Time domain equalizer for DMT modulation

A method of recovering a signal from a transmitted analog signal, comprising the steps of:

• receiving, over a transmission channel, an analog signal corresponding to modulated digital symbols in a plurality of frequency subchannels;
• analog-to-digital converting the analog signal into a discrete datastream;
• applying a time-domain equalizer to the datastream, the time-domain equalizer corresponding to a digital filter having coefficients derived according to a mean-squared error minimization that is constrained according to a spectral flatness;
• applying a discrete Fourier transform to the equalized datastream to recover symbols from each of the plurality of subchannels;
• applying a frequency domain equalizer to remove a channel response of the transmission channel;
• and
• after the step of applying a time-domain equalizer and before the step of applying a discrete Fourier transform, removing a cyclic prefix from each of a plurality of blocks of symbols in the datastream.

A transceiver for discrete multitone communications, comprising:

• means for receiving a training signal;
• means for regenerating the training signal in the receiving modem;
• means for operating a digital circuit in the receiving modem to derive coefficients for a time-domain equalizer digital filter from the received and regenerated training signals, by performing a minimization process constrained by a spectral flatness term in the frequency domain of the equalizer characteristic;
• and
• means for storing a filter vector based on the derived coefficients, for use in a time-domain equalizer digital filter, wherein means for operating digital circuit comprises:
• means for deriving correlation matrices from the received and regenerated training signals;
• means for deriving eigenvectors and eigenvalues for a combination of the correlation matrices;
• means for evaluating a cost function including a mean squared error term and a weighted spectral flatness term, the evaluating being performed using the derived eigenvectors and eigenvalues, over multiple trial values of a weighting constant;
• means for selecting a value of the weighting constant that provides a minimum value of the cost function;
• and
• means for deriving an optimal filter vector using the selected weighting constant value.

A transceiver comprising:

• means for receiving, over a transmission channel, an analog signal corresponding to modulated digital symbols in a plurality of frequency subchannels;
• means for analog-to-digital converting the analog signal into a discrete datastream;
• means for applying a time-domain equalizer to the datastream, the time-domain equalizer corresponding to a digital filter having coefficients derived according to a mean-squared error minimization that is constrained according to a spectral flatness;
• means for applying a discrete Fourier transform to the equalized datastream to recover symbols from each of the plurality of subchannels;
• means for applying a frequency domain equalizer to remove a channel response of the transmission channel;
• and
• means for removing a cyclic prefix from each of a plurality of blocks of symbols in the datastream, wherein the cyclic prefix is removed after applying a time-domain equalizer and before applying a discrete Fourier transform.