Fixed-point refinement of OFDM-based adaptive equalizers: An heuristic approach

In this paper we address the quantization of adaptive equalizers for OFDM-based WLAN. Traditional simulation-based approaches require extremely long limes to quantize complex systems. We have tackled this issue threefold: i) generating a minimization wordlength procedure tuned for adaptive equalizers; ii) providing a quantization evaluation criterion that allows fast Monte-Carlo simulations with high relative errors; and iii) simulating only a subset of subcarriers to extract quantization information. One of the main results is that running several single-subcarrier simulations is possible to achieve acceptable results in simulation lime and performance degradation compared to the simulation of complete OFDM symbols. We use as a practical example a 220 Mbps OFDM RLS equalizer combined with ABL pre-equalization.