Adaptive bitrate maximizing TEQ design for DMT-based systems

In a previous paper, we proposed a bitrate maximizing (BM) design criterion for the time-domain equalizer (TEQ) in a discrete multitone receiver. This BM-TEQ and the closely related BM per-group equalizers (PGEQ) get close to the performance of the so-called per-tone equalization (PTEQ). In this paper, we show that the BM-TEQ criterion, despite its nonlinear nature, is well suited for a recursive Levenberg-Marquardt (RLM) based design. This adaptive BM-TEQ also allows us to track slow variations of the transmission channel and the noise. This RLM-based design uses the same second-order statistics (SOS) as the earlier presented recursive least-squares (RLS) based adaptive PTEQ and opens up a complete range of adaptive BM equalizers: from the computationally efficient RLS-based PTEQ with largest memory cost, over the RLM-based BM-PGEQ with intermediate memory cost, towards an RLM-based BM-TEQ with considerably smaller memory cost, but larger equalizer updating complexity.