iofdmpaperjun04
Transcript of iofdmpaperjun04
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 1/11
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 52, NO. 6, JUNE 2004 1711
Interleaved Orthogonal Frequency DivisionMultiplexing (IOFDM) System
V. G. S. Prasad and K. V. S. Hari , Senior Member, IEEE
Abstract—In orthogonal frequency division multiplexing(OFDM) systems, for every block of data samples, an overheadof samples of cyclic prefix (CP) or zero padding (ZP) is addedto combat frequency selective channels. The code rate, which isdefined as the ratio
( + )
, is a measure of the efficiencyof transmitting user information. In this paper, a new systemis proposed to increase the code rate without increasing thenumber of subcarriers and without increasing the bandwidth. Theproposed system considers appending the zeros (ZP) once forevery blocks of data samples, which would increase the coderate to ( + ) . It is assumed that the channel is notvarying over the transmission of consecutive data blocks. Inorder to recover the data blocks in a computationally efficient
manner, an interleaving scheme is proposed, and the proposedsystem is called the interleaved OFDM (IOFDM) system. Variousissues such as computational complexity, peak-to-average powerratio (PAPR), and the effect of synchronization errors on theperformance of the IOFDM system are also presented. Based ona numerical simulation study, the average bit-error-rate (BER)performance of the IOFDM system is shown to beveryclose to thatof the OFDM system for a moderate increase in computationalcomplexity and delay.
Index Terms—Code rate, cyclic prefix, interblock interference,interleaving, OFDM, zero padding.
I. INTRODUCTION
I N recent years, orthogonal frequency division multiplexing(OFDM) has been adopted as a standard for various applica-
tions like digital audio/video broadcasting (DAB/DVB), wire-
less LANs, etc. Conventional OFDM systems transform infor-
mation symbol blocks and then insert redundancy in the form
of either cyclic prefix (CP) or zero padding (ZP) [ 1]–[3]. The
length of CP/ZP should be longer than the channel delay spread
to avoid interblock interference (IBI) arising due to the fre-
quency-selective nature of the channel.
Itwas proposedin [4]and[5] that ZP assures symbolrecovery
even when channel nulls occur on some subcarriers, which is
not possible with the use of CP. However, there is an increase in
receiver complexity.
The redundancy due to the CP/ZP causes reduction in thecode rate of the communication system. The loss in the code
rate may be around 10–25%. The loss is more when the finite
Manuscript received June 20, 2002; revised June 21, 2003. The associate ed-itor coordinating the review of this manuscript and approving it for publicationwas Dr. Sergios Theodoridis.
V. G. S. Prasad is with Emuzed India Pvt. Ltd., Bangalore, India.K. V. S. Hari is with the Department of Electrical Communication En-
gineering, Indian Institute of Science, Bangalore 560 012, India (e-mail:[email protected]).
Digital Object Identifier 10.1109/TSP.2004.827179
impulse response (FIR) channel order is comparable with the
number of subcarriers used in the OFDM system.
In this paper, we propose a scheme called interleaved or-
thogonal frequency division multiplexing (IOFDM), which has
a higher code rate than the conventional OFDM. IOFDM en-
hances the code rate without bandwidth expansion and without
increasing the number of subcarriers but with a moderate in-
crease in computational complexity and delay.
In the proposed IOFDM system, it is assumed that the
channel is not varying over the transmission of consecutive
data blocks. It is proposed that ZP is used once for every
transmission of consecutive data blocks, which wouldincrease the code rate. However, there is an introduction of
IBI. In order to recover the data symbols in a computationally
efficient manner, an interleaving scheme is proposed. In [6],
a scheme called vector OFDM (VOFDM) was presented to
increase the code rate. Even though the VOFDM transmitter
and the IOFDM transmitter look similar, the receiver sections
are entirely different. The IOFDM system is computationally
more efficient than the VOFDM system, as shown later.
Peak-to-average power ratio (PAPR) of the IOFDM system and
the effect of synchronization errors on the performance of the
IOFDM system are also presented.
The paper is organized as follows. In Section II, the conven-
tional OFDM system is briefly described. Section III presentsthe proposed IOFDM system. Section IV presents the perfor-
mance of theIOFDM systemusingsimulationresults. Thepaper
is concluded in Section V.
A. Notation
Underlined lowercase letters are used to represent vectors,
and boldfaced upper case letters are reserved for matrices. The
superscripts , , and stand for transpose, complex conju-
gate transpose, and inverse operation, respectively. repre-
sents the identity matrix, and represents the
discrete Fourier transform (DFT) matrix. denotes
the convolution of and . denotes the expecta-tion operator on , and represents Kronecker delta func-
tion. denotes the estimate of .
II. OFDM SYSTEM
A. OFDM Transmitter
The block diagram in Fig. 1 describes the discrete-time base-
band model of an OFDM system [7]. The bit stream is mapped
to an information symbol sequence using a modulation
scheme like binary phase shift keying (BPSK), quaternary phase
1053-587X/04$20.00 © 2004 IEEE
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 2/11
1712 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 52, NO. 6, JUNE 2004
Fig. 1. Block diagram of an OFDM system.
shift keying (QPSK), 16-QAM, etc. The sequence is parsed
into blocks of length
. The symbol block is then mapped to a precoded block of length through
a precoder such that
(1)
The role of is to effectively convert a frequency-selective
channel into a number of flat channels. It is assumed that the
frequency-selective channel is FIR in nature and that the upper
bound on its order is available, i.e., ,
where denotes the channel between the transmitter and the
receiver. A transmitted block
of length is formed as follows:
(2)
Setting the last elements of to zeros is known as zero
padding. The ZP will avoid IBI between the transmitted blocks.
After parallel-to-serial conversion of , the sequence
is then serially transmitted through a transmitting antenna.
B. OFDM Receiver
At the receiver, the received sequence in the presence of
noise is given by
(3)
where denotes complex additive white Gaussian noise(AWGN). It is assumed that the impulse response is con-
stant over the transmission of . Since the channel induced
IBI between the transmitted blocks is avoided due to the ZP, one
can focus at each received block
of length separately. After serial-to-parallel con-
version of , the received block is given by
(4)
where is the channel convolution matrix with the
th element being and
.
Define the -transforms:1 ,
, and . Then,
the -transform of is given by
(5)
Choosing distinct points on the complex -plane,
the precoder is designed such that
(6)
where is chosen to impose power constraint, i.e.,
. Equation (6) enables
proper symbol recovery at the receiver. The precoder can be
constructed by considering to be an FIR filter of order
with zeros being and [7]. The
polynomial can be uniquely determined by Lagrange
interpolation through the points and is obtained as
follows:
(7)
Evaluating (5) at , we get
(8)
Using (6), (8) is simplified as
(9)
Equation (9) reveals that OFDM converts the frequency-selec-
tive channel into a number of flat channels.
1c ( m ) denotes the element present in the ( m + 1 ) st row and ( k + 1 ) st
column of C , and C ( z ) denotes the Z -transform of the ( k + 1 ) st column of C .
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 3/11
PRASAD AND HARI: IOFDM SYSTEM 1713
In order to decode , we define
and write the set of equations given by (9) in a matrix/vector
form as follows:
(10)
where diag , and
......
. . ....
It is assumed that perfect channel state information (CSI) is
available at the receiver. Using the available CSI, we obtain the
decision vector of
length as follows:
(11)
where . After detecting from
, the estimated symbol sequence can be retrieved.
C. Improving the Code Rate
Even though OFDM simplifies equalization at the receiver,
the redundancy due to the CP/ZP causes reduction in the code
rate. The code rate of the OFDM system, which is denoted by
, is given by
(12)
One way to increase the code rate is by increasing the number of
subcarriers in the OFDM system. Let us consider the number of
subcarriers in the OFDM system to be , where . The
code rate would then be equal to , which
is greater than . Fig. 2 shows the improvement in the code
rate, which is denoted by , as a function of , for
different values of and . It is to be noted that the code rate
improvement flattens out after a moderate value of . The code
rate improvement for moderate values of is between 16 and
32%, which is significant. Since the OFDM systems are verysensitive to frequency offsets, the larger the number of subcar-
riers implies a decrease in the frequency spacing, and therefore,
frequency accuracy becomes more and more critical. In addi-
tion, the PAPR increases with the number of subcarriers.
In VOFDM,2 vectors of length are used as
the cyclic prefix once for every transmission of consecu-
tive symbol vectors of length . The code rate in the VOFDM
system, which is denoted by , is given by
(13)
2d a e stands for the smallest integer b such that b a .
Fig. 2. 0 versus P .
D. Problem Statement
Given the constraint on the number of subcarriers and thebandwidth, obtain a scheme that improves the code rate of an
OFDM system.
III. IOFDM SYSTEM
We now present the idea of processing consecutive data
blocks of length using the ZP of length . This would yield
the same code rate as given by . The proposed IOFDM
system is presented in this section.
For the sake of illustrating the principle behind the
proposed method, we will first consider the case when
. Considering two consecutive precoded blocks ,
, and zeros, we construct the transmitted block of length
as follows:
(14)
It is assumed that the channel is constant over the trans-
mission of .
At the receiver, the corresponding received block
of length in the absence
of noise is given by
(15)
where is the channel convolution matrix, where theth element is . The -transform of is given
by
(16)
Evaluating (16) at , we get
(17)
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 4/11
1714 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 52, NO. 6, JUNE 2004
Therefore, we have equations with unknowns. We need
additional equations to retrieve the information symbols
. Considering additional
distinct points on the complex -plane, we get
(18)
In order to decode , we define
and write the set of equations given by (17) and (18) in a ma-
trix/vector form as follows:
(19)
where diag ,
and
......
. . ....
. . .
......
. . ....
. . .
Solving the set of equations given by (19) is an expensive
process because we need to invert the matrix . In
addition, the choice of and, therefore, the inversionof is not straightforward to establish.
To address this issue, we propose to construct the transmitted
block of length by interleaving the elements
of and as follows:
(20)
The corresponding -transform of the received block of
length in the absence of noise is given by
(21)
Evaluating (21) at and and then
using (6), we obtain the equations needed to retrieve
as follows:
(22)
Solving the set of equations given by (22) is not an expensive
process because number of 2 2 Vandermonde matrix in-
versions are needed to retrieve
[8]. We can further reduce the computational complexity of
the receiver by preprocessing the symbol block at
the transmitter. The preprocessing results in the precoded block
being
(23)
where diag is the preprocessing matrix.
Because of preprocessing, (22) becomes
(24)
Solving the set of equations given by (24) needs two-point
inverse discrete Fourier transform (IDFT) operations. Next, we
generalize the above idea by considering consecutive symbolblocks.
A. IOFDM Transceiver
We consider consecutive symbol blocks
and then preprocess the symbol block
as follows:
(25)
where diag is the preprocessing matrix.
Now, the preprocessed block is the effective symbol block.
It is linearly mapped to a precoded block through the pre-coder such that
(26)
A transmitted block of length is formed
by interleaving the elements of and
padding zeros as follows:
(27)
After parallel-to-serial conversion of , the sequence
is then serially transmitted through a transmitting antenna. Be-cause of interleaving, we call the proposed scheme interleaved
orthogonal frequency division multiplexing (IOFDM). The
block diagram in Fig. 3 describes the discrete-time baseband
model of an IOFDM transmitter. The interleaving process at
the IOFDM transmitter for is shown in Fig. 4.
At the receiver, the received sequence is given by
(28)
where is complex AWGN. Here, it is assumed that the
channel is constant over the transmission of . Since
IBI between the transmitted blocks is avoided due to the ZP,
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 5/11
PRASAD AND HARI: IOFDM SYSTEM 1715
Fig. 3. Block diagram of an IOFDM transmitter.
Fig. 4. Interleaving process at the IOFDM transmitter for P = 3 .
one can focus at each received block of length , which
is given by
(29)
where is the channel convolutionmatrix, and
. The -transform of the re-
ceived block is given by
(30)
Let be the distinct points on the complex
-plane, where they are related to as follows:
and
(31)
Evaluating (30) at , we get and
(32)
The -transforms evaluated at are
obtained as follows:
......
.... . .
...
(33)Using (6) and (31), (32) is simplified as follows:
(34)
In order to decode , we define
and write the set of equations given by (34), for a fixed , in a
matrix/vector form as follows:
(35)
where diag ,
, and
. After frequency
domain equalization and -point IDFT, we obtain the decision
vector
of length as follows:
(36)
where . After detecting
from , the estimated symbol sequence can be re-
trieved. The block diagram in Fig. 5 describes the discrete-time
baseband model of an IOFDM receiver. Fig. 6 shows the dein-
terleaving process at the IOFDM receiver for .
Remarks:
• For , the IOFDM system is equivalent to the con-
ventional OFDM system.
• In OFDM, if subcarriers are used, the precoder
would be designed using distinct points
on the complex -plane. It would also be assumed that
the channel nulls do not occur at
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 6/11
1716 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 52, NO. 6, JUNE 2004
Fig. 5. Block diagram of an IOFDM receiver.
Fig. 6. Deinterleaving process at the IOFDM receiver for P = 3 .
, for proper data recovery. In IOFDM, the condition on
the channel nulls remains similar, but the precoder is
designed using only distinct points .
• Equation (36) reveals that IDFT operations are
required to retrieve . This
is less expensive than the matrix inversion.
• There is an increase in the processing delay as data
blocks have to be received.
B. Example of the IOFDM System
Let be the distinct points on
the complex -plane. For this choice of , the points, the matrices , , , , and the constant are
as shown in the equation at the bottom of the page, where is
the overlap-add operation matrix, which is defined as
follows:
where is the th column of . For this choice of ,
we can exploit the computational advantage of fast Fourier
transform (FFT) to implement the IOFDM system.
Remarks:
• The form of receiver considered is not robust to channel
nulls at subcarrier locations . Other solutions
(like the pseudoinverse solution [4] and [5]) can be used
for ZP-IOFDM, which are robust to channel nulls at sub-
carrier locations, but these methods may not give the com-
putational advantage. However, the benefit of code rate
improvement and PAPR improvement (discussed in a later
section) still exists with the use of IOFDM.
C. Computational Complexity
Even though IOFDM improves the code rate, there is a mod-
erate increase in computational complexity. In the conventional
OFDM system, for a carrier system, the transmitter has to
operate a general precoder matrix of size , which
could result in a computation cost of complex multi-
plications (CMs). In case of the IOFDM system, it would re-
quire CMs for precoder operation and an additional
CMs for diagonal transformation, as in (25), resulting in a total
number of CMs to be , which would be less than the
previous case. For the case when is the IDFT matrix (as in the
example above), the reduction in CMs is analogous to to savings
obtained due to splitting the -point IDFT into smaller but-
terflies.
The number of CMs required to retrieve consecutivesymbols at the receiver is
taken as the complexity measure. Let and beinteger powers
of 2, i.e., and . The number of CMs required
to find the -point DFT is .
For the choice of , we compare
the complexity of IOFDM with the complexities of OFDM and
VOFDM. In the conventional OFDM system, the number of
CMs required to evaluate is .
The number of CMs required to evaluate and to get
from as in (11) is . Therefore, the com-
plexity of the conventional OFDM system, which is denoted by
, is given by
(37)
diag
diag
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 7/11
PRASAD AND HARI: IOFDM SYSTEM 1717
In the proposed IOFDM system, the number of CMs required
to evaluate both and is
. In order to get
from as in (36), we need CMs. Thus,
the complexity of the IOFDM system, which is denoted by ,
is given by
(38)
To obtain the computational complexity of the VOFDM
system, let and be the transmitted block and the
received block, respectively. The -transform of the received
block in the absence of noise is given by
(39)
Rewriting (39) in terms of polyphase components [9], we get
(40)
In VOFDM, the symbol vector is given by
(41)
where
......
. . ....
and .
The number of CMs required to evaluate both
and is .
In order to get from as in (41), we need CMs.
Therefore, the complexity of the VOFDM system, which is
denoted by , is given by
(42)
Table I shows the comparison of the IOFDM system’s com-plexity and the VOFDM system’s complexity for .
Remarks:
• It is clear that the IOFDM system requires more com-
plexity than the OFDM system to decode data sym-
bols.
• The IOFDM system is computationally more efficient
than the VOFDM system.
D. Peak-to-Average Power Ratio
For the choice of , the dis-
crete-time baseband representation of the transmitted signal in
TABLE ICOMPUTATIONAL COMPLEXITY FOR K = 2 5 6
the conventional OFDM system, consisting of subcarriers,
is given by
(43)
Because of the superposition of many statistically independent
subcarrier signals, the OFDM signal exhibits high instantaneous
peak signal level with respect to average signal level. This high
peak-to-average nature of the OFDM signal results in harmonic
distortion, unless the transmitter’s power amplifier exhibits lin-
earity throughout the dynamic range of the signal. PAPR is used
as a measure to compare the instantaneous peak signal level
with the average signal level. PAPR of the conventional OFDM
system, which is denoted by , is defined as
(44)
where is the average symbol energy. The worst-case PAPR
that can happen in the OFDM system is
(45)
where is the symbol having largest power in the consid-
ered modulation alphabet.
In an IOFDM system that is implemented by interleavingconsecutive data blocks of length , the transmitted signal
is given by
and (46)
The PAPR of the IOFDM system, which is denoted by , is
defined as
(47)
The worst-case PAPR that can happen in the IOFDM system is
(48)
Comparing (45) and (48), we can say that both the IOFDM
system and the OFDM system, which consist of subcarriers,
have the same worst-case PAPR. Therefore, the IOFDM system
with interleaving data blocks of length has less worst-case
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 8/11
1718 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 52, NO. 6, JUNE 2004
PAPR compared with the conventional OFDM system with
subcarriers.
E. Effect of Synchronization Errors
In this section, we present the effect of synchronization errors
on the performance of both the OFDM system and the IOFDM
system. Let be the distinct points on the complex
-plane, which are uniformly separated on the unit circle.In the conventional OFDM system, the transmitted contin-
uous-time baseband signal over an OFDM symbol period
is described by
(49)
where is the symbol duration on each subcarrier, is the
ZP duration, is the sampling period which is defined as
, and
else.
In the presence of a carrier frequency offset , the received
signal in the absence of noise is given by
(50)
where denotes the FIR channel
between the transmitter and the receiver. The received signal
is sampled at the time instances
, where denotes the normalizedtimingoffset.
The discrete-time baseband representation of the received signal
is
(51)
where denotes the normalized carrier frequency
offset, and
Denoting as the -transform of , we get
(52)
Evaluating (52) at , we get,
(53)
where
In an OFDM system consisting of subcarriers, (53) be-
comes
(54)
Assuming and then
performing zero-forcing equalization, we get
(55)where
The sequence denotes the inter carrier interference (ICI)
and has an average power of
(56)
In the proposed IOFDM system, the continuous-time base-
band representation of the transmitted signal over an
IOFDM symbol period is given by
(57)
In the presence of a carrier frequency offset and a normal-
izedtiming offset , the received signal in the absence noise
is given by
(58)
where
and
Denoting as the -transform of and then evaluating
at , we get, and
, (59), shown at the bottom of the page. After
(59)
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 9/11
PRASAD AND HARI: IOFDM SYSTEM 1719
performing zero-forcing equalization and -point IDFT, we get,
and
(60)
where we have (61), shown at the bottom of the page. The se-
quence denotes ICI and has an average power of
(62)
Since , both the IOFDM system im-
plemented by interleaving data blocks of length and the
OFDM system consisting of subcarriers are equally af-
fected by synchronization errors.
IV. SIMULATION RESULTS
In this section, we present the average bit-error-rate
(BER) performance of the IOFDM system in various
channel models. QPSK modulation technique is used tomap the bit stream into an information symbol sequence. Let
be the distinct points on the
complex -plane. It is assumed that perfect CSI is available at
the receiver.
A. BER Performance in GSM Channels
In our simulations, the channel bandwidth is assumed
to be 5 MHz. The entire bandwidth is divided into
. To make subcarriers orthogonal
to each other, the symbol duration on each subcarrier, which
is denoted by , is s. Zero padding of duration
is used to avoid IBI arising due to the frequency-selective
nature of the channel. The value of is chosen according to
the channel used in simulation. The 12-tap GSM typical urban
(TU) channel and the hilly terrain (HT) channel are used.
We choose s in the TU channel and s
in the HT channel. The channel taps are characterized by the
Jakes’ Doppler spectrum [10]. In order to gain insight into the
average BER performance, we have taken 4000 information
symbol blocks of length per trial and averaged over
25 trials. In effect, for each (signal energy per informa-
tion bit to noise power spectral density ratio) value considered,
the BER values are obtained based on 51.2 bits.
Example 1: We study the performance of the IOFDM system
in the TU channel with a Doppler spread of 20 Hz. In Fig. 7, the
performance of the IOFDM system with different values of is
presented. The performance of the IOFDM system is very close
to that of the OFDM system ( ). It is slightly worse at low
values.
Fig. 7. Performance comparison of IOFDM systems in TU channel.
Fig. 8. Performance comparison of IOFDM systems in HT channel.
Example 2: We study the performance of the IOFDM system
in the HT channel with a Doppler spread of 20 Hz. Fig. 8 shows
the performance comparison of the IOFDM systems. As in the
previous example, the performance of the IOFDM system is
very close to that of the conventional OFDM system.
B. BER Performance in HIPERLAN/2 Channels
In the HIPERLAN/2 standard, the available channel band-
width is 20 MHz. The entire spectrum is divided into
subchannels. The symbol duration on each subcarrier is
s. A fixed duration of 0.8 s is used as ZP to avoid IBI,
(61)
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 10/11
1720 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 52, NO. 6, JUNE 2004
Fig. 9. Performance comparison of IOFDM systems in channel model B.
Fig. 10. Performance comparison of IOFDM systems in channel model E.
i.e., s and . Jakes’ Doppler spectrum corre-
sponding to a terminal speed of 3 m/s is assumed for all taps in
the channel. To gain insights into the average BER performance,
we have taken 8000 information symbol blocks, of length
, per trial and averaged over 25 trials. In effect, for each
value considered, the BER values are obtained based on
25.6 bits. Channel model B and channel model E are con-sidered in our simulations.
Example 3: We study the performance of the IOFDM system
in channel model B. Fig. 9 shows the performance comparison
of IOFDM systems in channel model B.
Example 4: We consider the case of using IOFDM for chan-
nels that are longer than the ZP used in an OFDM system. In
channel model E, the minimum length of the ZP needed to avoid
IBI is . The HIPERLAN/2 standard uses in gen-
eral. Because of the reduced ZP, the performance of the con-
ventional OFDM system is poor. To avoid this degradation, we
use the complete ZP, i.e., , and interleave more than two
precoded blocks, i.e., . Fig. 10 shows the performance
Fig. 11. Effect of normalized frequency offset of 1 = 0 : 0 5 on theperformance of various systems.
comparison of IOFDM systems in channel model E. As shown
in Fig. 10, the IOFDM system with and perform
better than the conventional OFDM system. This demonstrates
that for channels having longer lengths compared with the ZP,
we can improve the performance as well as increase the code
rate by adopting IOFDM.
The simulation results demonstrate that the IOFDM system
hasthe potentialto be an effective systemfor improvingthe code
rate of an OFDM system without loss of performance. Fig. 11
shows the effect of normalized frequency offset of on
the performance of various systems in the TU channel.
V. CONCLUSION
A new OFDM system based on interleaving the elements of
consecutive precoded blocks has been presented. The pro-
posed IOFDM system promises higher code rate compared with
the conventional OFDM system without bandwidth expansion,
without increasing number of subcarriers, and with moderate
increase in computational complexity. The IOFDM system im-
plemented by interleaving data blocks of length has less
worst-case PAPR compared with the OFDM system consisting
subcarriers. Both systems are equally affected by synchro-
nization errors. Simulation results verify that the performance
of the IOFDM system in terms of BER is very close to that of
the conventional OFDM system.
ACKNOWLEDGMENT
The authors would like to thank the reviewers for their valu-
able comments, which have helped the technical content of the
paper significantly.
REFERENCES
[1] J. A. C. Bingham, “Multicarrier modulation for data transmission: Anidea whose time has come,” IEEE Commun. Mag., pp. 5–14, May 1990.
[2] L. J. Cimini Jr., “Analysis and simulation of a digital mobile channelusing orthogonal frequency division multiplexing,” IEEE Trans.Commun., vol. COM-33, pp. 665–675, July 1985.
8/8/2019 iofdmpaperjun04
http://slidepdf.com/reader/full/iofdmpaperjun04 11/11
PRASAD AND HARI: IOFDM SYSTEM 1721
[3] R. W. Heath Jr and G. B. Giannakis, “Exploiting input cyclostationarityfor blind channel identification in OFDM systems,” IEEE Trans. SignalProcessing, vol. 47, pp. 848–856, Mar. 1999.
[4] B. Muquet, M. de Courville, P. Duhamel, and G. B. Giannakis, “OFDMwith trailing zeros versus OFDM with cyclic prefix: Links, comparisonsand application to the HIPERLAN/2 system,” in Proc. Int. Contr. Conf.,New Orleans, LA, June 2000, pp. 1049–1053.
[5] S. Zhou and G. B. Giannakis, “Finite-alphabet based channel estimationfor OFDM and related multicarrier systems,” IEEE Trans. Commun.,
vol. 49, pp. 1402–1414, Aug. 2001.[6] X.-G. Xia, “Precoded and vector OFDM robust to channel spectral nullsand with reduced cyclic prefix length in single transmit antenna sys-tems,” IEEE Trans. Commun., vol. 49, pp. 1363–1374, Aug. 2001.
[7] Z. Liu, G. B. Giannakis, S. Barbarossa, and A. Scaglione, “Transmit-antenna space-time block coding for generalized OFDM in the presenceof unknown multipath,” IEEE J. Selected Areas Commun., vol. 19, pp.1352–1364, July 2001.
[8] V. G. S. Prasad and K. V. S. Hari, “Interleaved orthogonal frequency di-vision multiplexing system,” in Proc. IEEE ICASSP, Orlando, FL, May2002, pp. 2745–2748.
[9] P. P. Vaidyanathan, Multirate Systems and Filter Banks. EnglewoodCliffs, NJ: Prentice-Hall, 1993.
[10] S. A. Fechtel, “A novel approach to modeling and efficient simulationof frequency selective fading radio channels,” IEEE J. Select. AreasCommun., vol. 11, pp. 422–431, Apr. 1993.
V. G. S. Prasad was born in Andhra Pradesh, India,in 1977. He received the B.Tech. degree from theNational Institute of Technology, Warangal, India,in 2000 and the M.E. degree from the Departmentof Electrical Communication Engineering, IndianInstitute of Science, Bangalore, India, in 2002.
Since March 2002, he has been a DSP engineer atEmuzed India Pvt. Ltd., Bangalore. His research in-terests include signal processing for communicationsystems and space-time and space-frequency codedtechniques for frequency selective channels.
K. V. S. Hari (SM’97) received the B.E. degreefrom Osmania University, Hyderabad, India, in1983, the M.Tech. degree from the Indian Instituteof Technology, New Delhi, India (IIT Delhi), in1985, and the Ph.D. degree from the University of California San Diego, La Jolla, in 1990.
He has been an Associate Professor with the De-partment of Electrical Communication Engineering,Indian Institute of Science (IISc), Bangalore, India,
since February 1998. His research interests arein statistical signal processing. He has worked onspace-time signal processing algorithms for direction-of-arrival estimation,acoustic signal separation using microphone arrays, and MIMO wirelesscommunication systems. He has also worked on MIMO wireless channel mea-surements and modeling and is the co-author of the IEEE 802.16 standard onwireless channel models for fixed broadband wireless communication systems.He was a visiting faculty at Helsinki University of Technology, Espoo, Finland,from May to July 2002; Stanford University, Stanford, CA, from September1999 to December 2000; and the Royal Institute of Technology, Stockholm,Sweden, from July to September 1995. He was an Assistant Professor withthe Department of Electrical and Computer Engineering, IISc, from February1992 to January 1998 and a scientist at Osmania University, from December1990 to January 1991. He was also with the Defence Electronics Research Lab,Hyderabad, from December 1985 to July 1987.