Near-optimal training sequences for MIMO OFDM systems with nulled subcarriers

In MIMO OFDM systems, the quality of channel estimation becomes important, especially for a large number of antennas. On the other hand, it is desirable to minimise the overhead, making the design of training sequences a crucial task. Although this problem has been solved efficiently for OFDM systems using all subcarriers, it is still an open question for systems with nulled subcarriers (such as IEEE802.11a and IEEE802.11n), where some of the subcarriers are left unmodulated in order to fit specified spectrum masks. In this paper we present a numerical method for designing MIMO OFDM training sequences for arbitrary systems using nulled subcarriers. It is shown that substantial gains can be had over simple training sequences, such as only transmitting on non-overlapping subcarriers. The performance of these new training sequences are measured in both mean-squared error and bit-error rate