Analysis and design of pilot-aided multicarrier systems over doubly selective channels with a local subcarrier processing constraint

In the near future, the need for portable multi-antenna wireless devices supporting high data-rates in a harsh mobile environment is anticipated. Typically, in this scenario, the underlying wireless channels are time- and frequency selective, hence doubly selective (DS). Multicarrier modulation (MCM) schemes, wherein a stream of data is split into sub-streams and transmitted on a set of subcarriers in parallel, is frequently used on wireless communication systems. DS channels generate both inter-symbol interference (ISI) as well as inter-carrier interference (ICI) for MCM transmissions, thereby making the design of MCM receivers a challenging task. The challenge is compounded by strict constraints on processing power as a result of the demand for light, portable, low-power devices, and by the lack of channel state information (CSI) in practical wireless receivers. On one hand, information-theoretic analysis of MCM systems has focused on the performance of optimal MCM reception without complexity constraints. On the other hand, practical MCM receiver designs use sub-optimal techniques requiring limited processing power. Moreover, the performance limits of these practical designs is seldom compared to what is promised by optimal reception. Local subcarrier processing (LSP) using CSI acquired via pilot-aided channel estimation is arguably the most popular practical MCM receiver design technique. The LSP constraint allows the use of only a small subset of local observations and dominant ICI coefficients to retrieve information transmitted on each subcarrier. We characterize the performance limits imposed by LSP-constrained reception on generic MCM schemes. Our approach provides a new framework within which the performance limits of practical reception strategies with processing complexity constraints can be characterized and compared to optimal reception for any MCM scheme. Short descriptions of the main issues addressed by this dissertation are listed below.; We characterize a lower-bound on the ergodic LSP-constrained achievable-rate of channel-estimation based generic multi-antenna MCM receivers operating over DS channels. In doing so, we assume the use of pilot-aided MCM transmission, and independent and identically distributed (i.i.d.) Gaussian codebooks. We use this lower-bound to evaluate the high-SNR spectral efficiency in two distinct regimes of high signal-to-noise ratio (SNR). In the regime in which practical wireless systems are likely to operate, we find that multiple receive antennas are both necessary and sufficient to achieve spectral efficiency when i.i.d. Gaussian codebooks are used. A single receive antenna is insufficient for LSP, whereas it is known to be sufficient for optimal reception. We also demonstrate that this lower-bound can be tightened by incorporating a larger number of observations and ICI coefficients into LSP at the expense of increased processing complexity. There is another regime of even higher SNRs in which practical wireless systems are unlikely to be operating. In this regime, residual ICI and ISI result in an achievable-rate ceiling.; This LSP-constrained achievable-rate metric is a versatile tool for analysis and design of MCM systems over DS channels. As an example, we use this metric to compare the performance of various MCM schemes under a reception complexity constraint. As yet another example, we design novel LSP-constrained achievable-rate maximizing beamforming and combining vectors for multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems operating on DS channels. We demonstrate that our schemes enjoy large gains over traditional approaches. Additionally, we find that our schemes are robust to the use of predicted (as opposed to perfect) CSI at the transmitter.; Channel estimation errors are a major hindrance to the performance of MCM reception over DS channels. We show that channel re-estimation coupled with successive-decoding can be ap...