Techniques for adaptively adjusting decision levels of a PAM-N decision feedback equalizer

A method for adaptively adjusting a plurality of decision levels of a N-pulse amplitude modulation (PAM-N) decision feedback equalizer (DFE), comprising:

• setting feedback coefficients of the DFE;
• mitigating, using the feedback coefficients, effects of interference from data symbols that are adjacent in time to an input PAM-N data symbol;
• computing a least error respective of one of the plurality of decision levels, wherein the least error indicates a difference of a pseudo equalized input PAM-N data symbol from an optimal position of the one of the plurality of decision levels, wherein one of the plurality of decision levels corresponds to a modulation level used to modulate data in the input PAM-N data symbol, wherein computing the least error further comprises subtracting the pseudo equalized input PAM-N data symbol from each of the plurality of decision levels generated by the calibration unit to produce N error values, comparing two error values corresponding to two adjacent decision levels, generating N−1 decisions based on the comparison results, wherein the generated N−1 decisions are used to determine the least error out of the N computed error values, recovering the data based on the N−1 decisions, and selecting the least error out of the computed N error values;
• and
• adaptively setting the plurality of decision levels based, in part, on the least error, thereby compensating for gain changes resulted by a cable on which the input PAM-N data symbol is received.

A multi-level pulse amplitude modulation (PAM-N) decision feedback equalizer (DFE), comprising:

• a coefficient computation unit for setting feedback coefficients of the DFE;
• a feedback unit that mitigates, using the feedback coefficients, effects of interference from data symbols that are adjacent in time to an input PAM-N data symbol;
• an error-and-decision unit for at least computing a least error value respective to one of a plurality of decision levels, wherein the least error value indicates a difference of a pseudo equalized input PAM-N data symbol from an optimal position of the one of the plurality of decision levels, wherein the one of the plurality of decision levels corresponds to a modulation level used to modulate data in the input PAM-N data symbol;
• and
• a calibration unit for adaptively setting the plurality of decision levels based, in part, on the least error value, thereby compensating for gain changes resulting from a cable on which the input PAM-N data symbol is received and further compensating for embedded offsets of the error-and-decision unit, wherein the calibration unit is further configured to independently adjust each of the plurality of the decision levels, the adjustment of each of the plurality of the decision levels is determined by Sk(j)=Sk(j−1)+μ(sign(Ek(j))), wherein ‘k’ is the one of the plurality of decision levels, μ is an adaptation speed factor, ‘Ek’ is the least error;
• and ‘j’ is an index of the input PAM-N data symbol.