29-10-2012, 02:19 PM
A Class of Nonlinear Signal-Processing Schemes for Bandwidth-Efficient OFDM Transmission With Low Envelope Fluctuation
A Class of Nonlinear Signal.pdf (Size: 447.31 KB / Downloads: 42)
Abstract
This paper presents a wide class of digital signal-processing
schemes for orthogonal frequency-division multiplexing
(OFDM) transmission which combine a nonlinear operation in the
time domain and a linear filtering operation in the frequency domain.
The ultimate goal of these schemes is to reduce the envelope
fluctuation of ordinary OFDM, while keeping its high spectral
efficiency and allowing a low-cost, power-efficient implementation.
An appropriate statistical model concerning the transmitted
frequency-domain blocks is developed, which is derived from
well-established results on Gaussian stochastic processes distorted
by memoryless nonlinearities. This model can be employed for
performance evaluation by analytical means, with highly accurate
results whenever the corresponding conventional OFDM
signals exhibit quasi-Gaussian characteristics. Cases where the
signal-processing scheme is repeatedly used, in an iterative way,
are treated through an extension of the proposed statistical modeling.
A set of numerical results is presented and discussed so as
to show the practical interest of both the proposed schemes and
the analytical methods for evaluation of their performance. For
the sake of comparisons, this paper includes numerical results
concerning the partial transmit sequence technique, which is an alternative
peak-to-mean envelope power ratio-reducing technique
of higher complexity, often recommended due to its distortionless
nature. The superior performance/complexity tradeoffs through
the proposed class of nonlinear signal-processing schemes is
emphasized.
INTRODUCTION
AWELL-KNOWN, major drawback of conventional
orthogonal frequency-division multiplexing (OFDM)
transmission schemes [1] is their strong envelope fluctuation
and high peak-to-mean envelope power ratio (PMEPR),
leading to power amplification difficulties. In order to avoid the
out-of-band radiation levels which are inherent to nonlinear distortion,
power amplifiers for OFDM transmission are required
to have linear characteristics and/or a significant input backoff
has to be adopted. Therefore, a reduced power efficiency is the
price to pay for a high bandwidth efficiency.
PERFORMANCE RESULTS
In the following, we present a set of performance results
concerning the proposed class of nonlinear signal-processing
schemes. We consider an OFDM modulation with
subcarriers and an -QAM constellation, with a Gray mapping
rule, on each subcarrier (this value of is high enough
to allow the Gaussian approximation of the OFDM signals).
The set of multiplying coefficients
has a trapezoidal shape, with for the data subcarriers
(in-band region), dropping linearly to 0 along the first
out-of-band subcarriers at both sides of the in-band
region, which means nonzero subcarriers. The nonlinear
operation is chosen to be an ideal envelope clipping [see (1)].
Both transmit and receive analog filter characteristics have an
SRRC shape with and a one-sided bandwidth .
The transmitter employs a power amplifier which is quasi-linear
within the range of variations of the input envelope, and the
coherent detection operates under perfect synchronization and
channel estimation.
CONCLUSIONS AND COMPLEMENTARY REMARKS
In this paper, we presented and evaluated a wide class of digital
signal-processing schemes for OFDM transmission which
combine a nonlinear operation in the time domain, and a linear
filtering operation in the frequency domain. The ultimate goal of
these schemes is to reduce substantially the envelope fluctuation
of ordinary OFDM, while keeping its high spectral efficiency
and allowing a low-cost, power-efficient implementation.
An appropriate statistical model concerning the transmitted
frequency-domain blocks has been developed so as to provide
analytical support for a computationally efficient evaluation of
power-bandwidth tradeoffs. A detailed evaluation of OFDM
transmission techniques which employ the signal-processing
schemes considered here was carried out, involving computations
of power spectra, BER performances, and achieved
PMEPR values. Such evaluation has taken advantage of our
statistical characterization of the transmitted blocks and included
other implementation issues (such as the impact of the
time-windowing procedures). A set of performance results
was presented and discussed, showing that the proposed basic
schemes can provide a significant PMEPR reduction while
keeping a high spectral efficiency.