Академический Документы
Профессиональный Документы
Культура Документы
Abstract—This paper focuses on the use of guard intervals [4], [6], [7].
(GIs) between successive symbols for iterative channel estima- In the context of OFDM GIs, BCE corresponds to us-
tion (CE) in a time domain synchronous orthogonal frequency ing zero padding (ZP) instead of NZP sequences. Efficient
division multiplexing (TDS-OFDM) system. A hybrid of zero
padding (ZP) and nonzero padding (NZP) frame structures is equalizers are designed using the ZP scheme in [8], while
proposed to create an energy efficient scheme that can maintain a subspace algorithm in [9] provides a lower limit on the
accurate CE as well as a low transmission bit error rate (BER). number of symbols required for CE in the ZP scheme. In
The proposed hybrid scheme, i.e., joint zero padding-nonzero [10] and [11], the CE schemes presented in [8] and [9]
padding (ZP-NZP) scheme, is tested in Rayleigh fading channel are combined, and a generalized BCE scheme is proposed.
environments using computer simulations. The results show
that, for the same BER performance, the proposed joint ZP- However, the scheme performs better only at high signal-to-
NZP scheme outperforms the existing NZP scheme on the basis noise (SNR) values, i.e., SNR ≥ 20 dB. In [12] and [13],
of the reduction in the required signal-to-noise ratio (SNR) three different GI techniques used for CE, i.e., CP, ZP and
while its CE results are comparable to the existing NZP scheme. NZP, are compared, and a lower bound on the CE error is
obtained for the NZP scheme. According to the analysis in
[13], NZP OFDM outperforms the other schemes. In [14], the
I. I NTRODUCTION
CE results of [13] are improved by considering an iterative
We consider the problem of iterative channel estimation data aided CE scheme. In [15], the processing of an extended
(CE) in time domain synchronous orthogonal frequency received OFDM symbol is proposed which provides a new
division multiplexing (TDS-OFDM) systems. OFDM is the look at the TDS-OFDM circular convolution and CE. The
technology of choice for next generation wireless commu- work presented in [15] has been extended in [16] and [17]
nications for its robustness to channel imperfections for by considering the cyclic reconstruction of NZP sequences.
high bit-rate transmissions [1]. TDS-OFDM is a variant of It should be noted that all of the existing schemes use
conventional cyclic prefixed OFDM (CP-OFDM); it replaces either the same NZP or ZP sequence in every OFDM
the CPs in CP-OFDM systems with nonzero padding (NZP) symbol. In this paper, we propose a hybrid of ZP and NZP
sequences in TDS-OFDM systems [2]. The NZP sequence is GI techniques to create an energy efficient joint ZP-NZP
used for CE and synchronization purposes. Accurate CE is scheme. The scheme is energy efficient because we limit the
needed to ensure reliable transmissions. use of NZP sequences (high power pilots) to only a subset of
Various researches have proposed different CE methods for OFDM symbols. We use the iterative CE algorithm presented
OFDM systems. Different algorithms used in the literature in [15] for CE after some novel preprocessing of the received
for CE can be classified into three main categories [3]: pilot signal. The proposed joint ZP-NZP scheme does not require
aided channel estimation (PCE), blind channel estimation an initial CE. It uses OFDM frames comprising of both ZP
(BCE) and semi-blind channel estimation (SCE). PCE meth- and NZP OFDM symbols.
ods provide good channel estimates. However, their main The organization of the remaining sections is as follows.
drawback is the relatively high overhead as some data carriers The system model is explained in section II. Section III
are used for pilots [3], [4], [5]. BCE and SCE methods provides the details of the proposed iterative CE scheme.
use information transmitted during guard intervals (GIs) for In section IV, the equalization process is described. Section
CE. The need to achieve higher data rates motivates the V presents the numerical analysis. Section VI discusses the
search for BCE methods as it does not require pilots that simulation results. Finally, we conclude the paper in section
can utilize up to 20% of the data subcarriers [4]. However, VII.
the performances of BCE methods are generally not as good
as those of PCE methods. Besides, BCE methods also face II. S YSTEM M ODEL
the synchronization problem [4]. Compared to BCE methods, The system diagram for the proposed joint ZP-NZP
SCE methods lead to lower CE error, better synchronization OFDM scheme is shown in Fig. 1. Let N be the number
and faster convergence. These SCE methods are more suit- of OFDM subcarriers and D = [D0 , D1 , ..., DN −1 ] be a set
able for time selective channels than their blind counterparts, of N QAM symbols at the input of the inverse fast Fourier
and are more reliable for practical communication systems transform (IFFT) block. The GI sequence of length c, denoted
by g = [g0 , g1 , ..., gc−1 ], is appended to the IFFT output.
For an NZP symbol, g = p where p = [p0 , p1 , ..., pc−1 ]
represents the pseudorandom sequence. For a ZP symbol, g
is a zero vector. The output of the parallel-to-serial converter,
i.e., s = [d0 , d1 , ..., dN −1 , g0 , g1 , ..., gc−1 ], which can be
rewritten as s = [s0 , s1 , ..., sN2 −1 ] where N2 = N + c,
is transmitted through a Rayleigh fading channel having
the channel impulse response (CIR) h = [h0 , h1 , ..., hL−1 ],
where L represents the length of CIR. It is assumed that L
is known. Additive white Gaussian noise (AWGN) corrupts
the signal in the channel. Fig. 2. Structure of ZP-NZP frame.
The N2 samples of the received OFDM symbol, denoted
by r = [r0 , r1 , ..., rN2 −1 ], and the last c samples of the
previous symbol is used for CE. The first N samples of r
represent the received data samples. The equalized data vec-
tor, denoted by d̂ = [dˆ0 , dˆ1 , ..., dˆN −1 ], is used as an input to
the FFT block to get another vector D̂ = [D̂0 , D̂1 , ..., D̂N −1 ]
of equalized QAM symbols.
C. Channel Estimation
After our novel preprocessing, we use the algorithm
k th
where s represents the k OFDM symbol in the frame. The presented in [15] for iterative CE. The received prepro-
ratio of NZP-to-ZP OFDM symbols in the given transmit cessed OFDM signal as shown in Fig. 4(a) is consid-
frame can be represented by M1 : M2 . Note that M1 = 0 ered for iterative CE. The proposed algorithm uses two
corresponds to a ZP OFDM system, while M2 = 0 corre- received vectors, i.e., the k th received symbol rk and
sponds to an NZP OFDM system. Fig. 2 depicts the structure the extended received symbol r̃k as depicted in Fig.
of a ZP-NZP transmit frame for a TDS-OFDM system with 4(a). These are defined as rk = [r0k , r1k , ..., rN k
2 −1
] and
k k−1 k−1 k k k
M1 = M2 = 1. In Fig. 2, each GI is indicated whether the r̃ = [rN2 −c , ..., rN2 −1 , r0 , r1 , ..., rN2 −1 ] respectively. The
ZP sequence or the NZP sequence is used. extended signal after the m th iteration is depicted in Fig.
4(b). The iterative CE procedure is explained with the help
B. Preprocessing of ZP-NZP OFDM Frame of Fig. 4 in the following steps.
The algorithm presented in [15] is used for CE. The Preparation: An extended NZP sequence p̃ is defined
algorithm requires preprocessing of each received ZP-NZP by appending c zeros to p, and is expressed as p̃ =
frame. It is shown in Fig. 3(a) that the tail of the k th received [p̃0 , p̃1 , ..., p̃2c−1 ]. The frequency domain NZP sequence, i.e.,
OFDM symbol is free from any signal overlapping caused P̃ = [P̃0 , P̃1 , ..., P̃2c−1 ], can be computed by taking the FFT
by g. The reason for this overlapping-free tail is the ZP of p̃. A constant α = 0.125 is predefined to give a 12.5 %
transmitted signal can be expressed in the frequency domain
as S = [S0 , S1 , ..., SN2 −1 ] where Sj is computed from
Sj = Rj /H˜m j (4)