Low-Complexity Algorithms for Channel Estimation in Optimised Pilot-Assisted Wireless OFDM Systems

Doctoral Thesis

2009

Permanent link to this Item
Authors
Supervisors
Journal Title
Link to Journal
Journal ISSN
Volume Title
Publisher
Publisher

University of Cape Town

License
Series
Abstract
Orthogonal frequency division multiplexing (OFDM) has recently become a dominant transmission technology considered for the next generation fixed and mobile broadband wireless communication systems. OFDM has an advantage of lessening the severe effects of the frequency-selective (multipath) fading due to the band splitting into relatively flat fading subchannels, and allows for low-complexity transceiver implementation based on the fast Fourier transform algorithms. Combining OFDM modulation with multilevel frequency-domain symbol mapping (e.g., QAM) and spatial multiplexing (SM) over the multiple-input multiple-output (MIMO) channels, can theoretically achieve near Shannon capacity of the communication link. However, the high-rate and spectrumefficient system implementation requires coherent detection at the receiving end that is possible only when accurate channel state information (CSI) is available. Since in practice, the response of the wireless channel is unknown and is subject to random variation with time, the receiver typically employs a channel estimator for CSI acquisition. The channel response information retrieved by the estimator is then used by the data detector and can also be fed back to the transmitter by means of in-band or out-of-band signalling, so the latter could adapt power loading, modulation and coding parameters according to the channel conditions. Thus, design of an accurate and robust channel estimator is a crucial requirement for reliable communication through the channel, which is selective in time and frequency. In a MIMO configuration, a separate channel estimator has to be associated with each transmit/receive antenna pair, making the estimation algorithm complexity a primary concern. Pilot-assisted methods, relying on the insertion of reference symbols in certain frequencies and time slots, have been found attractive for identification of the doubly-selective radio channels from both the complexity and performance standpoint. In this dissertation, a family of the reduced-complexity estimators for the single and multiple-antenna OFDM systems is developed. The estimators are based on the transform-domain processing and have the same order of computational complexity, irrespective of the number of pilot subcarriers and their positioning. The common estimator structure represents a cascade of successive small-dimension filtering modules. The number of modules, as well as their order inside the cascade, is determined by the class of the estimator (one or two-dimensional) and availability of the channel statistics (correlation and signal-to-noise power ratio). For fine precision estimation in the multipath channels with statistics not known a priori, we propose recursive design of the filtering modules. Simulation results show that in the steady state, performance of the recursive estimators approaches that of their theoretical counterparts, which are optimal in the minimum mean square error (MMSE) sense. In contrast to the majority of the channel estimators developed so far, our modular-type architectures are suitable for the reconfigurable OFDM transceivers where the actual channel conditions influence the decision of what class of filtering algorithm to use, and how to allot pilot subcarrier positions in the band. In the pilot-assisted transmissions, channel estimation and detection are performed separately from each other over the distinct subcarrier sets. The estimator output is used only to construct the detector transform, but not as the detector input. Since performance of both channel estimation and detection depends on the signal-to-noise power vi ratio (SNR) at the corresponding subcarriers, there is a dilemma of the optimal power allocation between the data and the pilot symbols as these are conflicting requirements under the total transmit power constraint. The problem is exacerbated by the variety of channel estimators. Each kind of estimation algorithm is characterised by its own SNR gain, which in general can vary depending on the channel correlation. In this dissertation, we optimise pilot-data power allocation for the case of developed low-complexity one and two-dimensional MMSE channel estimators. The resultant contribution is manifested by the closed-form analytical expressions of the upper bound (suboptimal approximate value) on the optimal pilot-to-data power ratio (PDR) as a function of a number of design parameters (number of subcarriers, number of pilots, number of transmit antennas, effective order of the channel model, maximum Doppler shift, SNR, etc.). The resultant PDR equations can be applied to the MIMO-OFDM systems with arbitrary arrangement of the pilot subcarriers, operating in an arbitrary multipath fading channel. These properties and relatively simple functional representation of the derived analytical PDR expressions are designated to alleviate the challenging task of on-the-fly optimisation of the adaptive SM-MIMO-OFDM system, which is capable of adjusting transmit signal configuration (e.g., block length, number of pilot subcarriers or antennas) according to the established channel conditions.
Description

Reference:

Collections