain a reasonable amount of truth. The trellis complexity of
Paper approved by M. Fossorier, the Editor for Coding and Communication turbo codes, however, has never been properly studied. In this
Theory of the IEEE Communications Society. Manuscript received August 1,
1997; revised July 6, 1999 and April 13, 2000. This work was supported in part paper, it is analyzed and related to that of the constituent codes
by Qualcomm, Inc. and Agenzia Spaziale Italiana. This paper was presented and interleaver. By the introduction of the uniform interleaver,
in part at the Sixth Communication Mini-Conference, Phoenix, AZ, November the evaluation of the “average” complexity of a turbo code of a
1997, and at the 2000 International Symposium on Information Theory, Sor-
rento, Italy, June 2000. given length can also be performed.
R. Garello and G. Cancellieri are with the Dipartimento di Elettronica ed Strong connections exist between the “true” trellis com-
Automatica, Università di Ancona, 60131 Ancona, Italy. plexity parameters (minimized by coordinate permutation) and
G. Montorsi and S. Benedetto are with the Dipartimento di Elettronica, Po-
litecnico di Torino, 10129 Torino, Italy. the free distance of the code , that characterizes the
Publisher Item Identifier S 0090-6778(01)04090-9. error probability performance at high signal-to-noise ratios. As
0090–6778/01$10.00 © 2001 IEEE
794 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO. 5, MAY 2001

a matter of fact, large values of trellis complexity quantities A. Complexity: Past Work on the Subject
are a necessary (but not sufficient) condition to achieve a large The literature on the trellis complexity of block codes is very
[5], [6]. rich. The subject was introduced in some classical papers [4],
Finally, a procedure to reduce turbo code complexity by coor- [8]–[10]. Many other papers deal with the subject, covering
dinate permutation, suggested by interleaver properties, is pre- methods to compute the complexity parameters, build the trellis,
sented. Its application is particularly interesting for the class of and establish bounds, or studies on the complexity reduction
row-by-column block interleavers, where it allows an impres- by coordinate permutation. Among them we can cite [11]–[26].
sive reduction that helps to understand the interleaver proper- Recent results on the complexity of convolutional codes are re-
ties. ported in [17] and [27]. For a complete list of references on the
In Section II, the definitions of the trellis complexity param- subject, see [28].
eters are recalled. In Section III, past work on interleaver theory
is reviewed. In Section IV, interleaver theory is revisited and ex-
B. Minimal Span Generator Matrix (MSGM) Matrices
tended to noncausal interleavers. The input–output trellis com-
plexity of interleavers is studied in Section V. In Section VI, A procedure for practically computing the state and branch
the connections between various interleaver parameters are ex- profile of a code has been presented in [20] and [22]. We say that
plored. Convolutional and block interleavers are considered in a generator matrix for with the LR (left/right) property
Section VII. In Section VIII, turbo codes are introduced, and has MSGM form. Given a row of
their complexity studied. , let be the smallest such that the
largest such that , and .
has the LR property if and
II. TRELLIS COMPLEXITY DEFINITIONS
for any pair of rows . Any generator matrix can be
For a linear binary block code of rate , we will use the easily transformed into an MSGM form [22]. For any
following notations referred to its minimal trellis [7]: we say that a row is edge-active in if , and is
• any codeword is an -tuple state-active in if and . Given
; in MSGM form, is the number of rows that are state-active
• is the state space at time its at time , and is the number of rows that are edge-active at
cardinality, i.e., the number of states at time , and time .
the state (space dimension) profile; Example 1: Consider the simple block code
• , is the trellis section with and . The
at time its cardinality, i.e., the generator matrix
number of branches at time , and
the branch profile.
Following [7], the state space can be defined as
, where is the projection of over the
interval and is the subcode of , composed by the has the LR property. It follows then
codewords with support enclosed in .
The trellis complexity parameters considered in this paper are
as follows.
• The maximum state complexity , defined as the
maximum of the state profile
We have . The minimal trellis
for is depicted in Fig. 1(a).

• The maximum branch complexity , defined as the C. Complexity Minimization by Coordinate Permutation
maximum of the branch profile Coordinate permutations can strongly change the complexity
parameters. In other words, given , there can exist an equiv-
alent code , where is an -long permutation,
mapping any sequence into ,
with , such that , and/or
• The average branch symbol complexity , defined
, and/or (in general, there may occur dif-
as the average number of branch symbols per information
ferent permutations to minimize different complexity param-
bit
eters). As a consequence, one can base a “real” measure of the
complexity of the code upon the parameters

(See Section II-D for a definition of that holds when more

than 1 bit labels each branch.)
GARELLO et al.: INTERLEAVER PROPERTIES AND THEIR APPLICATIONS TO THE TRELLIS COMPLEXITY ANALYSIS OF TURBO CODES 795

In general, the problem of finding the best permutation for

complexity minimization is NP-complete [25], [26].

D. Sectionalization
So far, we have considered an “atomic” representation of ,
i.e., a bit at every time . For clearness, in the fol-
lowing we will denote by , and , the complexity
parameters obtained when a sectionalization occurs such that a
group of bits corresponds to any time . The super-
script will be omitted only when obvious from the context.
As pointed out in [16], trellis sectionalization varies the state
and branch complexities, up to pathological cases like .
When , the definition of the average branch symbol com-
plexity becomes

is not affected by sectionalization; moreover, it is

strictly tied to the computational complexity of the decoding
algorithms [22]. This suggests that is perhaps the most
appropriate parameter for complexity considerations. Optimal
sectionalization of the code trellis to reduce its complexity has
been studied in [29]. Several authors also allow for sectional-
ization into varying numbers of bits per trellis section, as
in [29].

III. INTERLEAVER THEORY: PAST WORK ON THE SUBJECT

The theory of interleavers was established in the two classical
papers [1] and [2]. In [1] causal interleavers were deeply ana-
lyzed: several definitions and results, equivalent to those pre-
sented in this paper, will be referred in the following. More-
over, many significant results on the spreading properties of in-
terleavers (a subject that we have not considered in this paper)
and on their implementation were presented.
Other results were presented in [2] (see also [30]): Forney
Fig. 1. Minimal trellises for the codes of (a) Example 1 and (b) Example 2. invented periodic convolutional interleavers that are employed
in many applications and international standards.
Two considerations are important. First, classical interleaver
Example 2: Given the code of Example 1, the following per- theory only concerned causal devices. As a consequence, some
mutation: quantities defined in our paper (like, for example, equivalence
classes of interleavers, or the interleaver code) were not strictly
necessary, and not introduced in [1] and [2]. Second, other re-
sults on causal interleavers were not explicitly presented in [1]
leads to . The gener- and [2], and are poorly documented or scattered elsewhere in
ator matrix the literature.
As a matter of fact, we think that most of the results on causal
interleavers presented in this paper are to be considered as a
re-interpretation of previous ones, even when an explicit refer-
ence is missing. Among the main contribution of this paper are
has the LR property. It follows the general mathematical framework and all results concerning
noncausal interleavers, a subject which was not studied in the
past.
Recently, we partially presented our results in a preliminary
form in [31]. Similar definitions for causal interleavers can be
found in [32] together with some results about the interleaver
We have . spread. Some other results concerning the cycle decomposition
The minimal trellis of is depicted in Fig. 1(b). of an interleaver and its implementation are presented in [33].
796 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO. 5, MAY 2001

Fig. 2. The delay function and some parameters of the interleaver of Example 3.

IV. INTERLEAVERS: BASIC DEFINITIONS We say that is periodic with period , if there exists a
We begin to revisit interleaver theory by introducing some positive integer such that for all
basic definitions. They can be considered an extension of the
definitions introduced in [1] for causal interleavers. i.e.,
An interleaver is a device characterized by a fixed per-
mutation of the infinite time axis . maps bi-in- In this paper, we will only consider periodic interleavers, be-
finite input sequences into permuted cause nonperiodic permutations are not used in practice. The
output sequences , with period is usually referred to, in the turbo code literature,
. Although irrelevant for the interleaver proper- as the interleaver length or size, and is a crucial parameter in
ties, for simplicity we will assume in the following that is the determining the code performance. Often, it is also directly re-
binary alphabet . lated to the latency introduced in the transmission chain; this
is not strictly correct, as we will see soon that the minimum
A. Delays and Period delay introduced by the interleaver–deinterleaver pair is instead
We define the delay function as the characteristic latency . The ambiguity stems from the
common suboptimal implementation of a block interleaver (see
Section VII-B).
The interleaver action can then be described as Example 3: Consider this interleaver and its action on the
binary sequences , as shown at the bottom of the page.
has period , minimum and maximum delay
, and characteristic latency .
We also define the following.
Its delay function is depicted in Fig. 2.
The maximum delay
In the following, we will describe an interleaver of period
only by its values in the fundamental interval
. For the interleaver of Example 3 we
The minimum delay have .

B. Interleaver Equivalence

The characteristic latency Formal definition of equivalence classes of interleavers, not

introduced in classical theory, will prove useful in our frame-
work.
GARELLO et al.: INTERLEAVER PROPERTIES AND THEIR APPLICATIONS TO THE TRELLIS COMPLEXITY ANALYSIS OF TURBO CODES 797

We say that two interleavers are equivalent if one differs from and . Clearly,
the other only by a pure delay of the input or output sequence: . All the elements of have the same period and char-
and are equivalent if there exists a acteristic latency of and .
pair of integers such that for all Example 5: Given the interleaver of Example 3, its in-
verse interleaver is: . For the pair , we
have . The interleavers
and are deinterleavers for . For
or, equivalently the pair we have , and for the pair
we have .

E. Causal and Canonical Interleavers

We will denote an equivalence class of interleavers by the For practical purposes, an interleaver must be physi-
symbol . It can be easily proved that the characteristic cally realizable, or causal, i.e., it must satisfy the property
latency and the period are invariant for all interleavers
belonging to .
It is useful to introduce a subclass of com-
i.e., i.e., with
posed by all interleavers that are equivalent to by a pure delay
of the output sequences: and are equivalent up
To each interleaver with a certain (pos-
to translation if there exists an integer such that
sibly negative if is not causal), we can associate the (equiva-
lent causal) canonical interleaver with
Example 4: Consider the interleaver of Example 3 and
characterized by: . For
the two interleavers of period four:
, we have . (It is clear that an in-
and . Their delay functions are, re-
finite number of other interleavers belonging to could
spectively: and .
have been chosen as canonical for but, for simplicity, we have
We have: (with for and
chosen the only one belonging to ).
for ), and . The characteristic
Example 6: Let us consider the interleaver of period
latency is still equal to nine for both and .
with . has: ,
and . Its canonical interleaver has .
C. Inverse Interleaver We have:
Given , its inverse interleaver is . The .
delay functions of and are tied by: .
has the same period and characteristic latency of , and its min- F. Canonical Deinterleaver and Minimum Latency
imum/maximum delays are: and
Given a causal canonical interleaver , and its in-
.
verse , we know that all deinterleavers be-
long to and have the same characteristic latency
D. Deinterleavers and Latency of . Among them, we consider the equivalent causal canon-
The action of an interleaver must be inverted at ical interleaver associate with and call it the (equivalent
the receiver side by a deinterleaver such that any causal) canonical deinterleaver . By definition, has
sequence that enters is permuted into an and . The most interesting fea-
output sequence that is a delayed version of ture of regards the latency, and is clarified by the following
, with (in general, we corollary.
admit negative ). is the latency, i.e., the delay Corollary 1: Given a causal canonical interleaver with
introduced in the transmission chain by the interleaver/deinter- characteristic latency and its causal canonical deinterleaver
leaver pair. , the latency introduced by the pair is min-
Given , its inverse interleaver is certainly a imum and is equal to , i.e., for the deinterleaved sequence we
deinterleaver of , and it yields . Moreover, all the have: .
elements of are deinterleavers for . This simple lemma Proof: By definition, . Then
follows. . By the definition of canonical causal interleaver associ-
Lemma 1: Given an interleaver with period and ated with we have . From Lemma 1
characteristic latency , and its inverse , all it follows .
deinterleavers for are characterized by a permutation of Example 7: Given the interleaver and its causal canonical
the kind , i.e., belong to . All of them interleaver of Example 6, the inverse of is the interleaver
have period and characteristic latency . The latency with . The causal canonical deinterleaver
introduced by an interleaver/deinterleaver pair is equal is . The latency introduced by the pair
to . is . In fact, for a generic sequence
Proof: If is a deinterleaver for , we must have: at the output of we have the equation shown at the bottom
, but: . Then of the next page, and thus for all .
798 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO. 5, MAY 2001

Clearly,
are three sequences belonging to the code. The
generators of are all possible shifted versions of those three
sequences. The minimal trellis for is depicted in Fig. 4.
Any branch is labelled by 2 bits
Fig. 3. Interleaver code. that are the input and the output bits at time . The interleaver
is not causal. The number of branches exiting any state is equal
This Corollary coincides with the results presented in [1, Sec- to two only for .
tion II], where it was shown that the latency introduced by the in- For the connections between the state space of the interleaver
terleaver/deinterleaving (called unscrambler) pair is equal to the code and that of a real machine able to exactly realize the
maximum delay (called the encoding delay) when the min- interleaver input/output relationship, see Sections V-C and VII.
imum delay is equal to zero. It also coincides with Theorem 1
of [32] on latency (called minimum total delay). A. Interleaver Trellis Complexity
The latency is important for system applications having For causal interleavers, it is well known and intuitive that the
stringent delay requirements. The previous corollary shows that state space size is constant, because at each time 1 bit comes in
the minimum latency introduced by an interleaver/deinterleaver and 1 bit goes out. When more general interleavers (noncausal,
causal pair is equal to the characteristic latency . Practical too) are considered, the following general theorem holds for the
motivations can lead to latencies greater than this minimum state profile of the interleaver code (simply in
value (see Section VII-B for the two-register implementation of the following).
block interleavers). For this reason, in the turbo code literature, Theorem 1: For every interleaver code
the minimum latency and the period are often confused.

V. THE INPUT–OUTPUT INTERLEAVER CODE AND where

ITS COMPLEXITY
The study of the state space of a causal interleaver is intuitive
and does not need a specific theory. However, in this paper we
also study noncausal interleavers which are important for some The state space is
applications, for example, turbo code schemes. For this reason,
we introduce the new concept of input–output interleaver code.
The “code” formed by the set of all output sequences
of an interleaver with unconstrained input sequences Proof: We can use the definition
spans over all . As a consequence, it has a trivial state space (see Section II). The sequences , where is an
of cardinality for any . all-zero sequence with the exception of (then is
Given an interleaver , we introduce the (input–output) in- an all-zero sequence, with the exception of ), form
terleaver code defined as the set of all input/output inter- a set of generators for . All sequences with
leaver sequence pairs , where the input sequences span and are generators of and all those with
over all (see Fig. 3). Every codeword is a bi-infinite se- and are all-zero sequences. The remaining are linear
quence , where independent sequences.
The state space of an interleaver code at time is composed
by two components: consists of the input bits
is the pair composed by the input and the output bit at time . A that must be kept in memory because they will be output in the
minimal trellis for will have 2 bits labeling any future, while consists of the bits that have been
branch; the number of branches exiting any state will be equal output previously in the past and must be kept in memory until
to two for all times only for causal interleavers. time to give the corresponding input bit.
Example 8: Consider an interleaver with period Example 9: For the interleaver of the previous example we
and . All sequences of are of the kind have the following.
• For and .
• For , which contributes with to
, and , which contributes with to .
Then and .
GARELLO et al.: INTERLEAVER PROPERTIES AND THEIR APPLICATIONS TO THE TRELLIS COMPLEXITY ANALYSIS OF TURBO CODES 799

Fig. 4. Minimal trellis for the interleaver code of Example 8.

• For • For . Then and

and . .
• For . Then and
.
B. The Trellis Complexity of a Causal Interleaver
For a causal, realizable, interleaver, one always has: . The interleavers belonging to an equivalence class have an-
The previous Theorem 1 becomes the following corollary. other invariant (other than the period and the characteristic la-
Corollary 2: For a causal interleaver, given the set tency).
, we have and Lemma 2: Any interleaver belonging to the equivalence
, constant with . class has a canonical interleaver with the same con-
Proof: By the definition of causal interleaver, . straint length.
Given , consider the set . There are two possibilities: if Proof: Given , its equivalent canonical causal in-
then ; otherwise, if then terleaver is with . For a
. In both cases . generic , with we have
A causal interleaver has a constant number of states for . Its equivalent canonical inter-
each : we will call the constraint length of . As a leaver is with
consequence, the trellis complexity parameters of the interleaver . Let us compute of Theorem 1. For we have
code for a causal interleaver are simply . For we have

Example 10: Consider the (noncausal) interleaver of pe-

where . Then
riod two with
.

C. Minimal Encoder of a Causal Interleaver

To compute the state profile of the interleaver code , we have Given a causal interleaver , we are interested in the con-
the following. struction of a finite-state machine able to realize its input/output
relationship. In analogy with the code terminology, we will call
• For . Then and
it an encoder for . If the interleaver is not causal, an encoder
.
can be built for a causal equivalent.
• For and . Then
Following [34] (Section III-C), an encoder can be viewed as
and .
The equivalent causal canonical interleaver for is with the set of all the admissible input–output sequence pairs of the
interleaver, i.e., as a code over the product alphabet .
Formally, we could then build the minimal space state of this
code (as in [7]), and a minimal (systematic) encoder for it (as in
[34]), which exactly realizes the input/output relationship.
This means that a minimal (with the minimum possible
We have the following. number of states) encoder for a causal interleaver has a space
800 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO. 5, MAY 2001

state in one-to-one correspondence with the space state of the average absolute delay is and it coincides with the average
interleaver code . of the state profile.
An important consequence of Corollary 2 is that a minimal Given a causal interleaver and a causal deinterleaver , the
encoder for a causal interleaver does not need to keep memory sum of the average delays through and is clearly equal to the
of , or bits, but only at each time . They are the input introduced latency . This lemma follows from Corollary
bits forming that must be kept in memory to be released 3.
in the future, and their number is constant because, in a causal Lemma 4: The sum of the constraint lengths of a causal in-
interleaver, at any time an input bit arrives and one of the terleaver/deinterleaver pair is equal to the latency
input bits in memory, or itself, is output. See Section VII
for more details on the practical implementation of the encoder.

VI. CONNECTIONS AMONG INTERLEAVER PARAMETERS

When and are the canonical causal interleaver/deinter-
In this section, we explore the relations between the various leaver pair, the introduced latency is equal to the characteristic
interleaver parameters. The connection between the state spaces latency, and then we have the following key property.
of an interleaver and its inverse is clarified by the following Corollary 4: The sum of the constraint lengths of a causal
lemma. canonical interleaver/deinterleaver pair is equal to the
Lemma 3: Given an interleaver and its inverse characteristic latency
, and their interleaver codes and , their state space
profiles are equal: .
Proof: For any , there is a one-to-one correspondence be-
tween and for , and and for . In fact,
. By posing one has: This corollary coincides with [1, Theorem 4], where it was
. shown that the sum of the constraint lengths of the interleaving
A strong relation exists between the delay function of an in- and the deinterleaving (called minimum storage capacities) is
terleaver and its state space. equal to the latency.
Theorem 2: Given an interleaver of period , the Corollary 4 completely clarifies the fundamental role of the
average of the absolute values of the delays is equal to the av- characteristic latency of an interleaver. In fact, given , its
erage of the state profile characteristic latency is equal to the following:
• the difference between the maximum and the minimum
delay of ;
• the difference between the maximum and the minimum
delay of any interleaver equivalent to ;
Proof: Consider a position with a certain delay . • the difference between the maximum and the minimum
If is negative, belongs to the sets for all the fol- delay of any deinterleaver for ;
lowing positions . If is positive, belongs • the maximum delay of the canonical interleaver asso-
to the sets for all the previous positions . In both ciate with ;
cases, increases of units the sum • the maximum delay of the canonical deinterleaver as-
. sociate with ;
• the sum of the constraint lengths of and ;
A. Causal Interleavers • the minimum latency introduced in the transmission chain
We have seen that for a causal interleaver the state profile is by an interleaver/deinterleaver pair involving or any in-
constant and equal to . The previous Theorem 2 becomes the terleaver equivalent to .
following corollary.
Corollary 3: For a causal interleaver, the average delay is VII. CONVOLUTIONAL AND BLOCK INTERLEAVERS
equal to the constraint length The interleaver parameters and properties previously intro-
duced hold for all periodic interleavers. In the following, we
will introduce a way to describe general periodic interleavers
that will be called convolutional interleavers, as opposed to the
special and important subclass of block interleavers.
Example 11: Given the interleaver of Example 10, we have
and . The average absolute delay is 2. Since A. Convolutional Interleavers
and , it coincides with the average of the state Given a periodic interleaver of period ,a
profile. Given the causal canonical interleaver of the same convenient and completely general way to represent the permu-
Example, we have and . The average absolute tation is the following:
delay is 2 and coincides with the constraint length .
Finally, given the interleaver with of Ex-
(1)
ample 9, we have , and . The
GARELLO et al.: INTERLEAVER PROPERTIES AND THEIR APPLICATIONS TO THE TRELLIS COMPLEXITY ANALYSIS OF TURBO CODES 801

Fig. 6. Minimal memory implementation of an interleaver.

Fig. 5. Structure of a convolutional interleaver.
Proof: For a causal interleaver, we proved in Corollary 3
that the average delay is equal to the constraint length, so that
where is a finite basic permutation of length
, and is the th element of a shift vector of elements
taking values in .
A visual implementation of the convolutional interleaver de-
scribed in (1) is shown in Fig. 5, where the reader can recognize
the similarities with the familiar structure of convolutional in- From (1) we then have
terleavers described by Ramsey and Forney in [1] and [2]. The
input stream is permuted according to the basic permutation .
The th output of the permutation register is sent to a delay line
with a delay , whose output is read and sent out. Notice
that in the realization of Fig. 5 the permutation is always non-
causal, except for the case of the identity permutation. As a con-
sequence, it is in general not physically realizable. In the next
subsection, we will describe a minimal realization of any peri-
odic causal interleaver.
Example 12: A simple class of interleavers stems from the
choice of the identity basic permutation

In this section we describe an implementation with minimal

memory requirements of the encoder for a generic periodic
causal interleaver. The encoder is realized through a single
As an example, choosing , and memory with size equal to the constraint length . The input
yields the equation shown at the bottom of the and output bits at a given time are read and stored with a single
page. Read-Modify-Write operation acting on a single memory cell
1) Implementation of Convolutional Interleavers with Min- whose address is derived as follows (see Fig. 6).
imal Memory Requirement: We have already proved in Corol- At time we read from a memory cell, and write
lary 2 that the minimal memory requirement for the implemen- into it. As a consequence, the successive contents of
tation of a causal interleaver is the constraint length , which that memory cell are
in turn is equal to the average of the delay profile. An important Thus, the permutation induces a partition
property that relates the shift vector to the constraint length of the set of integers with the following equivalence relation-
of causal interleavers is given in the following lemma. ship:
Lemma 5: The constraint length of a causal interleaver of
period is related to the shift vector through
and the constraint length is also the cardinality of this parti-
tion.
With the previous considerations in mind, we can devise the
following algorithm to compute the addresses to be used in
the minimal memory interleaver of Fig. 6.
802 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO. 5, MAY 2001

1) Set the running variable . so that and the address sequence is given by
2) For all .
3) If , then the input symbol is directly trans-
ferred to the output. In this case the memory contents are
Example 15: The equivalent causal canonical interleaver of
not modified (the address will be conventionally set
Example 10 has the following parameter:
to 0).
4) If then and .
5) If then .
Since the interleaver is periodic, the address sequence
is also periodic. Its period can be obtained by deriving the fol-
lowing expression of the th power of the permutation in the
fundamental interval
so that lcm and the address sequence is given
by

To derive the period of the sequence we must use the cycle

decomposition of the basic permutation B. Block Interleavers
Block interleavers, a particular case of the general class of
periodic interleavers previously described, are very important
for applications, and form the basis of most turbo code schemes.
For each cycle of the basic permutation, we can define A block interleaver is generated by a permutation of length
the following basic period: , made periodic on all , and so yielding the
infinite permutation

(2)

where is the length of the cycle and is its th element. The period of is clearly equal to the length of . A
The periods of a cycle then corresponds to the sum of all the block interleaver (apart when is the identity) is not causal,
shift vector elements associated to the cycle elements. and has a nonpositive minimum delay .
Finally, it can be proved that the period of the address se- The maximum delay is and then the
quence is times the least common multiple of all periods characteristic latency is .
Block interleavers are a particular case of convolutional in-
lcm terleavers, obtained by posing in the general
representation (1). A convolutional interleaver with period
Example 13: As an example, consider the causal periodic is equivalent to a block interleaver if there exists an
permutation with period ; we such that the set is a set
have of adjacent numbers. As an example, this happens if the shift
vector in the representation (1) has the form

The basic permutation can be decomposed as follows:

So that Among all possible choices to construct a causal interleaver

lcm equivalent to a given block interleaver, two are particularly in-
teresting, and will be described in the following.
Applying the algorithm previously described, we have in fact 1) The Canonical Causal Interleaver of a Block Inter-
that the address sequence is periodic with period 12 leaver: For a block interleaver , generated by an -length
permutation with a certain minimum delay , we
can consider its equivalent causal canonical interleaver
Example 14: The delay permutation has the following pa- , with , with
rameters: , and . By definition of canonical
interleaver, is the best choice in terms of latency. By
using the canonical deinterleaver, the latency is equal to
. The constraint length of is charac-
terized by the following lemma.
GARELLO et al.: INTERLEAVER PROPERTIES AND THEIR APPLICATIONS TO THE TRELLIS COMPLEXITY ANALYSIS OF TURBO CODES 803

Lemma 6: The canonical interleaver of a block middle of the permutation, bits have been anticipated and
interleaver with minimum delay has constraint bits will be released in the future).
length Example 17: The case yields this per-
mutation of length and the corresponding state profile
computed by applying Theorem 1

Proof: By definition, we have iff . Since

, if we compute the set of Theorem
1 for , we have .
Lemma 6 shows that a minimal encoder for needs only
cells. The algorithm previously presented in Sec- VIII. APPLICATION TO TURBO CODES
tion VII-A-1 can be applied to realize it.
2) The Two-Register Causal Interleaver of a Block Inter- For simplicity, in this paper we focus on turbo codes, or par-
leaver: In practice, to make a block interleaver causal, the allel concatenated convolutional codes (PCCC), of rate- ob-
causal interleaver with is often tained from the following:
used instead of the canonical interleaver. corresponds to • two equal binary systematic convolutional encoders
an encoder implementation largely used in practice, which and of rate- and constraint length ;
consists of two registers of length used alternatively, one • a block interleaver of length .
for writing the input bits of the current block, and the other Every information frame of bits is read out in two different
for reading the output permuted bits of the preceding block. orders for the two encoders. The input bits of are not per-
Clearly, in this case has a maximum delay muted, whereas those of are permuted by . Therefore, given
usually larger than , and then it the information frame , the
leads to a nonminimum latency. If also the deinterleaver for th input bit is for and for . The two en-
is realized by the same two-register strategy, the introduced coders are systematic, and produce two parity check bits
latency is equal to . and , respectively. The input bits of the second encoder are
Example 16: The block interleaver with not transmitted, while the information bits and the two parity
and has and check bits are multiplexed from top to bottom. At the end of the
. Its canonical interleaver is with information frame of bits, the two encoders are terminated,
, with a minimum . i.e., driven back to the zero state. As explained in [35], this can
The two-register causal interleaver is with be done in bit times. In this way, other triplets are generated.
, with . Let us define . By termination, the turbo code be-
By using the canonical deinterleaver of with comes an block code with parameters .
, the pair introduces Every codeword will be denoted by ,
a latency equal to 7. with .
By using the two-register deinterleaver for with
, the pair A. The Trellis Complexity of Turbo Codes
introduces a latency equal to 16. To compute the minimal trellis of a turbo code and its com-
3) The Trellis Complexity of a Block Interleaver: The state plexity measures, one can directly apply the procedure based on
profile of noncausal interleavers has been characterized in The- the MSGM form of a generator matrix described in Section II.
orem 1. For a block interleaver, the minimal trellis has a partic- We will consider a sectionalization of bits for each time
ular form: it only has one state at the beginning and at the end . The following theorem ties the state space of a turbo
of its period. code (simply in the following) to the state profile of
Corollary 5: For a block interleaver for and the constituent block interleaver .
. Theorem 3: Given a turbo code constituted by two con-
4) Row-by-Column Block Interleavers: Row-by-column in- stituent encoders of constraint length and a block interleaver
terleavers are a class of block interleavers widely used in prac- with state profile , the state profile of the turbo code is
tice. Given , the input bits are written row-wise equal to
into a matrix with rows and columns and read column-
wise. The corresponding permutation is with

Proof: Denote by an
input sequence, with all bits equal to zero, with the exception
An row-by-column interleaver has parame- of , and by the corresponding
ters , and encoded sequence. The sequences , with ,
. The minimal trellis of a form a set of generators for . At time , the state space is given
row-by-column interleaver is very regular. In particular, when by . There are sequences , with
is a power of two, it is easy to prove that (in the and . All their projections are independent
804 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO. 5, MAY 2001

and a number of admissible positions with .

Moreover, there are numbers to be considered,
because , then .
Fixed , we can compute the sum in the
same way, by considering the contribution to of any
number , with . With the same method of
before, we obtain . Finally we have
.
Note that the state profile can be a noninteger number,
because it corresponds to the “average” of the state profiles of all
interleavers with length . Even if a uniform interleaver does
not really exist, all the theorems of this section are referred to
it (instead of the average over all the interleavers). This way,
the theorems look more immediate. This is in fact the reason
why the uniform interleaver has become a standard tool for turbo
Fig. 7. State profile for the interleaver code and the turbo code of Example 18. code analysis.
The “average” of the maximum state complexity over all
generators of . There are sequences , with block interleavers of length is easily computed.
and . All their projections are independent Corollary 6: A uniform block interleaver of length has a
generators of . All projections , with and maximum state complexity equal to:
, are generators of . All projections , with Proof: The maximum of the state profile
and are elements of . The maximum is reached in the middle point .
number of independent generators among them is . Given Theorem 3 and Theorem 4, we can now establish the
Example 18: Consider a turbo code composed by two equal “average” of the state profile and the “average” of the maximum
four-state, rate- , convolutional encoders with state complexity over all turbo codes employing a block inter-
and , and a block rectangular interleaver leaver of length .
with and . In Fig. 7 we Corollary 7: The state profile of a turbo code formed by
report the state profile of the turbo code and of the constituent two constituent encoders of constraint length and a uniform
interleaver. The largest difference is equal to according block interleaver of length is equal to
to Theorem 3.
with
B. The Average Complexity of Turbo Codes
A “uniform interleaver” of length , introduced in [36], and its maximum state complexity to
is a probabilistic device that acts as the “average” of all possible
block interleavers of length . Formally, given a binary input with
frame of Hamming weight maps
into one of the output frames of weight with probability
.
C. Reducing the Trellis Complexity of Turbo Codes
The following theorem allows to compute the state profile of
a block uniform interleaver of length , that coincides with the A terminated turbo code is a block code ,
“average” of the state profile over all block interleavers of and its complexity parameters , and strongly de-
length . pend on the constituent interleaver, as explained by Theorem 3.
Theorem 4: A uniform block interleaver of length has a It would be of great interest to find a permutation of the
state profile equal to code sequences that minimizes the complexity parameters, i.e.,
search for , and . Finding the best permutation
with for minimizing turbo code trellis complexity is still an open re-
search problem at this stage, and we cannot offer the final solu-
tion. (Note that, in general, the problem of finding the best per-
Proof: At time is the average of the state mutation is NP-complete [25], [26].) Instead, we will present a
space dimension over all block interleavers of length permutation choice, based on a result on interleaver code com-
For any block interleaver, The- plexity minimization, and show through examples how it can
orem 1 holds, and . Fixed , we first compute substantially decrease the trellis complexity of the turbo code.
the sum . Consider all permutations. Any 1) Minimizing the Complexity of an Interleaver
number , with , contributes to a Code: In this section, we will use coordinate permuta-
number of times equal to the number of permutations such that tions of the interleaver code sequences. Given a sequence
, with . Among all possible permu- , a permutation acting on will be
tations, there are permutations with this property. denoted by the pair where and are infinite
In fact, there are permutations such that , permutations and .
GARELLO et al.: INTERLEAVER PROPERTIES AND THEIR APPLICATIONS TO THE TRELLIS COMPLEXITY ANALYSIS OF TURBO CODES 805

Fig. 8. State and branch profile for the turbo code of Example 19, evaluated Fig. 9. State and branch profile for the turbo code of Example 20, evaluated
directly and after applying p . directly and after applying p .

Note that for two equivalent interleavers ,

there always exists a permutation ,
where and are pure delay permutations, such that
.
Given an interleaver and its code , we can look
for some permutation that minimizes the com-
plexity parameters of . The obvious solution for any is
the permutation . In this case is composed
by all pairs , for any input sequence . The code
has trivial dynamics with , and
. Obviously, the permutation yields
the same minimization results.
2) Application to the Reduction of the State Com-
plexity of Turbo Code: Given a turbo code and a code
sequence ,
Fig. 10. State profile for the turbo codes of Example 21, evaluated directly and
we will consider permutations acting on of the kind after applying p .
where and are permutations of
and .
D. The Trellis Complexity of Turbo Codes with Rectangular
To reduce the complexity of the turbo code employing a block
Interleavers
interleaver , we consider two permutations. They derive di-
rectly from those minimizing the complexity of the interleaver Impressive results in terms of reduction of turbo code trellis
code, which had been shown to be or, equivalently, complexity through the application of the two previously intro-
. The two permutations are1 duced permutations can be obtained for the class of row-by-
column block interleavers. By using the results obtained in Sec-
tion VII-B-4 and Theorem 3, we know that when is a power
of two
Example 19: Consider a turbo code composed by the
same constituent convolutional encoders of Example 18,
and a “palindrome” interleaver with :
By applying ( , respectively) when
. In Fig. 8
, we obtain a consistent reduction to
we report the state and branch profile of the turbo code
evaluated directly and through the permutation , by
applying the MSGM method of [22] recalled in Section II.
The curves obtained after applying show a significant Example 20: Consider a turbo code composed by the same
complexity reduction: the maximum state complexity passes constituent convolutional encoders of Example 18, and a block
from (65536 states) to (16 states). The rectangular interleaver with and . In
same profiles are obtained by applying . Fig. 9 we report the state and branch profile of the turbo code
1Since the permutation  has length N , while the turbo codewords have evaluated directly and through the permutation , by ap-
length n = N +  after termination, we extend  by posing  (i) = i for plying the MSGM method. The curves obtained after applying
N  i <n. show a significant complexity reduction. The average
806 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 49, NO. 5, MAY 2001

branch symbol complexity decreases from 253 625 branch sym- [8] J. K. Wolf, “Efficient maximum likelihood decoding of linear block
bols per information bits to 77 625 branch symbols per informa- codes using a trellis,” IEEE Trans. Inform. Theory, vol. IT-24, pp. 76–80,
Jan. 1978.
tion bits. [9] J. L. Massey, “Foundation and methods of channel encoding,” in Proc.
Example 21: Consider a turbo code with the same con- Int. Conf. Information Theory and Systems, Berlin, Germany, Sept. 1978.
stituent convolutional encoders of Example 18, and rectangular [10] G. D. Forney Jr., “Coset codes—Part II: Binary lattices and related codes,”
IEEE Trans. Inform. Theory, vol. 34, pp. 1152–1187, Sept. 1988.
interleavers with different length , but equal column [11] D. J. Muder, “Minimal trellises for block codes,” IEEE Trans. Inform.
number . Let us consider , Theory, vol. 34, pp. 1049–1053, Sept. 1988.
and . In Fig. 10 we report the state [12] T. Kasami, T. Takata, T. Fujiwara, and S. Lin, “On the optimum bit orders
with respect to the state complexity of trellis diagrams for binary linear
profile of the two turbo codes, evaluated directly and through codes,” IEEE Trans. Inform. Theory, vol. 39, pp. 242–245, Jan. 1993.
the permutation . The result is striking: the two curves [13] , “On complexity of trellis structure of linear block codes,” IEEE
obtained directly show a dependence of the maximum state Trans. Inform. Theory, vol. 39, pp. 1057–1064, May 1993.
[14] A. D. Kot and C. Leung, “On the construction and dimensionality of
complexity on , and, in particular, a value of equal to linear block code trellis,” Proc. 1993 IEEE Int. Symp. Information
18 for and to 130 for . Instead, the two curves Theory, p. 291, Jan. 1993.
obtained after applying yield a significant complexity [15] Y. Berger and Y. Be’ery, “Bounds on the trellis size of linear block
codes,” IEEE Trans. Inform. Theory, vol. 39, pp. 203–209, Jan. 1993.
reduction, and, more important, a maximum state complexity [16] G. D. Forney Jr., “Dimension/length profiles and trellis complexity
that is independent from , and given by . Using of linear block codes,” IEEE Trans. Inform. Theory, vol. 40, pp.
the algorithm described in [37], we have computed the free 1741–1752, Nov. 1994.
[17] V. Sidorenko and V. V. Zyablov, “Decoding of convolutional codes
distance of the two turbo codes, and verified that it is also using a syndrome trellis,” IEEE Trans. Inform. Theory, vol. 40, pp.
independent from ; this suggests that a simple increase of 1663–1666, Sept. 1994.
the interleaver length may not be beneficial, and, instead, that [18] F. R. Kschischang and G. B. Horn, “A heuristic for hordering a linear
block code to minimize trellis state complexity,” in Proc. 32nd Annu.
the interleaver in a PCCC construction should be chosen aiming Allerton Conf. Communication, Control, and Computing, Sept. 1994,
at maximizing the “true” trellis complexities parameters. pp. 75–84.
[19] S. J. Dolinar Jr., L. L. Ekroot, A. B. Kiely, R. J. Mc Eliece, and W. Lin,
“The permutation trellis complexity of linear block codes,” in Proc. 32nd
IX. CONCLUSION Annu. Allerton Conf. Communication, Control, and Computing, Sept.
1994, pp. 60–74.
In this paper, we have defined the main parameters that char- [20] F. R. Kschischang and V. Sorokine, “On the trellis structure of block
acterize an interleaver, by extending classical concepts to non- codes,” IEEE Trans. Inform. Theory, vol. 41, pp. 1924–1937, Nov. 1995.
[21] C. C. Lu and S. H. Huang, “On bit-level trellis complexity of Reed-
causal interleavers, too. After introducing the input/output in- Muller codes,” IEEE Trans. Inform. Theory, vol. 41, pp. 2061–2064,
terleaver code, we have evaluated its complexity, and explored Nov. 1995.
the connections between the various parameters. We have then [22] R. J. McEliece, “On the BCJR trellis for linear block codes,” IEEE
Trans. Inform. Theory, vol. 42, pp. 1072–1092, July 1996.
focused on causal interleavers and on block interleavers, that are [23] A. B. Kiely, S. J. Dolinar Jr., R. J. McEliece, L. L. Ekroot, and W. Lin,
both important for applications. These concepts have been ap- “Trellis decoding complexity of linear block codes,” IEEE Trans. In-
plied to the evaluation of the trellis complexity of turbo codes. form. Theory, vol. 42, pp. 1687–1697, Nov. 1996.
[24] A. Engelhart, J. Maucher, and V. Sidorenko, “Heuristic algorithms for
We have also faced the problem of finding a time axis permuta- ordering a linear block code to reduce the number of nodes of the min-
tion that reduces the complexity of turbo codes. imal trellis,” Proc. 1998 IEEE Int. Symp. Information Theory, p. 206,
Aug. 1998.
[25] G. B. Horn and F. R. Kschischang, “On the intractability of permuting a
ACKNOWLEDGMENT block code to minimize trellis complexity,” IEEE Trans. Inform. Theory,
vol. 42, pp. 2042–2048, Nov. 1996.
The authors wish to thank the Editor and the reviewers for [26] A. Vardy, “The intractability of computing the minimum distance of a
code,” IEEE Trans. Inform. Theory, vol. 43, pp. 1757–1766, Nov. 1997.
their valuable comments that helped to refocus and improve the [27] R. J. McEliece and W. Lin, “The trellis complexity of convolutional
original manuscript. They are also grateful to F. Chiaraluce and codes,” IEEE Trans. Inform. Theory, vol. 42, pp. 1855–1864, Nov. 1996.
G. D. Forney, Jr., for their appropriate comments. [28] S. Lin, T. Kasami, T. Fujiwara, and M. Fossorier, Trellises and Trellis-
Based Decoding Algorithms for Linear Block Codes. Norwell, MA:
Kluwer, 1998.
REFERENCES [29] A. Lafourcade and A. Vardy, “Optimal sectionalization of a trellis,”
IEEE Trans. Inform. Theory, vol. 42, pp. 689–703, May 1996.
[1] J. L. Ramsey, “Realization of optimum interleavers,” IEEE Trans. In- [30] G. D. Forney Jr., “Interleavers,” U.S. Patent 3 652 998, Mar. 1972.
form. Theory, vol. IT-16, pp. 338–345, May 1970. [31] S. Benedetto, R. Garello, and G. Montorsi, “The trellis complexity
[2] G. D. Forney Jr., “Burst-correcting codes for the classic bursty channel,” of turbo codes,” in Proc. Sixth Communication Theory Mini-Conf.,
IEEE Trans. Commun. Technol., vol. COM-19, pp. 772–781, Oct. 1971. GLOBECOM’97, Phoenix, AZ, Nov. 1997, pp. 60–65.
[3] C. Berrou, A. Glavieux, and P. Thitimajshima, “Near Shannon limit [32] K. Andrews, C. Heegard, and D. Kozen, “A theory of interleavers,” 1997
error-correcting coding and decoding: Turbo-codes,” in Proc. ICC’93, IEEE Int. Symp. Information Theory, 1997. Recent Results and Tech.
Geneva, Switzerland, May 1993, pp. 1064–1070. Rep. 97-1634, Comput. Sci. Dept., Cornell Univ..
[4] L. R. Bahl, J. Cocke, F. Jelinek, and J. Raviv, “Optimal decoding of linear [33] M. Mondin and F. Daneshgaran, “Realization of permutations with min-
codes for minimizing symbol error rate,” IEEE Trans. Inform. Theory, imal delay with applications to turbo coding,” Proc. 1996 IEEE Int.
vol. IT-20, pp. 284–287, Mar. 1974. Symp. Information Theory and Its Applications (ISITA’96), 1996.
[5] A. Lafourcade and A. Vardy, “Lower bounds on trellis complexity of [34] M. D. Trott, S. Benedetto, R. Garello, and M. Mondin, “Rotational in-
block codes,” IEEE Trans. Inform. Theory, vol. 41, pp. 1938–1954, Nov. variance of trellis codes. Part I: Encoders and precoders,” IEEE Trans.
1995. Inform. Theory, vol. 42, pp. 751–765, May 1996.
[6] , “Asymptotically good codes have infinite trellis complexity,” [35] D. Divsalar and F. Pollara, “Turbo Codes for Deep-Space Communica-
IEEE Trans. Inform. Theory, vol. 41, pp. 555–559, Mar. 1995. tions,” JPL, TDA Prog. Rep. 42-120, Feb. 1995.
[7] G. D. Forney Jr. and M. D. Trott, “The dynamics of linear codes over [36] S. Benedetto and G. Montorsi, “Unveiling turbo-codes: Some results
groups: State spaces, trellis diagram and canonical encoders,” IEEE on parallel concatenated coding schemes,” IEEE Trans. Inform. Theory,
Trans. Inform. Theory, vol. 39, pp. 1491–1514, Sept. 1993. vol. 42, pp. 409–428, Mar. 1996.
GARELLO et al.: INTERLEAVER PROPERTIES AND THEIR APPLICATIONS TO THE TRELLIS COMPLEXITY ANALYSIS OF TURBO CODES 807

[37] R. Garello, S. Benedetto, and P. Pierleoni, “Computing the free distance Sergio Benedetto (SM’90–F’97) received the
of turbo codes and serially concatenated codes with interleavers: Algo- Laurea in ingegneria elettronica (summa cum laude)
rithms and applications,” J. Select. Areas Commun., to be published. from Politecnico di Torino, Italy, in 1969.
From 1970 to 1979, he was with the Istituto di Elet-
tronica e Telecomunicazioni, first as a Research En-
gineer, then as an Associate Professor. In 1980, he
Roberto Garello (M’90) was born in Torino, Italy, was made a Professor in Radio Communications at
on December 18, 1965. He received the “Laurea in the Università di Bari. In 1981, he rejoined to Po-
Ingegneria Elettronica” degree (summa cum laude) in litecnico di Torino as a Professor of Data Transmis-
1990, and the Ph.D. degree in electronic engineering sion Theory in the Dipartimento di Elettronica. From
in 1994, both from Politecnico di Torino. 1980 to 1981, he spent nine months at the System
In 1993, he spent six months visiting the Signal Science Department of University of California, Los Angeles, as a Visiting Pro-
and Information Processing Laboratory, ETH, fessor, and three months at the University of Canterbury, New Zealand, as an
Zurich, Switzerland, and the Laboratory for Infor- Erskine Fellow. He has co-authored two books in signal theory and probability
mation and Decision Systems, MIT, Boston, MA. and random variables (in Italian), and the books “Digital Transmission Theory”
From October 1994 to August 1997, he was a Design (Englewood Cliffs, NJ: Prentice-Hall, 1987) and “Optical Fiber Communica-
Engineer at Marconi Communications, Genova, tions Systems” (Boston, MA: Artech House, 1996), as well as about 200 papers
working on high-rate all-digital modems and signal processing applications. for leading engineering journals and conferences. Active in the field of digital
From September 1997 to September 1998, he was with the Dipartimento di transmission systems since 1970, his current interests are in the field of optical
Elettronica of Politecnico di Torino working on the DSP implementation of a fiber communications systems, performance evaluation and simulation of dig-
turbo decoder. Since November 1998, he has been an Associate Professor at the ital communication systems, trellis-coded modulation and concatenated coding
Dipartimento di Elettronica ed Automatica of Università di Ancona, Italy. His schemes.
main interests include teaching, coding theory, and spread-spectrum systems. Dr. Benedetto is currently the Area Editor for Signal Design, Modulation and
Detection of the IEEE TRANSACTIONS ON COMMUNICATIONS.

Guido Montorsi (S’93–M’95) was born in Turin,

Italy, on January 1, 1965. He received the Laurea
degree in ingegneria elettronica in 1990 from
Politecnico di Torino, Turin, Italy, with a master’s
thesis concerning the study and design of coding Giovanni Cancellieri was born in Florence, Italy,
schemes for HDTV, developed at the RAI Research in 1952. He graduated from the University of
Center, Turin. In 1992, he spent the year as Visiting Bologna, Bologna, Italy, with a degree in electronic
Scholar in the Department of Electrical Engineering engineering and in physics.
at Rensselaer Polytechnic Institute, Troy, NY. In He developed research in telecommunication sys-
1994, he received the Ph.D. degree in telecommu- tems, with particular attention to networks and fibre
nications from the Dipartimento di Elettronica of optics components. In 1980, he joined the University
Politecnico di Torino. of Ancona, where he is currently Full Professor of
Since December 1997, he has been an Assistant Professor at the Politecnico di Telecommunications. He is co-author of more than
Torino. His current interests include the area of channel coding, particularly on 100 papers and nine books. He is a consultant for the
the analysis and design of concatenated coding schemes and study of interative Italian Ministry of Research for funding projects of
decoding strategies. advanced products and large enterprises.