Beruflich Dokumente
Kultur Dokumente
discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/3318291
CITATIONS
DOWNLOADS
VIEWS
326
1,913
409
3 AUTHORS, INCLUDING:
Pierre Siohan
Cyrille Siclet
Orange Labs
SEE PROFILE
SEE PROFILE
1170
I. INTRODUCTION
1171
(FIR) prototype filter of arbitrary length, which are derived in Section III;
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;
several design examples presenting a large set of orthogonal and nearly orthogonal prototype filters with different
pulse shapes. Design charts, tables, and graphical displays
are provided in Section V, allowing different comparisons
in terms of orthogonality and timefrequency localization.
denote the set of integer and real numNotations:
correspond to the space of
bers, respectively.
square-summable continuous and discrete-time functions,
respectively. Vectors and matrices are denoted with bold italic
letters, for instance, . Superscript stands for transposition.
with lower case letters, for inWe denote discrete filters of
, and their -transform with upper case letters, such
stance,
.
and
designate the real part and imaginary
as
part of the complex-valued number , respectively. Similarly,
, we get
and
for a filter
. Superscript denotes complex
. The
conjugation. For a filter
tilde notation denotes paraconjugation:
.
is the
norm of .
(2)
(3)
(4)
(5)
In order to get simplified, easier-to-manipulate expressions, the
following notations can be introduced:
(6)
(7)
We also denote
(8)
A. Continuous-Time Formulation
(9)
(1)
is the signaling interval,
the spacing bewhere
and
the real and imagtween two successive carriers,
inary parts, respectively, of the QAM complex-valued symbols
we want to transmit, and
a symmetrical real-valued
is defined by
(11)
(12)
Assuming a distortion-free channel, the transmitted symbols are
, if and only if the set of
perfectly recovered, i.e.,
constitutes an orthonormal basis of its span, i.e.,
functions
(13)
1172
where
B. Discrete-Time Formulation
In the previous section, we considered continuous-time
with its classical real-valued inner product
signals in
. Now, our goal is to obtain a discrete-time signal
with its classical real-valued inner product
,
in
which is defined by
(14)
corresponds to the transmission of
Since the duration
complex-valued symbols, the critical sampling period is defined
by
(15)
is critically sampled, and therefore, the
This means that
spectral efficiency is still maximum. In order to get a causal diswith length equal to
is truncrete-time prototype
and is delayed by
cated to the interval
time units. Moreover, in order to get a prototype filter with norm still approximately equal to
has to be normalized with a multiplicative factor equal to
(16)
and
Thus, based on (8) and using the fact that
, the baseband transmitted discrete-time
that
OFDM/OQAM signal is such that
(20)
C. OFDM/OQAM Transmultiplexer
Equation (17) looks like the expression of a signal obtained at
sub-bands
the output of a synthesis bank of a filterbank with
on each sub-band.
and with an expansion factor equal to
and
denote the input signal and the filter
Indeed, if
on the th subband of this synthesis bank, respectively, then the
is written as [10, p. 117]
output signal
(21)
and
(22)
and with
(23)
can be defined by (23) since
has
We can remark that
modulo . Similarly, (19) looks
just to be equal to
like the expression of the output symbols of an analysis bank of
sub-bands and with a decimation factor
a filterbank with
on each sub-band. Thus, if
denotes the input
equal to
and
denote the filter and
signal of this bank and
the output signal on the th subband, respectively, then [10, p.
117]
(24)
, let us rewrite (19) in
Using the fact that
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
) by
(19)
and
and
(26)
1173
trary length
works cited previously.
appear as
(27)
The orthogonality conditions of the modulated transmultiplexer we have just obtained naturally depend on the analysis
, respectively.
and synthesis filterbanks, i.e.,
To get a compact representation of this system, we use the
polyphase approach, which leads us to IFFT-based implementations, to the inputoutput relation and, finally, allows us to
get the mathematical orthogonality conditions.
A. Polyphase Approach
Then, by defining
and
, which are the -transand
, respectively, we get the transmultiforms of
plexer scheme depicted in Fig. 1. In this way, we get a transmultiplexer with the following particularities.
The input and output signals are real-valued symbols.
is twice the ex The number of channels
, whereas for claspansion and decimation factors
sical transmultiplexers that are critically decimated [10]
or
or oversampled [12], we necessarily have
, respectively. Exact reconstruction in this case
is nevertheless made possible by the fact that the input
symbols are real-valued; therefore, the present transmultiplexer is equivalent to a critically decimated one with
complex-valued inputs [cf. (15)].
Before expansion by a factor , each real-valued input is
premodulated, whereas at the receiver side, the dual operation occurs after decimation, the real part being extracted
afterwards (cf. Fig. 1).
, with depending on the length of the pro A delay
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 implementation scheme involves two integer parameters and that
allow us to handle the case of causal prototype filters with arbi-
(30)
.Wecanthenrewrite
and
where
asfunctionsof
andderive thepolyphasematricesof thesynthesis(modulator)andanalysis(demodulator)parts,whicharede(type 2) and
(type 1) [10], respectively:
noted
(31)
(32)
1174
Fig. 2.
where is the
antidiagonal matrix. Therefore, we
can directly derive from Fig. 2 the IFFT-based modulator and
demodulator schemes depicted in Figs. 3 and 4, respectively.
Furthermore, Figs. 3 and 4 only represent two basic schemes
that can be further improved, as shown in Appendix A.
C. InputOutput Relation
B. IFFT-Based Implementations
Direct implementation of the systems depicted in Figs. 1 and
2 would be very costly. However, as the modulator and demodulator are related to a modulated transform, we examine, in this
section, how it is possible to get an efficient IFFT-based implementation.
and
as functions of
In order to get the matrices
, we now introduce the matrices
and
,
which are given by
diag
diag
(33)
(34)
(37)
Fig. 4.
1175
(39)
D. Mathematical Expression of Orthogonality
Perfect orthogonality is provided when there is no ISI, i.e.,
, nor ICI, i.e.,
.
Using filterbank terminology, we may also say that the orthogonal transmultiplexer checks the perfect reconstruction (PR)
, where
property with the delay if and only if
is the
identity matrix. In Appendix C, we derive
an equivalent formulation of this latter equality, which leads
us to the following theorem showing how the conditions of
perfect orthogonality can be directly related to the polyphase
components of the prototype filter.
be a symmetrical real-valued prototype
Theorem 1: Let
filter that is assumed to be causal and time-limited with arbibe its
type
trary length , and let
sub-band OFDM/OQAM
I polyphase components. Then, a
signal, using this prototype, can be transmitted through a distortion-free channel and reconstructed without ISI nor ICI, with
denoting the integer part functhe delay
tion by superior value if and only if
(40)
..
.
..
..
..
..
..
..
.
(38)
1176
of
(41)
seen in several publications related to OFDM/OQAM over wireless channels [2], [7], [8], [16]. In this paper, in order to assess
our results on a solid basis, we will use two different localization measures.
Let be a discrete-time signal with finite energy, i.e.,
, and with Fourier transform denoted
. Its second, and in frequency,
order moment in time, which is denoted
, are given by
which is denoted
(45)
and
where
shown that
minimize
and
, respectively. It can be
(43)
(46)
Moreover, using (39), we notice that
so that
mod
Moreover, for a real-valued signal,
It can be deduced from [17] that if
define a localization measure denoted such that
(44)
V. DESIGN RESULTS
The quality of a given prototype filter can be evaluated by
means of the resulting ISI and ICI and, generally, by its ability
to satisfy a given criterion. In this section, two situations are considered: a first one where the prototype is obtained by truncation
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 conditions. In both cases, the timefrequency localization is used as
a criterion either for evaluation, as in Section V-B, or for optimization, as in Section V-C.
A. TimeFrequency Localization Measures
It is well known that the wireless channel introduces time and
frequency dispersion. To limit the resulting distortion, a solution proposed in [2] is to use basic signals being localized in
time and frequency with the same timefrequency scale as the
channel itself. In this context, time and frequency dimensions
are equally important. Indeed, timefrequency localization is
now often considered as being the right criterion, as it can be
.
, we can
(47)
(48)
(49)
(50)
(51)
where and are the gravity centers in time and frequency,
and
, which are secondrespectively, which minimize
(42)
1177
shown at the bottom of the page, where is the roll-off param. Therefore, the continuous-time expression of
eter
SRRC functions is given by
(52)
(56)
(53)
(54)
The optimum is still
(58)
is a real-valued number,
are some
where
is the Gaussian function
real-valued coefficients, and
.
For OFDM/OQAM signals, we suppose that
, and in order to satisfy (13), we also suppose that
is such that
. Given that
, the corresponding filters are multiplied by a
. The discrete-time version
normalization factor equal to
of the SRRC, OFDP, and EGF prototype filters are then obtained
(except for
by truncation to the interval
OFDPs, which no longer need to be truncated) and thanks to
(16).
for the EGF, as
Besides, when setting
in [2] and [21], has to be approximately within the
range to get the best timefrequency localization measures [22].
, the maximum disIn Table I, we have compared, for
[cf. (44)] obtained for a few EGF and SRRC functortion
and
tions. They show that for a given length, except for
, the EGF clearly outperform the best SRRC funcand
tion. Some other simulations carried out with
give the results reported in Figs. 5 and 6. They lead us to the
ratio, the
conclusion that for any value of and a given
EGF can probably provide better results than the SRRC functions. They also show that in both cases, the distortion mainly
ratio.
depends on the
We can deduce from a result previously obtained for cosinemodulated filterbanks [22] that for OFDM/OQAM systems, the
(55)
1178
TABLE I
MAXIMUM DISTORTION (D
) FOR SRRC(r ) AND EGF() WITH
DIFFERENT VALUES OF r AND ,AND FOR M = 4
of the
coefficients given in [20]. This table shows that the
distortion of OFDPs slightly increases when the number of
. It also shows that OFDPs
carriers increases beyond
have a smaller distortion than SRRCs for the same
ratio. Moreover, the distortion of OFDPs is also smaller than
and
. Finally, for
the distortion of EGFs for
and
, OFDPs are between the EGFs
and
.
obtained for
C. Design Examples Satisfying the Discrete Orthogonality
Conditions
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).
From Table I and Figs. 5 and 6 and as illustrated with previous publications (see, for instance, [23]), it also clearly apratios may
pears that to get low distortion values, high
be required. Furthermore, the truncation and sampling operations do not allow direct control of the timefrequency localization. On the other hand, if we choose to directly optimize the
prototype filter to get a maximum timefrequency localization
under the PR constraint (40), we are, at least, ensured to get
neither ISI nor ICI distortion. This optimization can be directly
under
conducted by optimizing the prototype coefficients
the constraint (40), or it can be also carried out using a set of lattice coefficients that corresponds to an efficient implementation
for
of each pair of polyphase components
, [10, pp. 380383], [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 coefficriterion as the optimized lattice
cients according to the
(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
for the EGF
general, the best localization measures, i.e.,
for the SRRC functions. As stated previ[21], [22] and
ously, the OL method is the only one that provides a perfect
distortion theoretically equal to 0
orthogonality, i.e., a
but, in practice, due to numerical inaccuracies, around
or
for double precision computers. Furthermore, as our
results reported in Table III show, it appears that for the two
), despite the introduction of the PR
measures used ( and
constraint, we have the following.
Compared with the SRRC function, the OL method always provides better localized prototype filters, even for a
smaller length than for SRRC filters.
Compared with the OFDP, OL filters have a better localization measure for a given length. This is not surprising
since OFDPs are not optimized according to the maximization of the timefrequency localization.
Compared with the EGF, we can see that the OL method
generally provides better results. This is particularly true
, i.e., a pulse length
for the shortest filter lengths (
limited to the duration of one complex-valued symbol).
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. 8.
= 1 for
= 2:846e 02.
EGF with
0:1442;D
Fig. 9. OL for
7:002e 15.
= 1024 and
= 1024 and
= 2048;
= 2048;
= 0:9049; D
1180
If
VI. CONCLUSION
We have presented an application of the filterbank theory to
the analysis and design of OFDM/OQAM systems. In a first
step, we have derived the transmultiplexer scheme that can be
associated with OFDM/OQAM systems with symmetrical FIR
prototypes of arbitrary length. Then, using a polyphase decomposition of this prototype, we have proposed fast implementation schemes of the OFDM/OQAM, modulator, and demodulator, which are based on the IFFT, and we have also found the
discrete orthogonality conditions related to the polyphase components of this prototype. Systems that do not exactly satisfy the
discrete orthogonality conditions create ISI and ICI. For such
nonorthogonal systems, and assuming a distortion-free transmission channel, we have derived the analytical expressions of
the ISI and ICI distortions created by imperfect filterbanks. Finally, we have provided a set of design examples that clearly
show that the numerical optimization of a discrete orthogonal
prototype has, at least, two important advantages compared with
truncation and sampling of prototypes being orthogonal in continuous time. Indeed, the direct design of orthogonal prototypes
in discrete time not only avoid ISI and ICI distortions for distortion-free channels, but they also allow us to obtain short prototype filters with very good localization measures. This naturally
tends to significantly reduce the computational cost of implementation.
is even,
, and for
(63)
Moreover, the FFT in (63) is an FFT with real-valued inputs; therefore, it can be carried out thanks to an FFT with
complex-valued inputs, instead of
real-valued inputs [25, p. 82].
is odd,
, and
If
(64)
Equation (64) corresponds to an odd DFT with real-valued
inputs that can also be realized thanks to an FFT with
complex-valued inputs instead of
real-valued inputs
[25, p. 83]. Thus, as in [23], it is possible to use only an
FFT of size .
Similarly, at the demodulation part, the IFFT followed by a rois also equivalent to an FFT
tation by
. Then, we will only
with a permutation of the inputs
keep the real part of the outputs of the FFT so that further reduction of the complexity is also possible.
APPENDIX B
COMPUTATION OF THE TRANSFER MATRIX
In Section III, we have shown that the transfer matrix
can be written
(65)
where
(66)
REDUCTION
OF THE
APPENDIX A
COMPLEXITY OF
THE
IFFT-BASED
IMPLEMENTATIONS
It can be seen from Fig. 3 that for
(60)
Let us define two integers
and
,letusfirstnotethatadelay
placedbetween
Toevaluate
an expansion and a decimation by leads to a transfer function
such that
, with
if is a mulotherwise. Thus, we obtain
tiple of and
, and from (66), we get
. Therefore, the entries of matrix are
by
or
(67)
(61)
if
otherwise
(62)
(68)
if
if
(69)
1181
(70)
(71)
Moreover, using (72), shown at the bottom of the page, we obtain from (70)(72) that
and finally
(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
, the
of the page. Using the decomposition
can be simplified as follows when
expression of
is even:
(75)
Furthermore, it can be checked by inspection from (29) and (69)
for
. Thus,
that
we finally obtain the expression of the transfer matrix given by
(38) and (39).
for
(72)
for
if
is even
otherwise
(74)
1182
APPENDIX C
ORTHOGONALITY CONDITIONS
The matrix
for
(81)
ACKNOWLEDGMENT
for
(76)
(77)
. Moreover,
This relationship is always true for
and
lead to the same relationship; thus
(78)
(79)
(80)
where
ness as
is the unique value of that has the same even. Moreover, as in Appendix B, using the fact that
for
, we obtain
[1] S. B. Weinstein and P. M. Ebert, Data transmission by frequency-division multiplexing using the discrete Fourier transform, IEEE Trans.
Commun. Technol., vol. COM-19, pp. 628634, Oct. 1971.
[2] B. Le Floch, M. Alard, and C. Berrou, Coded orthogonal frequency
division multiplex, Proc. IEEE, vol. 83, pp. 982996, June 1995.
[3] B. Hirosaki, An orthogonally multiplexed QAM system using the discrete Fourier transform, IEEE Trans. Commun., vol. 29, pp. 982989,
July 1981.
[4] R. W. Chang, Synthesis of band-limited orthogonal signals for multichannel data transmission, Bell Syst. Tech. J., vol. 45, pp. 17751796,
Dec. 1966.
[5] B. R. Saltzberg, Performance of an efficient parallel data transmission
system, IEEE Trans. Commun. Technol., vol. 15, pp. 805811, Dec.
1967.
[6] A. Vahlin and N. Holte, Optimal finite duration pulses for OFDM,
IEEE Trans. Commun., vol. 44, pp. 1014, 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. 118,
1997.
[8] H. Blcskei, P. Duhamel, and R. Hleiss, Design of pulse shaping
OFDM/OQAM systems for high data-rate transmission over wireless
channels, in IEEE Int. Conf. Commun., Vancouver, BC, Canada, June
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. 979995, 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
modem using orthogonally multiplexed QAM technique, IEEE Trans.
Commun., vol. COMM-34, pp. 587592, June 1986.
[12] R. Hleiss, P. Duhamel, and M. Charbit, Oversampled OFDM systems,
in IEEE Int. Conf. Digital Signal Process., Greece, July 1997.
[13] N. J. Fliege, Orthogonal multiple carrier data transmission, Eur. Trans.
Telecommun., vol. 3, no. 3, pp. 255264, MayJune 1992.
[14] T. Q. Nguyen and R. D. Koilpillai, The theory and design of arbitrarylength cosine-modulated filter banks and wavelets, satisfying perfect
reconstruction, IEEE Trans. Signal Processing, vol. 44, pp. 473483,
Mar. 1996.
[15] T. Karp and N. J. Fliege, MDFT filter banks with perfect reconstruction, in Proc. IEEE Int. Symp. Circuits Syst., vol. 1, Seattle, WA, Apr.
1995, pp. 744747.
[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. Vilb, On the uncertainty principle in discrete signals, IEEE Trans. Circuits Syst. II, vol. 39, pp. 394395, June 1992.
[18] M. I. Doroslovacki, Product of second moments in time and frequency
for discrete-time signals and the uncertainty limit, Signal Process., vol.
67, no. 1, pp. 5976, May 1998.
[19] P. R. Chevillat and G. Ungerboeck, Optimum FIR transmitter and receiver filters for data transmission over band-limited channels, IEEE
Trans. Commun., vol. COMM-30, pp. 19091915, Aug. 1982.
[20] A. Vahlin and N. Holte, Optimal finite duration pulses for OFDM, in
Proc. IEEE Global Commun. Conf., San Francisco, CA, Nov. 1994, pp.
258262.
[21] C. Roche and P. Siohan, A family of extended Gaussian functions with
a nearly optimal localization property, in Proc. Int. Workshop MultiCarrier Spread-Spectrum, Germany, Apr. 1997, pp. 179186.
1183
[22] P. Siohan and C. Roche, Cosine-modulated filterbanks based on extended gaussian functions, IEEE Trans. Signal Processing, vol. 48, pp.
30523061, Nov. 2000.
[23] G. Cariolaro and F. C. Vagliani, An OFDM scheme with a half complexity, IEEE J. Select. Areas Commun., vol. 13, pp. 15861599, Dec.
1995.
[24] U. Reimers, Digital video broadcasting, IEEE Commun. Mag., pp.
104110, June 1998.
[25] M. Bellanger, Digital Processing of Signals: Theory and Practice. New York: Wiley, 1989.