Reception of Short QAM Data Packets Using Sequential and Partial GLRT

Two methods are proposed to improve error rate performance for reception of short packets with quadrature amplitude modulation (QAM). Both approaches are based on the same idea as generalized likelihood ratio test (GLRT) but have reduced computational cost that scales linearly in both packet length and signal constellation size. The first method, sequential GLRT, is based on demodulation of QAM symbols sequentially using GLRT decision rule and treating previously demodulated symbols as known data. The second method, partial GLRT, uses partial search over a portion of packet while remaining data symbols are demodulated with hard decisions, and Euclidean distance between these decisions and received signal samples is used as a decision metric to improve detection reliability. Computer simulation results showed that partial GLRT provides bit error rate (BER) very close to achievable limit even for very short QAM packets where the number of data symbols is equal to the QAM constellation size. The loss at high signal-to-noise ratio is about 0.3 dB for 16-QAM and about 0.085 dB for 64-QAM. For sequential GLRT, BER performance is worse, but this method has lower computational cost and can be useful for choosing a suitable trade-off between BER performance and complexity.