Time domain automatic equalizer with frequency domain control

Claim 2

An equalizer comprising: (a) a transversal filter having a plurality of taps separated by time delay intervals t, said transversal filter receiving input signals from a transmission channel having a distortion transfer function D(w) and an undistorted transfer function H(w), said transversal filter providing output signals, (b) means for sampling said output signals from said transversal filter at spaced time intervals t, (c) means for transforming said sampled output signals into frequency representations thereof, (d) means for providing frequency representations associated with the undistorted transfer function H(w) of the transmission channel, (e) means responsive to said frequency representations of said sampled output signals and said frequency representations associated with said undistorted transfer function for providing samples of a transfer function substantially equal to 1/D(w), (f) means for converting said samples of said transfer function substantially equal to 1/D(w) into a time domain sample set, and (g) means for applying said time domain sample set as weights for said taps for providing time domain equalization for said transmission channel.

Claim 3

An equalizer as recited in claim 1 wherein said means for transforming said sampled output signals into frequency representations comprise means for generating the discrete Fourier transform of said sampled output signals, and said means for converting said samples of said transfer function substantially equal to 1/D(w) into a time domain sample set comprises means for generating the inverse discrete Fourier transform.

Claim 4

An equalizer as recited in claim 2 wherein said transversal filter has N taps, where N is an even integer, said input signal is real, said means for sampling the output signal from said transversal filter comprises means for sampling N samples, each sample spaced by a time interval t, and said means for generating the discrete Fourier transform comprises means for generating spectral components X n , where n has values in one of the groups consisting of n=0, 1 . . . N/2 and n=0, N/2, N/2+ 1 . . . N-1.

Claim 5

An equalizer as recited in claim 3 wherein spectral components X 0 and X N/2 are real.

Claim 6

An equalizer as recited in claim 1 wherein said means for providing samples of said transfer function substantially equal to 1/D(w) comprises means for generating samples of D*(w)/((RD).sup. 2 + (ID) 2 ), where RD and ID are real and imaginary parts of the transfer function D(w), and D*(w) is the complex conjugate of D(w).

Claim 7

An equalizer as recited in claim 5 wherein said means for computing samples of D*(w)/((RD) 2 + (ID) 2 ) comprises: (a) means for generating samples of D*(w) upon receipt of one input signal by said transversal filter, (b) means for generating samples of 1/((RD) 2 + (ID) 2 ) upon receipt of another input signal by said transversal filter, and (c) means for generating samples of the product D*( w).sup.. 1/((RD) 2 + (ID) 2 )= 1/D( w).

Claim 8

An equalizer as recited in claim 6 wherein said means for generating samples of D*(w) comprises means for changing the algebraic sign of ID.

Claim 9

An equalizer as recited in claim 6 wherein said means for generating samples of 1/((RD) 2 + (ID) 2 ) comprises feedback circuit means for comparing a frequency representation of said received another input signal with a frequency representation of said undistorted transfer function H(w).

Claim 10

An equalizer as recited in claim 6 wherein said one and another input signals comprise predetermined training signals.

Claim 11

An equalizer as recited in claim 6 wherein said transversal filter has N taps, and said means for sampling said output signals from said transversal filter comprises means for sampling N samples and said means for transforming said sampled output samples into frequency representations comprises means for generating spectral components, X n where, n is a frequency index and n=0, 1 . . . N-1, said equalizer further comprising means for storing a sample of D*(w) and 1/((RD) 2 + (ID) 2 ) for each value n.

Claim 12

An equalizer as recited in claim 10 wherein said means for transforming said sampled output signals comprises means for generating the discrete Fourier transform, said output signals are real functions of time and n has values in only one of the groups consisting of n=0, 1 . . . N/2 and n=0, N/2, N/2+1 . . . N-1, where N is an even integer.

Claim 13

An equalizer as recited in claim 11 wherein spectral components X 0 and X N/2 are real.

Claim 14

Apparatus for adjusting the tap gains of a transversal filter having an input and an output in equalizing distortion through a transmission channel upon response to a sequence of at least two predetermined training pulses, H(w) being the undistorted transfer function of the transmission channel and D(w) being the distortion transfer function and H(w).sup.. D(w) being the transfer function of the transmission channel, said tap gains adjusted to correspond to a transfer function of the transversal filter substantially given by 1/D(w)= D*(w)/((RD) 2 + (ID) 2 ), where, (D*(w) is the complex conjugate of D(w), RD is the real part of D(w) and ID is the imaginary part of D(w), said apparatus comprising: (a) means for setting the tap gains of said transversal filter to correspond to samples of the transfer function 1/H(w), whereby the output of said transversal filter has a frequency spectrum H(w).sup.. D(w).sup.. 1/H(w)= D(w), (b) means responsive to receipt of said first training pulse from said transversal filter for generating values corresponding to samples of D*(w) and for storing same, (c) means for setting the tap gains of said transversal filter to correspond to samples of the transfer function D*(w) whereby the output of said transversal filter has a frequency spectrum H(w).sup..D( w).sup.. D*(w)= RH((RD) 2 + (ID) 2 )+ jIH((RD).sup. 2 + (ID) 2 ) where j= √-1 and RH and IH are samples of the real and imaginary parts respectively of the undistorted transfer function H(w), (d) means responsive to receipt of said second training pulse from said transversal filter and to samples of said undistorted transfer function H(w) for generating and storing values corresponding to samples of 1/((RD) 2 + (ID).sup. 2), (e) means for multiplying together said stored samples D*(w) and 1/((RD).sup. 2 + (ID).sup. 2) to obtain samples of 1/D( w), and (f) means for setting the tap gains of said transversal filter to correspond to the samples of transfer function 1/D( w) whereby the time domain equalization of received signals from said transmission channel is produced in said transversal filter.

Claim 15

Apparatus for adjusting the tap gains of a transversal filter having an input and an output in equalizing distortion through a transmission channel upon response to a sequence of at least two predetermined training pulses, H(w) being the undistorted transfer function of the transmission channel and D(w) being the distortion transfer function and H(w).sup.. D(w) being the transfer function of the transmission channel, said tap gains adjusted to correspond to a transfer function of the transversal filter substantially given by 1/D(w)= D*(w)/((RD) 2 + (ID) 2 ), where, (D*(w) is the complex conjugate of D(w), RD is the real part of D(w) and ID is the imaginary part of D(w), said apparatus comprising: (a) means for setting the tap gains of said transversal filter to correspond to samples of the transfer function 1/H(w), whereby the output of said transversal filter has a frequency spectrum H(w).sup.. D(w).sup.. 1/H(w)= D(w), (b) means responsive to receipt of said first training pulse from said transversal filter for generating values corresponding to samples of D*(w) and for storing same, (c) means for setting the tap gains of said transversal filter to correspond to samples of the transfer function D*(w) whereby the output of said transversal filter has a frequency spectrum H(w).sup..D( w).sup.. D*(w)= RH((RD) 2 + (ID) 2 )+ jIH((RD).sup. 2 + (ID) 2 ) where j= √-1 and RH and IH are samples of the real and imaginary parts respectively of the undistorted transfer function H(w), (d) means responsive to receipt of said second training pulse from said transversal filter and to samples of said undistorted transfer function H(w) for generating and storing values corresponding to samples of 1/((RD) 2 + (ID).sup. 2), (e) means for multiplying together said stored samples D*(w) and 1/((RD).sup. 2 + (ID).sup. 2) to obtain samples of 1/D( w), and (f) means for setting the tap gains of said transversal filter to correspond to the samples of transfer function 1/D( w) whereby the time domain equalization of received signals from said transmission channel is produced in said transversal filter.

Claim 16

Apparatus as recited in claim 13 further comprising means for sampling an output of said transversal filter and for providing the discrete Fourier transform of said sampled outputs.

Claim 17

Apparatus as recited in claim 14 wherein said transversal filter has N taps spaced apart by a time interval t and said sampling means comprises means for sampling N samples of said transversal filter output, each sample apart by a time interval t.

Claim 18

Apparatus as recited in claim 15 wherein means for setting the tapped gains of said transversal filter to correspond to samples of the transfer function 1/H(w) comprises means for generating samples of the inverse discrete Fourier transform of /H1/H(w).

Claim 19

Apparatus as recited in claim 15 wherein said means for setting the tapped gains of said transversal filter to correspond to samples of the transfer function D*(w) comprises means for generating samples of the inverse discrete Fourier transform of D*(w).

Claim 20

Apparatus as recited in claim 15 wherein said means for setting the tapped gains of said transversal filter to correspond to samples of the transfer function 1/D( w) comprises means for providing samples of the inverse discrete Fourier transform of 1/D( w).

Claim 21

Apparatus as recited in claim 13 wherein said means for generating and storing values corresponding to samples of 1((RD).sup. 2 + (ID).sup. 2) comprises feedback circuit means for comparing a frequency representation of said second training pulse with samples of said undistorted transfer function H(w).

Claim 24

A method as recited in claim 20 wherein the step of transforming said sampled outputs into a frequency representation comprises generating the discrete Fourier transform of said sampled outputs.

Claim 25

A method as recited in claim 21 wherein the step of converting samples of said inverse distortion transfer function into the time domain comprises generating the inverse discrete Fourier transform.

Claim 26

A method as recited in claim 20 wherein the step of generating frequency representations associated with an undistorted transfer function H(w) comprises the step of generating frequency representations of the inverse undistorted transfer function 1/H( w).

Claim 27

A method as recited in claim 23, wherein said inverse distortion transfer function is represented by 1/D( w) and the step of generating sample values thereof comprises: (a) means for setting the tap gains of said transversal filter to correspond to samples of the transfer function 1H( w), whereby the output of said transversal filter has a frequency spectrum H(w) ).sup.. D( w) .sup.. 1/H( w)= D(w), (b) transmitting a first training signal through said transmission channel, (c) generating samples of D*(w) upon receipt of said first training signal where D*(w) is the complex conjugate of D(w), (d) setting the tap gains of said transversal filter to correspond to samples of the transfer function D*(w) whereby the output of said transversal filter has a frequency spectrum: H(w).sup. . D(w).sup. . D*(w)= RH((RD).sup. 2 +(ID) 2 )+ jIH((RD) 2 + (ID).sup. 2), (e) storing values corresponding samples of D*(w), (f) transmitting a second training signal through said transmission channel, (g) generating values corresponding to samples of 1/((RD) 2 )+ (ID).sup. 2) in response to receipt of said second training signal and samples of said undistorted transfer function H(w), and (h) multiplying together said stored values corresponding to samples D*(w) and 1/((RD) 2 + (ID).sup. 2) to generate samples of said inverse transverse function 1D( w).