Method and apparatus for efficient AMLD matrix computations

A method of calculating code-specific channel matrices for use in a multi-stage Assisted Maximum Likelihood Detection (AMLD) wireless communication receiver wherein a candidate symbol sets are successively refined in sequential computational stages, comprising:

• calculating a code-specific channel matrix for use at one or more stages by computing a low rate matrix T that is substantially constant over a plurality of symbol periods, wherein the low rate matrix T includes information on a net channel response and combining weights, wherein the combining weights are substantially constant over two or more AMLD receiver stages, and the low rate matrix T is computed once and used in two or more AMLD receiver stages;
• computing high rate matrices that vary with each received candidate symbol;
• and
• for each symbol period, combining the high rate matrices with the low rate matrix to generate the code-specific channel matrix for refining a candidate symbol set at a particular stage of the AMLD receiver.

A method of calculating code-specific channel matrices for use in a multi-stage Assisted Maximum Likelihood Detection (AMLD) wireless communication receiver wherein candidate symbol sets are successively refined in sequential computational stages, comprising:

• calculating a code-specific channel matrix for use at one or more stages by computing a low rate matrix T that is substantially constant over a plurality of symbol periods and combining weights, wherein computing the low rate matrix T comprises:
• transforming the combining weights, a time domain quantity, to frequency domain;
• transforming a net channel response, a time domain quantity, to frequency domain;
• computing a linear convolution in the frequency domain of the combining weights and the frequency domain of the net channel response to form a low rate frequency domain matrix;
• transforming the low rate frequency domain matrix to a low rate time domain matrix;
• and
• selecting elements from the low rate time domain matrix to form the low rate matrix T;
• computing high rate matrices that vary with each received candidate symbol;
• and
• for each symbol period, combining the high rate matrices with the low rate matrix T to generate the code-specific channel matrix for refining a candidate symbol set at a particular stage of the AMLD receiver.

A multi-stage Assisted Maximum Likelihood Detection (AMLD) wireless communication receiver operative to consider received candidate symbols and iteratively refine received candidate symbol sets by eliminating candidates at one or more stages, comprising:

• a first stage comprising a linear equalizer followed by a single user detector;
• subsequent intermediate stages, each comprising a block linear equalizer followed by a constrained joint detector;
• and
• a final stage comprising a Rake front end followed by a sequence detector;
• wherein a code-specific channel matrix is calculated for use at the one or more stages, by
• computing a low rate matrix T that is substantially constant over a plurality of symbol periods;
• computing high rate matrices that vary with each received candidate symbol;
• and
• for each symbol period, combining the high rate matrices with the low rate matrix T to generate the code-specific channel matrix for refining the received candidate symbol sets at a particular stage of the AMLD receiver.