29-10-2012, 02:33 PM
Joint Turbo Equalization and Multiuser Detection of MC-CDMA Signals With Low Envelope Fluctuations
Joint Turbo Equalization.pdf (Size: 716.19 KB / Downloads: 28)
Abstract
In this paper, we consider the uplink transmission in
multicarrier code-division multiple-access (MC-CDMA) systems.
As other multicarrier signals, MC-CDMA signals have high envelope
fluctuations and a high peak-to-mean envelope power ratio
(PMEPR), which leads to amplification difficulties. This is particularly
important for the uplink transmission, since an efficient
low-cost power amplification is desirable at the mobile terminals
(MTs). Moreover, the transmission over time-dispersive channels
destroys the orthogonality between spreading codes, which might
lead to significant multiple-access interference levels. To reduce
the envelope fluctuations of the transmitted signals, while maintaining
the spectral efficiency, the MC-CDMA signal associated to
eachMTis submitted to a clipping device, followed by a frequencydomain
filtering operation. However, the nonlinear distortion effects
can be high when an MC-CDMA transmitter with reduced
envelope fluctuations is intended (e.g., a small clipping level and/or
when successive clipping and filtering operations are employed).
In this paper, we define an iterative receiver that jointly performs
a turbo multiuser detection and the estimation and cancellation
of the nonlinear distortion effects. Our performance results show
that the proposed receiver structure allows good performances,
very close to the linear receiver ones, even for high system load
and/or when a PMEPR as low as 1.7 dB is intended for each MT.
INTRODUCTION
MULTICARRIER code-division multiple-access (MCCDMA)
schemes combine an orthogonal frequencydivision
multiplexing (OFDM) modulation [1] with a
code-division multiple-access (CDMA) scheme [2]. Spreading
is performed in the frequency domain, and MC-CDMA
schemes are promising candidates for future broadband wireless
systems. Since the transmission over time-dispersive channels
destroys the orthogonality between spreading codes, a
frequency-domain equalizer (FDE) optimized under an MMSE
criterion is usually employed at the receiver [3], [4]. Since an
MMSE FDE does not perform an ideal channel inversion, we
are not able to fully orthogonalize the different spreading codes,
which can lead to severe interference levels, particularly for
fully loaded systems and/or when different powers are assigned
to different spreading codes. To improve the performance,
several receivers that perform turbo multiuser detection (MUD)
were proposed [5]–[7]. In [5], the use of soft information for
interference cancellation is exploited in MC-CDMA systems.
An iterative semiblind receiver for coded MC-CDMA systems,
able to deal with intracell and intercell interference, is proposed
in [6]. A novel low-complexity parallel interference cancellation
(PIC) receiver for turbo coded MC-CDMA systems is also
investigated in [7]. A promising iterative receiver for multicode
MC-CDMA signals was proposed in [8], based on the iterative
block decision-feedback equalizer (IB-DFE) concept [9]–[11],
allowing significant performance improvements.
NONLINEAR TRANSMITTER AND
RECEIVER STRUCTURES
To reduce the envelope fluctuations of the transmitted signals,
we employ the transmitter structure depicted in Fig. 2(a),
which is based on the nonlinear signal processing schemes
proposed in [15] for reducing the PMEPR of OFDM signals
while maintaining the spectral efficiency of conventional
OFDM schemes. Each time-domain sample is submitted to a
nonlinear device so as to reduce the envelope fluctuations on the
transmitted signal [see Fig. 2(b)].We assume that the nonlinear
device is an ideal envelope clipping with clipping level sM.
After a DFT operation, the clipped signal is then submitted to
a frequency-domain filtering procedure, through the set of multiplying
coefficients {Gk, k = 0, 1, . . .,N − 1}, to reduce the
out-of-band radiation levels inherent to the nonlinear operation.
Since the frequency-domain filtering procedure produces some
envelope fluctuation regrowth, limiting the achievable PMEPR,
C&F operations can be iteratively repeated so as to further
reduce the PMEPR of the transmitted signals.
IMPLEMENTATION COMPLEXITY ISSUES
The implementation complexity of our receivers can be
measured in terms of the number and size of DFT/inverse
DFT (IDFT) operations, the number of despreading/spreading
operations, and the computation charge required for the calculation
of the feedforward coefficients. In the case of the
IMUD receiver, we need L size-N DFT operations, one for
each antenna, and a pair of despreading/spreading operations
for the detection of each MT at each iteration (except for the
first iteration where only one despreading operation for each
MT is required). If we have estimation and compensation of
nonlinear effects, XP pairs of size-N DFT/IDFT operations
(X is the number of C&F operations) for the detection of each
MT at each iteration are also needed. As for the computation
of the feedforward coefficients, we need to invert the P × P
matrix in (37) for each MT at each iteration. Naturally, for slowvarying
channels, this operation is not required for all blocks. In
the case of the turbo-MUD receiver, the SISO channel decoding
needs to be implemented in the detection process of each MT,
with the soft-output Viterbi algorithm instead of a conventional
Viterbi algorithm. This can be the most complex part of the
turbo-MUD receiver.
PERFORMANCE RESULTS
In this section, we present a set of performance results
concerning the iterative receiver structures proposed in this
paper for the uplink of MC-CDMA systems with frequencydomain
spreading. We consider M = 32 data symbols for
each user, corresponding to blocks with length N = KM =
256, plus an appropriate CP. QPSK constellations, with Gray
mapping, are employed. To reduce the envelope fluctuations
of the transmitted signals (and the PMEPR) while maintaining
the spectral occupation of conventional MC-CDMA schemes,
each MT employs the clipping techniques combined with a
frequency-domain filtering operation proposed in [15] [the
power amplifiers are assumed to be linear for the (reduced)
dynamic range of the envelope fluctuations of the transmitted
signals]. The PMEPR of the transmitted signals (defined as in
[15]) are shown in Table I, together with the corresponding
average signal-to-nonlinear-self-interference ratio (SIR) values.
The receiver (i.e., the BS) knows the characteristics of the
PMEPR-reducing signal processing technique employed by
each MT.
CONCLUSION
In this paper, we have considered the uplink transmission
in MC-CDMA systems employing clipping techniques so as
to reduce the envelope fluctuations of the transmitted signals.
We have proposed an iterative receiver structure that combines
turbo MUD and the estimation and cancellation of the nonlinear
distortion effects that are inherent to the transmitted signals.
Our performance results have shown that the use of the channel
decoder outputs instead of the coded MUD outputs in the
feedback loop allows a significant performance improvement
at low and moderate SNRs, even for severely time-dispersive
channels and/or when a very low PMEPR MC-CDMA transmission
is intended.