Hong-Yu Liu

Constant amplitude zero auto-correlation (CAZA) sequences and periodic sequences have been widely used as preambles for synchronization and channel estimation in the literature. In this paper, we first construct generalized CAZA (GCAZA) sequences in the time domain. Then, their Cramer-Rao bounds containing no matrix inversion for the purpose of joint carrier frequency offset and channel impulse response estimation are derived. The properties of auto-correlation and the discrete Fourier transform of a GCAZA are then presented analytically. The latter also explains the relationship between the conventional frequency-domain CAZA and the proposed time-domain GCAZA. Finally, we conduct numerical simulations and compare the results with existing sequences in the literature for fading channels.

We modify a LC-CFO estimator in the literature to make it easier to implement. By using the Lagrange interpolation, three low-order functions are used as candidates for replacing the arctangent function used in the LC-CFO estimator. The simplest one among these low-order functions sets its output directly equal to its input. In consequence, up to 20 iterations are needed for convergence with the carrier frequency offset up to 0.5. Moreover, the steady-state mean square errors are very close to the Cramer–Rao bounds at median-to-high signal-to-noise ratios. Those performances are about the same as when using a conventional LC-CFO estimator.

For joint maximum-likelihood (ML) frequency tracking and channel estimation using orthogonal frequency-division multiplexing (OFDM) training blocks in OFDM communications over mobile wireless channels, a major difficulty is the local extrema or multiple-solution complication arising from the multidimensional log-likelihood function. To overcome this, we first obtain crude ML frequency-offset estimators using single-time-slot samples from the received time-domain OFDM block. These crude frequency estimators are shown to have unique closed-form solutions. We then optimally combine these crude frequency estimators in the linear-minimum-mean-square-error (LMMSE) sense for a more accurate solution. Finally, by alternatively updating the LMMSE frequency estimator and the ML channel estimator through adaptive iterations, we successfully avoid the use of a multidimensional log-likelihood function, hence obviating the complex task of global solution search and, meanwhile, achieve good estimation performance. Our estimators have mean square errors (MSEs) tightly close to Cramer-Rao bounds (CRBs) with a wide tracking range.

It is a general consensus that an orthogonal space-time block code can achieve full diversity, and due to its orthogonal nature, the multiple-input-multiple-output (MIMO) maximum-likelihood (ML) decoding metrics can be decoupled into single-input-single-output (SISO) ML metrics based on linear processing at the receiver, thus greatly reducing the decoding complexity. In fact, nonorthogonal codes also currently exist that can achieve better symbol-error-rate performance without rate reduction and complexity increase for correlated fading channels. In this paper, we show by detailed derivations that nonorthogonal linear space-time block codes can also be decoded by ML decoupling through receiver linear processing. Our derived expressions for the decoupled ML metrics automatically contain the design information for the receiver linear processors.

For digital data transmission using an orthogonal frequency-division multiplexing (OFDM) over fading channels, the interchannel- interference (ICI) term caused by the Doppler spread is correlated with the desired signal term. Nonetheless, many researchers have assumed the ICI term to be uncorrelated with the desired signal term. In this paper, we present the symbol-error-rate (SER) performance for the M-quadratic amplitude-modulation OFDM systems over Ricean fading using the exact correlated ICI model. The results are then compared with the uncorrelated ICI results. We find that the two results present some differences. The differences tend to increase as the Ricean factor is increased.

We derive the symbol error probability (SEP) expressions for orthogonal space-time block codes with any form of generalized complex orthogonal design (GCOD) employing arbitrary rectangular quadratic-amplitude modulation signaling over correlative fading channels. By first decorrelating the physical branches into uncorrelated virtual branches, a general expression for the moment generating function of the received signal-to noise ratio (SNR) is obtained, from which the SEP is derived. The result can be applied to correlative fading channels with any form of Gaussian fading gains. Moreover, channels having joint fading models, as well as mixed channel powers, are considered. We also discover that, for certain GCOD codes, different information symbols may exhibit different SEPs. The above features are demonstrated by two GCOD examples used for simulations. Theoretical performance curves are compared with Monte Carlo simulated results in excellent agreement.

We present exact symbol error rate (SER) performance analysis for M-QAM OFDM systems over Ricean and Rayleigh fading is analyzed. Both slow and fast quasi-static fading as well as frequency-selective and -nonselective channels are considered.