Performance evaluation of channel estimation techniques for MIMO-OFDM systems with adaptive sub-carrier allocation

摘要

Dynamic Sub-Carrier Allocation (DSA) strategies have been shown previously to achieve significant performance benefits when applied to OFDMA systems and further benefits for MIMOOFDMA systems. Analysis thus far has focussed on the assumption of ideal Channel State Information (CSI). In this paper, the impact of non-ideal CSI is investigated. Various channel estimation techniques are evaluated for application to MIMO-OFDMA systems. They are based on Least Squares (LS) Estimation with training pilots. ‘Conventional’ (as for MIMO-OFDM) CTP (combining training pilots and ‘improved’ (optimised for MIMO-OFDMA) STP (Separate training pilots) versions of both Frequency Domain Least Square (FDLS) and Time Domain Least Square (TDLS) channel estimation are considered, as are the options of both Space- Time Block coding (STBC) and Spatial Multiplexing (SM) as MIMO strategies. The STP-TDLS strategy is shown to significantly outperform other channel estimation options, achieving performance within 1dB of the ideal case. Subsequently, the use of the STP-TDLS channel estimation method in conjunction with the DSA algorithm is considered in order to determine the impact of non-ideal channel knowledge on the gain achieved by DSA. The performance for the cases of ideal CSI and CSI derived via STP-TDLS channel estimation are compared and evaluated for both the STBC and SM cases. The effects of non-ideal channel estimation in both the DSA mechanism and channel equalisation separately and together are evaluated. It is shown that STP-TDLS channel estimation works better in SM (only 1dB worse than ideal CSI case) than in STBC. Furthermore, it is shown that DSA is less sensitive than channel equalisation to nonideal CSI. The degradation of system performance in the realistic case of non-ideal CSI for both DSA and channel equalisation is a compound of the effects of the separate effects of non-ideal CSI error. It is shown here that in both STBC and SM cases, the effect is almost a linear addition of the two parts. Given the substantial benefits of DSA and its relative insensitivity to channel estimation errors, it is concluded that DSA remains a highly promising technique.