1170
5, MAY 2002
Analysis and Design of OFDM/OQAM Systems
Based on Filterbank Theory
Pierre Siohan, Senior Member, IEEE, Cyrille Siclet, and Nicolas Lacaille
Abstract—A discrete-time analysis of the orthogonal frequency
division multiplex/offset QAM (OFDM/OQAM) multicarrier
modulation technique, leading to a modulated transmultiplexer, is
presented. The conditions of discrete orthogonality are established
with respect to the polyphase components of the OFDM/OQAM
prototype filter, which is assumed to be symmetrical and with arbi-
trary length. Fast implementation schemes of the OFDM/OQAM
modulator and demodulator are provided, which are based on the
inverse fast Fourier transform. Non-orthogonal prototypes create
intersymbol and interchannel interferences (ISI and ICI) that,
in the case of a distortion-free transmission, are expressed by a
closed-form expression. A large set of design examples is presented
for OFDM/OQAM systems with a number of subcarriers going
from four up to 2048, which also allows a comparison between
different approaches to get well-localized prototypes.
Index Terms—Filterbanks, multicarrier modulation, polyphase
representation, pulse shaping, transmultiplexer.
I. INTRODUCTION
T HE principle of multicarrier modulation (MCM) consists
of splitting up a wideband signal at a high symbol rate into
several lower rate signals, each one occupying a narrower band.
A generally recognized advantage of MCM is its robustness
against different types of channel distortions, such as multi-
path propagation and narrowband interference. Orthogonal
frequency division multiplex (OFDM) is certainly, until now,
the most important class of MCM. The OFDM acronym often
recovers two different types of modulation. In the first one, as
proposed, for instance, in [1], each carrier is modulated using
quadrature amplitude modulation (QAM). In this scheme,
which is also called OFDM/QAM [2], QAM symbols are
shaped with a rectangular window. In a second category of
OFDM systems, which is also called orthogonally multiplexed
QAM in [3] (O-QAM) or OFDM with offset QAM in [2]
(OFDM/OQAM), the modulation used for each subcarrier is a
staggered offset QAM (OQAM). Both the OFDM/QAM and
OFDM/OQAM modulation schemes theoretically guarantee
orthogonality and a maximum and identical spectral efficiency.
Furthermore, in practice, they can be implemented thanks to
Manuscript received April 27, 2000; revised January 10, 2002. The associate
editor coordinating the review of this paper and approving it for publication was
Prof. Gregori Vazquez.
P. Siohan was with France Télécom R&D, DMR/DDH, Cesson-Sévigné
Cedex, France. He is now with IRISA-INRIA, Campus de Beaulieu, Rennes
Cedex, France.
C. Siclet is with the France Télécom R&D, DMR/DDH, Cesson-Sévigné
Cedex, France.
N. Lacaille was with France Télécom R&D, DMR/DDH, Cesson-Sévigné
Cedex, France. He is now with l’Université de Technologie de Belfort-Mont-
béliard, Belfort Cedex, France.
Publisher Item Identifier S 1053-587X(02)03151-3.
1053-587X/02$17.00 © 2002 IEEE
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO.
the discrete Fourier transform (DFT). An important difference
comes from the fact that OFDM/OQAM, unlike OFDM/QAM,
allows the introduction of an efficient pulse shaping, which
makes it less sensitive to the frequency offset due to the trans-
mission channel and to the receiver. Therefore, if OFDM/QAM
constitutes the modulation kernel of the famous so-called coded
OFDM (COFDM) system, OFDM/OQAM is now presented as
being a good candidate to get still higher bit rates over wireless
channels [2].
Based on the earliest works in the field; see, for example,
[3]–[5], the orthogonality of OFDM/OQAM has been presented
for some time, only using continuous pulses of infinite length,
as, for instance, the well-known raised-cosine function. The im-
portance of getting an orthogonal pulse-shaping satisfying given
criteria has only appeared recently. However, most of the solu-
tions already proposed do not take into account the truncation
and discretization effects of the pulse shape, i.e., a natural con-
sequence of DFT-based implementations. These effects imply a
loss of orthogonality, i.e., an intersymbol interference (ISI) and
an interchannel interference (ICI), which is implementation de-
pendent and arises even for distortion-free channels.
In [6], the authors take into account the truncation effect and
choose to minimize the out-of-band energy of a time-limited
pulse shape. Once again, the discretization step of the resulting
pulse shape is ignored, and this only leads to an approximation
of the discrete orthogonality conditions. In the case of wireless
transmission, due to the time and frequency dispersion created
by this type of channel, it seems, however, more appropriate to
look for well-localized pulse shapes [2], [7]. Independently of
the present work, some authors [8] have recently proposed to
discretize orthogonality conditions given in continuous-time [6]
in order to get discrete orthogonality conditions. The evalua-
tion in [8] is also based on the time-frequency localization cri-
terion. In this paper, our approach is different and more directly
connected to filterbank formalism, as presented, for instance, in
transmultiplexer theory [9]. We start by discretizing the basic
definition equation of OFDM/OQAM, and afterwards, we de-
rive the theoretical properties, the implementation scheme, and
the performances of the system. We represent all signals by their
complex envelope, and we assume, as usual, that the prototype
filter is real-valued and symmetrical. In this way, we are able to
get the following:
• a description of OFDM/OQAM systems being equiva-
lent to a modulated transmultiplexer, given in Section II,
leading to efficient implementations based on inverse fast
Fourier transform (IFFT) and presented in Section III;
• the discrete orthogonality conditions with respect to the
polyphase components of a causal finite impulse response
SIOHAN et al.: ANALYSIS AND DESIGN OF OFDM/OQAM SYSTEMS 1171

(FIR) prototype filter of arbitrary length, which are de- pulse shape that can be different from the rectangular window.
rived in Section III; At the demodulation side, the symbols are estimated by
• an analytical expression of the overall distortion due to
the ISI and to the interchannel interference ICI created by
imperfect filterbanks in distortion-free channels, which is
derived in Section IV; (2)
• several design examples presenting a large set of orthog-
onal and nearly orthogonal prototype filters with different
pulse shapes. Design charts, tables, and graphical displays
are provided in Section V, allowing different comparisons (3)
in terms of orthogonality and time–frequency localization.
Notations: denote the set of integer and real num-
bers, respectively. correspond to the space of
square-summable continuous and discrete-time functions, (4)
respectively. Vectors and matrices are denoted with bold italic
letters, for instance, . Superscript stands for transposition.
We denote discrete filters of with lower case letters, for in-
stance, , and their -transform with upper case letters, such (5)
as . and designate the real part and imaginary
part of the complex-valued number , respectively. Similarly, In order to get simplified, easier-to-manipulate expressions, the
for a filter , we get and following notations can be introduced:
. Superscript denotes complex
conjugation. For a filter . The
(6)
tilde notation denotes paraconjugation: .
is the norm of .

(7)
II. OFDM SYSTEMS USING OFFSET-QAM (OFDM/OQAM)
In this section, we recall the classical continuous-time for- We also denote . Then, (1) –(5) are rewritten as
mulation of OFDM/OQAM, and the discrete-time formulation
is then derived. We also introduce the notations that allow clas- (8)
sical multirate relations of the filterbank theory [10] to be re-
covered in discrete time.

A. Continuous-Time Formulation (9)

The principle of OFDM/OQAM is the transmission of offset Another way to formulate the OFDM/OQAM continuous-time
QAM symbols rather than QAM symbols. Then, the orthog- transmitted signal is to write it as an expansion over some basis
onality can be maintained with pulse shapes being different functions with coefficients
from the rectangular window [4]. The second specificity of
this scheme is that considering two successive carriers, the
time offset is introduced onto the imaginary part of the QAM (10)
symbols on one of the carriers, whereas it is introduced onto
the real part of the symbols on the other one. The number where is defined by
of carriers, which is denoted , is also assumed to be even
, and the continuous-time baseband transmitted (11)
signal writes [11] Then, (9) is rewritten as

(12)
Assuming a distortion-free channel, the transmitted symbols are
perfectly recovered, i.e., , if and only if the set of
(1) functions constitutes an orthonormal basis of its span, i.e.,

where is the signaling interval, the spacing be-

tween two successive carriers, and the real and imag-
inary parts, respectively, of the QAM complex-valued symbols
(13)
we want to transmit, and a symmetrical real-valued
1172 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO. 5, MAY 2002

where is the Kronecker delta. In particular, we get Then, assuming a distortion-free channel, if
. It is worthwhile mentioning that these conditions are and only if the set of functions constitutes an orthonormal
unchanged if is just defined modulo . Thus, it can be basis of its span, i.e.,
assumed, as, for example, in [2], that .

B. Discrete-Time Formulation
In the previous section, we considered continuous-time
signals in with its classical real-valued inner product (20)
. Now, our goal is to obtain a discrete-time signal
in with its classical real-valued inner product ,
which is defined by C. OFDM/OQAM Transmultiplexer
Equation (17) looks like the expression of a signal obtained at
the output of a synthesis bank of a filterbank with sub-bands
and with an expansion factor equal to on each sub-band.
(14) Indeed, if and denote the input signal and the filter
on the th subband of this synthesis bank, respectively, then the
Since the duration corresponds to the transmission of output signal is written as [10, p. 117]
complex-valued symbols, the critical sampling period is defined
by (21)

(15) This exactly corresponds to (17) with and

This means that is critically sampled, and therefore, the

spectral efficiency is still maximum. In order to get a causal dis- (22)
crete-time prototype with length equal to is trun- and with
cated to the interval and is delayed by (23)
time units. Moreover, in order to get a proto-
type filter with norm still approximately equal to We can remark that can be defined by (23) since has
has to be normalized with a multiplicative factor equal to just to be equal to modulo . Similarly, (19) looks
like the expression of the output symbols of an analysis bank of
(16) a filterbank with sub-bands and with a decimation factor
equal to on each sub-band. Thus, if denotes the input
Thus, based on (8) and using the fact that and signal of this bank and and denote the filter and
that , the baseband transmitted discrete-time the output signal on the th subband, respectively, then [10, p.
OFDM/OQAM signal is such that 117]

(24)

Using the fact that , let us rewrite (19) in

order to illustrate the analogy with (24):
(17)

with

(18)

At the demodulation stage, we get an estimate of the transmitted

real-valued symbol using the real-valued inner product of
and
(25)

Let us decompose with two integers and ( and

) by
(26)
(19)
Then, in order to get a causal system, (25) is rewritten, making
SIOHAN et al.: ANALYSIS AND DESIGN OF OFDM/OQAM SYSTEMS 1173

Fig. 1. Transmultiplexer model of OFDM/OQAM modulations.

the reconstruction delay appear as trary length , which is not considered in the
works cited previously.

III. DISCRETE ORTHOGONALITY FOR OFDM/OQAM

The orthogonality conditions of the modulated transmulti-
plexer we have just obtained naturally depend on the analysis
and synthesis filterbanks, i.e., , respectively.
To get a compact representation of this system, we use the
polyphase approach, which leads us to IFFT-based implemen-
tations, to the input–output relation and, finally, allows us to
(27) get the mathematical orthogonality conditions.

A. Polyphase Approach
It appears that (27) corresponds to (24) with
The analysis and synthesis filterbanks in Fig. 1 can be
and
expressed as functions of the prototype filter or of its
(28) -transform . Let us decompose as a function of its
polyphase components of order [10]

Then, by defining and , which are the -trans- with

forms of and , respectively, we get the transmulti-
plexer scheme depicted in Fig. 1. In this way, we get a transmul- (29)
tiplexer with the following particularities.
• The input and output signals are real-valued symbols. As is real-valued and symmetrical, using (22) and (28),
• The number of channels is twice the ex- and can be written as
pansion and decimation factors , whereas for clas-
sical transmultiplexers that are critically decimated [10]
or oversampled [12], we necessarily have or (30)
, respectively. Exact reconstruction in this case
is nevertheless made possible by the fact that the input where .Wecanthenrewrite and
symbols are real-valued; therefore, the present transmulti- asfunctionsof andderive thepolyphasematricesof thesyn-
plexer is equivalent to a critically decimated one with thesis(modulator)andanalysis(demodulator)parts,whicharede-
complex-valued inputs [cf. (15)]. noted (type 2) and (type 1) [10], respectively:
• Before expansion by a factor , each real-valued input is
premodulated, whereas at the receiver side, the dual oper-
ation occurs after decimation, the real part being extracted
afterwards (cf. Fig. 1).
• A delay , with depending on the length of the pro- (31)
totype filter, has to be included either at the transmitter
output or at the receiver input.
Compared with previous schemes proposed for orthogonal
MCM systems (see, for instance, [11] and [13]), our imple-
mentation scheme involves two integer parameters and that (32)
allow us to handle the case of causal prototype filters with arbi-
1174 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO. 5, MAY 2002

Fig. 2. General polyphase representation of the OFDM/OQAM transmultiplexer.

Fig. 3. OFDM/OQAM modulator realized with an IFFT.

Using these notations and the Noble identities [10, p. 119], where is the antidiagonal matrix. Therefore, we
the transmultiplexer, as depicted in Fig. 1, is equivalent to its can directly derive from Fig. 2 the IFFT-based modulator and
polyphase representation shown in Fig. 2. Let demodulator schemes depicted in Figs. 3 and 4, respectively.
with a -transform . Then, denotes the Furthermore, Figs. 3 and 4 only represent two basic schemes
-transform of . Besides, at the receiver side, that can be further improved, as shown in Appendix A.
denotes the -transform of with .
C. Input–Output Relation
B. IFFT-Based Implementations Let us now introduce a matrix notation where
Direct implementation of the systems depicted in Figs. 1 and and are column vectors with elements
2 would be very costly. However, as the modulator and demod- and , respec-
ulator are related to a modulated transform, we examine, in this tively. is the -transform of . Then, it
section, how it is possible to get an efficient IFFT-based imple- can be seen from the scheme depicted in Fig. 2 that we have
mentation. the relation
In order to get the matrices and as functions of (35)
, we now introduce the matrices and ,
which are given by where denotes the transfer function of the section in-
cluding the expanders, delay lines, and decimators. Hence
diag
with
diag (33)
(36)
Using these latter notations and (31) and (32), we get Thus, from (34) and (36), we get
with
(34) (37)
SIOHAN et al.: ANALYSIS AND DESIGN OF OFDM/OQAM SYSTEMS 1175

Fig. 4. OFDM/OQAM demodulator realized with an IFFT.

After some computation reported in Appendix B, it can be It can be noted that with this latter expression, we recover
shown that is as in (38), shown at the bottom of the page, a PR condition that is identical to the ones already obtained at
where first for PR cosine-modulated filterbanks [10, p. 379], [14] and
afterwards, with exception of the normalization constant, for the
PR orthogonal MDFT filterbanks [15].1
It is worthwhile mentioning that using the analogy with filter-
(39) banks again, each pair of polyphase filters
satisfying (40), for , can be implemented in
Figs. 3 and 4 by means of a cascaded lattice that structurally en-
D. Mathematical Expression of Orthogonality sures a perfect orthogonality [10, pp. 380–383], [14].
Perfect orthogonality is provided when there is no ISI, i.e.,
, nor ICI, i.e., . IV. ISI AND ICI FOR A DISTORTION-FREE CHANNEL
Using filterbank terminology, we may also say that the orthog- In this section, we give, in the case of a distortion-free
onal transmultiplexer checks the perfect reconstruction (PR) channel, explicit expressions of the ISI and ICI terms when
property with the delay if and only if , where using OFDM/OQAM with a prototype filter that is not perfectly
is the identity matrix. In Appendix C, we derive orthogonal.
an equivalent formulation of this latter equality, which leads A modulation of each subcarrier using OQAM- , i.e., a
us to the following theorem showing how the conditions of states modulation, implies that each real-valued symbol
perfect orthogonality can be directly related to the polyphase , coded with bits, is such that ,
components of the prototype filter. where is a fixed amplitude, and is an integer in the range
Theorem 1: Let be a symmetrical real-valued prototype .
filter that is assumed to be causal and time-limited with arbi- In practice, most often, prototype functions for communi-
trary length , and let be its type cation systems are obtained by truncation and discretization
I polyphase components. Then, a sub-band OFDM/OQAM of continuous-time functions that are no longer orthogonal
signal, using this prototype, can be transmitted through a dis- in discrete time. Then, as (40) is not exactly satisfied, these
tortion-free channel and reconstructed without ISI nor ICI, with nonorthogonal prototypes create ISI, i.e., interference between
the delay denoting the integer part func- symbols inside a given subband, and ICI, i.e., interference
tion by superior value if and only if between symbols belonging to different subbands. Using (36)
1In that way, this result is an illustration of the equivalence between these two
(40) aspects of multirate filterbanks.

.. .. ..
. . .
.. .. .. (38)
. . .
..
.
1176 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO. 5, MAY 2002

and (38), it can be seen that each component of seen in several publications related to OFDM/OQAM over wire-
is such that less channels [2], [7], [8], [16]. In this paper, in order to assess
our results on a solid basis, we will use two different localiza-
tion measures.
Let be a discrete-time signal with finite energy, i.e.,
, and with Fourier transform denoted . Its second-
(41) order moment in time, which is denoted , and in frequency,
which is denoted , are given by
where is the integer part function. Therefore, in the time
domain, we get (42), shown at the bottom of the page.
The total distortion is then given by .
For a -OQAM, we have , and thus,
the maximum distortion is (45)

where and minimize and , respectively. It can be

shown that

(43)
(46)
Moreover, using (39), we notice that Moreover, for a real-valued signal, mod .
so that It can be deduced from [17] that if , we can
define a localization measure denoted such that
(44)
(47)

where is the unreachable optimum. This

V. DESIGN RESULTS inequality related to the uncertainty principle can be applied in
The quality of a given prototype filter can be evaluated by particular when is symmetrical with even length. In order to
means of the resulting ISI and ICI and, generally, by its ability avoid the restriction mentioned above, we can use, instead of
to satisfy a given criterion. In this section, two situations are con- (45), the definition of second-order moments suggested in [18]
sidered: a first one where the prototype is obtained by truncation by Doroslovac̆ki
and discretization of orthogonal functions in continuous-time
and is therefore not perfectly orthogonal and a second one where
we design it in order to exactly satisfy the orthogonality condi-
tions. In both cases, the time–frequency localization is used as (48)
a criterion either for evaluation, as in Section V-B, or for opti-
mization, as in Section V-C.
(49)
A. Time–Frequency Localization Measures
It is well known that the wireless channel introduces time and (50)
frequency dispersion. To limit the resulting distortion, a solu-
tion proposed in [2] is to use basic signals being localized in
time and frequency with the same time–frequency scale as the (51)
channel itself. In this context, time and frequency dimensions
are equally important. Indeed, time–frequency localization is where and are the gravity centers in time and frequency,
now often considered as being the right criterion, as it can be respectively, which minimize and , which are second-

(42)
SIOHAN et al.: ANALYSIS AND DESIGN OF OFDM/OQAM SYSTEMS 1177

order moments in time and frequency, respectively. It can be shown at the bottom of the page, where is the roll-off param-
shown that eter . Therefore, the continuous-time expression of
SRRC functions is given by
mod (52)

(56)

(53) As for OFDPs, their expression in continuous-time is as fol-

lows [6], [20]:

Then, we obtain a second criterion, which is denoted

(57)
(54)

The optimum is still . where is the th prolate spheroidal wave function trun-
cated to the interval and where are
B. Design Examples Based on Continuous-Time Orthogonal some real-valued coefficients obtained thanks to an optimiza-
Functions tion procedure [20].
Last, the EGFs are defined by [21]
By truncation and sampling of a prototype function satis-
fying orthogonality conditions in continuous time, it is possible
to get a discrete prototype being nearly orthogonal, as long as
its length is sufficient. In our simulations, we have compared
three types of functions that, initially, were defined in contin- (58)
uous-time:
1) the well-known square root raised cosine (SRRC) func- where is a real-valued number, are some
tion (cf. for instance, [19]); real-valued coefficients, and is the Gaussian function
2) a family of pulses introduced by Vahlin and Holte in .
[20], which we will call optimal finite duration pulses For OFDM/OQAM signals, we suppose that
(OFDPs); , and in order to satisfy (13), we also suppose that
3) a new family of functions called extended Gaussian func- is such that . Given that
tions (EGFs); see [21]. , the corresponding filters are multiplied by a
In all cases, the truncation limits the pulse length to normalization factor equal to . The discrete-time version
and the discretization is carried out with a sampling period of the SRRC, OFDP, and EGF prototype filters are then obtained
. These three families are interesting from by truncation to the interval (except for
different points of view. Thus, SRRC functions mainly have OFDPs, which no longer need to be truncated) and thanks to
a historical interest, and we only consider them in order to (16).
make a comparison with some well-known functions. The Besides, when setting for the EGF, as
OFDPs have two important properties. The first is that they are in [2] and [21], has to be approximately within the
orthonormal with a finite duration so that they do not need to range to get the best time–frequency localization measures [22].
be truncated. The second is that they are optimized in order to In Table I, we have compared, for , the maximum dis-
get as much of the pulse energy as possible within a frequency tortion [cf. (44)] obtained for a few EGF and SRRC func-
band . Finally, an important property of the tions. They show that for a given length, except for and
EGF is the fact that they can also lead to an orthonormal basis , the EGF clearly outperform the best SRRC func-
after discretization and that, furthermore, they can be very tion. Some other simulations carried out with and
close to the achievable optimum bound in time–frequency give the results reported in Figs. 5 and 6. They lead us to the
localization [2], [22]. Nevertheless, none of these pulses keep conclusion that for any value of and a given ratio, the
strict orthonormality after both operations of truncation and EGF can probably provide better results than the SRRC func-
discretization. tions. They also show that in both cases, the distortion mainly
First, let us give the continuous-time expressions of SRRC depends on the ratio.
functions OFDP and EGF. For a transmission rate equal to , We can deduce from a result previously obtained for cosine-
the frequency expression of the SRRC function is as in (55), modulated filterbanks [22] that for OFDM/OQAM systems, the

(55)
1178 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO. 5, MAY 2002

TABLE I of the coefficients given in [20]. This table shows that the
MAXIMUM DISTORTION (D ) FOR SRRC(r ) AND EGF() WITH
distortion of OFDPs slightly increases when the number of
DIFFERENT VALUES OF r AND ,AND FOR M = 4
carriers increases beyond . It also shows that OFDPs
have a smaller distortion than SRRCs for the same
ratio. Moreover, the distortion of OFDPs is also smaller than
the distortion of EGFs for and . Finally, for
and , OFDPs are between the EGFs
obtained for and .

C. Design Examples Satisfying the Discrete Orthogonality

Conditions
From Table I and Figs. 5 and 6 and as illustrated with pre-
vious publications (see, for instance, [23]), it also clearly ap-
pears that to get low distortion values, high ratios may
be required. Furthermore, the truncation and sampling opera-
tions do not allow direct control of the time–frequency localiza-
tion. On the other hand, if we choose to directly optimize the
prototype filter to get a maximum time–frequency localization
under the PR constraint (40), we are, at least, ensured to get
neither ISI nor ICI distortion. This optimization can be directly
conducted by optimizing the prototype coefficients under
the constraint (40), or it can be also carried out using a set of lat-
tice coefficients that corresponds to an efficient implementation
Fig. 5. D values for SRRC prototype functions with r = 1 and M = 4
(dashed line), M = 8 (dotted line), and M = 16 (solid line). of each pair of polyphase components for
, [10, pp. 380–383], [22]. We have chosen
this second approach, which seems preferable, since it leads
to an unconstrained optimization problem with, furthermore, a
number of variables that are approximately reduced by a factor
of 2.
Let us refer to the optimization method of the lattice coeffi-
cients according to the criterion as the optimized lattice
(OL), knowing naturally that more generally, the optimization
of the lattice coefficients can be carried out with different types
of criteria, including, as in [20], the maximization of the in-band
energy. We can compare the OL method with the truncation and
discretization method applied to the EGF, OFDP, and SRRC
functions. We have chosen the parameter value, providing, in
Fig. 6. Estimated (lines) and experimental (dots) D values for EGF general, the best localization measures, i.e., for the EGF
prototype with  = 0:5 (solid line),  = 1 (dashed line), and  = 1:5 [21], [22] and for the SRRC functions. As stated previ-
2
(dashdot line) for M = 4 (o); M = 8 (+) and M = 16 ( ).
ously, the OL method is the only one that provides a perfect
orthogonality, i.e., a distortion theoretically equal to 0
maximum distortion related to discrete EGF prototypes can be but, in practice, due to numerical inaccuracies, around
estimated by [16] or for double precision computers. Furthermore, as our
results reported in Table III show, it appears that for the two
(59)
measures used ( and ), despite the introduction of the PR
Our results reported in Fig. 6 give a good illustration of the constraint, we have the following.
accuracy of this estimate for • Compared with the SRRC function, the OL method al-
and . ways provides better localized prototype filters, even for a
Concerning the SRRC function, it can be seen in Fig. 5, with smaller length than for SRRC filters.
a roll-off value , that an increase in the ratio also • Compared with the OFDP, OL filters have a better local-
tends to reduce the distortion. However, it is also clear that this ization measure for a given length. This is not surprising
decrease is not as fast nor as regular as with the EGF. since OFDPs are not optimized according to the maxi-
In Table II, we have reported for different values of mization of the time–frequency localization.
the maximum distortion related to discrete-time OFDPs with • Compared with the EGF, we can see that the OL method
time duration equal to and , which correspond generally provides better results. This is particularly true
to and , respectively. for the shortest filter lengths ( , i.e., a pulse length
These prototype filters have been computed thanks to the values limited to the duration of one complex-valued symbol).
SIOHAN et al.: ANALYSIS AND DESIGN OF OFDM/OQAM SYSTEMS 1179

TABLE II
MAXIMUM DISTORTION (D ) FOR THE VAHLIN PULSES

TABLE III
LOCALIZATION MEASURES OBTAINED FOR DISCRETE SRRC(r = 1), DISCRETE EGF( = 1), DISCRETE OFDP, AND FOR THE OL METHOD

Fig. 7. SRRC function with r = 1 for M = 1024 and L = 2048;  = Fig. 8. EGF with = 1 for M = 1024 and L = 2048;  =
0:0830; D 0
= 7:145e 02. 0:1442;D 0
= 2:846e 02.

Some exceptions occur for the highest values of for

which it may be possible that our optimization procedure
found local optima only.
Nevertheless, if one wants, for simplicity, to use analytical
expressions, it is also clear that EGF have to be preferred to the
OFDP and the SRRC functions to get well-localized prototype
filters.
As for MCM systems, the number of subcarriers is relatively
high (generally greater than 64 and sometimes up to 8192 [24]), Fig. 9. OL for M = 1024 and L = 2048;  = 0:9049; D =
and as the computational complexity is more often a crucial 7:002e 15.0
factor, the significant improvements we get for prototype filters
limited to one signaling interval are of a particular practical
interest.
Figs. 7–10 show time and frequency representations obtained
for OFDM/OQAM systems with . The comparison
for between the SRRC function, the EGF, and the OL
method can be made with Figs. 7–9. In Fig. 10, we can compare
the time and frequency responses of the EGF, with , and
of the OFDP, for . It is clear that the OFDP that is
optimized to minimize the out-of-band energy yields a better
Fig. 10. EGF (dashed line,  = 1;  = 0:9767;D = 1:476e 03) 0
result in the frequency domain. On the other hand, the EGF, and OFDP (solid line,  = 0:9328;D 0
= 5:399e 03) for M = 1024
which has a better localization in time, finally provides, with and L = 8192.
1180 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO. 5, MAY 2002

, a better time–frequency localization measure • If is even, , and for

than the OFDP with .
Concerning the comparison of the impulse responses for the
prototypes of length 2048 (see Figs. 7–9) with the ones with
length 8192 (Fig. 10), one has to take care of the different time (63)
scales used for the graphical displays.
Moreover, the FFT in (63) is an FFT with real-valued in-
puts; therefore, it can be carried out thanks to an FFT with
VI. CONCLUSION
complex-valued inputs, instead of real-valued in-
We have presented an application of the filterbank theory to puts [25, p. 82].
the analysis and design of OFDM/OQAM systems. In a first • If is odd, , and
step, we have derived the transmultiplexer scheme that can be
associated with OFDM/OQAM systems with symmetrical FIR (64)
prototypes of arbitrary length. Then, using a polyphase decom-
position of this prototype, we have proposed fast implementa- Equation (64) corresponds to an odd DFT with real-valued
tion schemes of the OFDM/OQAM, modulator, and demodu- inputs that can also be realized thanks to an FFT with
lator, which are based on the IFFT, and we have also found the complex-valued inputs instead of real-valued inputs
discrete orthogonality conditions related to the polyphase com- [25, p. 83]. Thus, as in [23], it is possible to use only an
ponents of this prototype. Systems that do not exactly satisfy the FFT of size .
discrete orthogonality conditions create ISI and ICI. For such Similarly, at the demodulation part, the IFFT followed by a ro-
nonorthogonal systems, and assuming a distortion-free trans- tation by is also equivalent to an FFT
mission channel, we have derived the analytical expressions of with a permutation of the inputs . Then, we will only
the ISI and ICI distortions created by imperfect filterbanks. Fi- keep the real part of the outputs of the FFT so that further re-
nally, we have provided a set of design examples that clearly duction of the complexity is also possible.
show that the numerical optimization of a discrete orthogonal
prototype has, at least, two important advantages compared with APPENDIX B
truncation and sampling of prototypes being orthogonal in con- COMPUTATION OF THE TRANSFER MATRIX
tinuous time. Indeed, the direct design of orthogonal prototypes
in discrete time not only avoid ISI and ICI distortions for distor- In Section III, we have shown that the transfer matrix
tion-free channels, but they also allow us to obtain short proto- can be written
type filters with very good localization measures. This naturally (65)
tends to significantly reduce the computational cost of imple-
mentation. where
(66)
APPENDIX A Toevaluate ,letusfirstnotethatadelay placedbetween
REDUCTION OF THE COMPLEXITY OF THE IFFT-BASED an expansion and a decimation by leads to a transfer function
IMPLEMENTATIONS such that , with if is a mul-
It can be seen from Fig. 3 that for tiple of and otherwise. Thus, we obtain
, and from (66), we get
. There-
(60)
fore, the entries of matrix are

Let us define two integers and by

(67)
or (61)

Let us also denote In order to only retain the nonzero terms in (67), we now define
if
if (68)
(62) otherwise
if
and for , we also introduce (69), shown at the bottom of
Then, two different cases appear. the page. Therefore, the entries in matrix can be rewritten as

if
(69)
if
SIOHAN et al.: ANALYSIS AND DESIGN OF OFDM/OQAM SYSTEMS 1181

shown in the first equation at the bottom of the page, where

(70)

(71)
and finally
Moreover, using (72), shown at the bottom of the page, we ob-
tain from (70)–(72) that

(73)

We can now notice that the argument of in (73) is real if

and are even; otherwise, it is purely imaginary. Thus,
being given, defining as the unique value of , which has
the same evenness as , we obtain (74), shown at the bottom
of the page. Using the decomposition , the
expression of can be simplified as follows when
is even:

(75)

Furthermore, it can be checked by inspection from (29) and (69)

that for . Thus,
we finally obtain the expression of the transfer matrix given by
(38) and (39).

for
(72)
for

if is even
(74)
otherwise
1182 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 50, NO. 5, MAY 2002

APPENDIX C the necessary and sufficient conditions of orthogonality on the

ORTHOGONALITY CONDITIONS polyphase components
The matrix given by (65) can also be rewritten as
for
(81)

ACKNOWLEDGMENT

Then, the orthogonality conditions can be derived from the The authors would like to thank the anonymous reviewers for
equality or equivalently, from the following valuable comments that have led to improvements in this paper
relation: and Dr. Vahlin, who provided data files allowing to include its
own prototype function, i.e., the OFDP, in our comparisons.

REFERENCES
[1] S. B. Weinstein and P. M. Ebert, “Data transmission by frequency-di-
vision multiplexing using the discrete Fourier transform,” IEEE Trans.
This latter expression is also equivalent to Commun. Technol., vol. COM-19, pp. 628–634, Oct. 1971.
[2] B. Le Floch, M. Alard, and C. Berrou, “Coded orthogonal frequency
division multiplex,” Proc. IEEE, vol. 83, pp. 982–996, June 1995.
[3] B. Hirosaki, “An orthogonally multiplexed QAM system using the dis-
crete Fourier transform,” IEEE Trans. Commun., vol. 29, pp. 982–989,
July 1981.
[4] R. W. Chang, “Synthesis of band-limited orthogonal signals for multi-
channel data transmission,” Bell Syst. Tech. J., vol. 45, pp. 1775–1796,
Dec. 1966.
[5] B. R. Saltzberg, “Performance of an efficient parallel data transmission
system,” IEEE Trans. Commun. Technol., vol. 15, pp. 805–811, Dec.
1967.
[6] A. Vahlin and N. Holte, “Optimal finite duration pulses for OFDM,”
IEEE Trans. Commun., vol. 44, pp. 10–14, Jan. 1996.
[7] R. Haas and J.-C. Belfiore, “A time-frequency well-localized pulse for
multiple carrier transmission,” Wireless Pers. Commun., vol. 5, pp. 1–18,
1997.
[8] H. Bölcskei, P. Duhamel, and R. Hleiss, “Design of pulse shaping
for (76) OFDM/OQAM systems for high data-rate transmission over wireless
channels,” in IEEE Int. Conf. Commun., Vancouver, BC, Canada, June
Then, using (68) and (69), (76) becomes 1999.
[9] A. N. Akansu, P. Duhamel, X. Lin, and M. de Courville, “Orthogonal
transmultiplexers in communication: A review,” IEEE Trans. Signal
Processing, vol. 46, pp. 979–995, Apr. 1998.
[10] P. P. Vaidyanathan, Multirate Systems and Filter Banks. Englewood
Cliffs, NJ: Prentice-Hall, 1993.
[11] B. Hirosaki, S. Hasegawa, and A. Sabato, “Advanced groupband data
(77) modem using orthogonally multiplexed QAM technique,” IEEE Trans.
Commun., vol. COMM-34, pp. 587–592, June 1986.
[12] R. Hleiss, P. Duhamel, and M. Charbit, “Oversampled OFDM systems,”
in IEEE Int. Conf. Digital Signal Process., Greece, July 1997.
This relationship is always true for . Moreover, [13] N. J. Fliege, “Orthogonal multiple carrier data transmission,” Eur. Trans.
and lead to the same relationship; thus Telecommun., vol. 3, no. 3, pp. 255–264, May–June 1992.
[14] T. Q. Nguyen and R. D. Koilpillai, “The theory and design of arbitrary-
length cosine-modulated filter banks and wavelets, satisfying perfect
reconstruction,” IEEE Trans. Signal Processing, vol. 44, pp. 473–483,
Mar. 1996.
[15] T. Karp and N. J. Fliege, “MDFT filter banks with perfect reconstruc-
tion,” in Proc. IEEE Int. Symp. Circuits Syst., vol. 1, Seattle, WA, Apr.
(78) 1995, pp. 744–747.
[16] P. Siohan and N. Lacaille, “Analysis of OFDM/OQAM systems based
on the filterbank theory,” in IEEE Global Commun. Conf., Brazil, Dec.
1999.
[17] L. C. Calvez and P. Vilbé, “On the uncertainty principle in discrete sig-
nals,” IEEE Trans. Circuits Syst. II, vol. 39, pp. 394–395, June 1992.
[18] M. I. Doroslovac̆ki, “Product of second moments in time and frequency
for discrete-time signals and the uncertainty limit,” Signal Process., vol.
(79) 67, no. 1, pp. 59–76, May 1998.
[19] P. R. Chevillat and G. Ungerboeck, “Optimum FIR transmitter and re-
ceiver filters for data transmission over band-limited channels,” IEEE
Trans. Commun., vol. COMM-30, pp. 1909–1915, Aug. 1982.
(80) [20] A. Vahlin and N. Holte, “Optimal finite duration pulses for OFDM,” in
Proc. IEEE Global Commun. Conf., San Francisco, CA, Nov. 1994, pp.
where is the unique value of that has the same even- 258–262.
[21] C. Roche and P. Siohan, “A family of extended Gaussian functions with
ness as . Moreover, as in Appendix B, using the fact that a nearly optimal localization property,” in Proc. Int. Workshop Multi-
for , we obtain Carrier Spread-Spectrum, Germany, Apr. 1997, pp. 179–186.
SIOHAN et al.: ANALYSIS AND DESIGN OF OFDM/OQAM SYSTEMS
[22] P. Siohan and C. Roche, “Cosine-modulated filterbanks based on ex-
tended gaussian functions,” IEEE Trans. Signal Processing, vol. 48, pp.
3052–3061, Nov. 2000.
[23] G. Cariolaro and F. C. Vagliani, “An OFDM scheme with a half com-
plexity,” IEEE J. Select. Areas Commun., vol. 13, pp. 1586–1599, Dec.
1995.
[24] U. Reimers, “Digital video broadcasting,” IEEE Commun. Mag., pp.
104–110, June 1998.
[25] M. Bellanger, Digital Processing of Signals: Theory and Prac-
tice. New York: Wiley, 1989.
Pierre Siohan (M’94–SM’99) was born in Camlez,
France, in October 1949. He received the Doctorat
degree from the École Nationale Supérieure des
Télécommunications (ENST), Paris, France, in 1989
and the Habilitation degree from the University of
Rennes, Rennes, France, in 1995.
In 1977, he joined the Centre Commun d’Études
de Télédiffusion et Télécommunications (CCETT),
Rennes, where his activities were first concerned
with the communication theory and its application to
the design of broadcasting systems. Between 1984
and 1997, he was in charge of the Mathematical and Digital Signal Processing
Group. From September 1997 to August 2001, he was an Expert Member with
the Terrestrial Broadcasting and Radio Access Networks Laboratory: one of
the R&D laboratories of France Télécom. He is currently Director of Research
with the Institut National de Recherche en Informatique et Automatique
(INRIA), Rennes. His current research interest are in the areas of filterbank
design for communication systems and joint source-channel coding.
1183
Cyrille Siclet was born in Champigny-sur-Marne,
France, in September 1976. He received the engineer
degree in telecommunications from the École
Nationale Supérieure des Télécommunications de
Bretagne (ENST Bretagne), Bretagne, France, and
the D.E.A. degree from the University of Rennes,
Rennes, France, in 1999. He is pursuing the Ph.D.
degree at the University of Rennes.
Since March 1999, he has been working for France
Télécom R&D, Rennes.
Nicolas Lacaille was born in Paris, France, in
1972. He received the engineer degree in electronic
engineering from the École Supérieure d’Ingénieurs
en Electrotechnique et Electronique, Paris, France,
in 1995. He received the D.E.A. degree from the
University of Rennes, Rennes, France, in 1998.
He joined the École Normale Supérieure de
Cachan in 1995. Since 1998 he has been with
Université Technologique de Belfort Montbéliard,
Belfort, France, where he is currently a Lecturer
with the Computer Sciences Department, where
he is mainly involved in computer architecture, signal processing, and
telecommunications.