Dynamic algorithm transforms for low-power reconfigurable adaptive equalizers

In this paper, we present low-power reconfigurable adaptive equalizers derived via dynamic algorithm transforms (DAT's). The principle behind DAT is that conventional signal processing systems are designed for the worst case and are not energy-optimum on average. Therefore, significant energy savings can be achieved by optimally reconfiguring the hardware in these situations. Practical reconfiguration strategies for adaptive filters are-presented. These strategies are derived as a solution to an optimization problem. The optimization problem has energy as the objective function and a constraint on the algorithm performance (specifically the SNR), The DAT-based adaptive filter is employed as an equalizer for a 51.84 Mb/s very-high-speed digital subscriber loop (VDSL) over 24-pair BKMA cable. The channel nonstationarities are due to variations in cable length and number of far-end crosstalk (FEXT) interferers. For this application,:the traditional design is based on 1 kft cable length and 11 FEXT interferers. Tt was found that up to 81% energy savings can be achieved when cable length varies from 1-0.1 kft and the number of FEXT interferers varies from 11 to 4, On the average, 53% energy savings are achieved as compared with the conventional worst-case design.