Parametric Channel Estimation for Pseudo-Random Tile-Allocation in Uplink OFDMA

We consider the uplink channel estimation of a multipath wireless channel used for orthogonal frequency division multiple access (OFDMA) transmission, where the uplink uses a pseudo-random ldquotilerdquo allocation pattern. A tile is made of small number of physically adjacent data subcarriers along with a few embedded pilot subcarriers and an uplink sub-channel allocated to an user in OFDMA systems such as IEEE 802.16d/e wireless MAN consists of several such pseudo-randomly chosen tiles. While the embedded pilots enable intra-tile channel interpolation, such an estimation will have an error floor which degrades performance substantially for highly frequency selective channels. We propose a parametric channel estimation method applicable to such irregular and sparsely spaced pilots, that does not exhibit an error-floor over the nominal operating range of signal to noise ratios, even for highly selective channels. The proposed algorithm exploits the pilot structure in each tile in estimating the delay subspace corresponding to the parametric channel description. Although this algorithm is more computationally complex when compared to the intra-tile linear interpolator, it offers a greatly enhanced bit-error probability (BEP) performance with a significantly lower pilot overhead. The uncoded BEP expression for the proposed estimator are analytically derived. Simulation results provided compares the mean squared error performance of this parametric channel estimator with the Cramer-Rao bound and also illustrates the significantly improved BEP performance over the existing methods.