Adjusting filter coefficients

Claim 2

A transversal filter apparatus comprising: K=N+n-1 successive shift register stages, where N and n are integers greater than one and m=N/n is an integer greater than zero, a first stage of said K successive stages having an input for receiving successive input signal samples; first register means for storing a first subset of N coefficients; second register means for storing a second subset of n coefficients; multiplier means responsive to said first and second register means for applying first and second ordered sets of K coefficients as weights to the contents of said K stages in respective first and second half cycles per input signal sample, the first N coefficients of said first set being said first subset and last n-1 coefficients of the first set being zero, the first N-1 coefficients of the second set being zero and the last n coefficients of said second set being said second subset; means for adding the thereby weighted contents of said K stages to produce a filter output signal in said first half cycle and to produce a coefficient adjusting output signal in said second half cycle; means for forming a difference between the filter output signal and a reference signal; means coupled to the second register means for updating the stored second subset of coefficients with said difference such that the second subset comprises the differences formed in the last n successive second half cycles; and means coupled to the first register means for adjusting m of the stored first subset of coefficients as a function of the coefficient adjusting output signal in a manner that each of the first subset of coefficients is adjusted only once for each n successive input samples received.

Claim 3

The apparatus of claim 1 wherein m=1.

Claim 4

The apparatus of claim 1 where m=2.

Claim 5

A transversal filter method comprising: receiving successive input signal samples at a first of K=N+n-1 successive shift register stages, where N and n are integers greater than one and m=N/n is an integer greater than zero; applying first and second ordered sets of K coefficients as weights to the contents of said K stages in respective first and second half cycles per input signal sample, the last n-1 coefficients of the first set and the first N-1 coefficients of the second set being zero; adding the thereby weighted contents of said K stages to produce a filter output signal in said first half cycle and to produce a coefficient adjusting output signal in said second half cycle; forming a difference between the filter output signal and a reference signal; updating the second set of coefficients with said difference such that the last n coefficients of said second set comprises the differences formed in the last n successive second half cycles; and adjusting m of the first N coefficients of the first set as a function of the coefficient adjusting output signal in a manner that each of the first N coefficients of the first set is adjusted only once for each n successive input samples received.

Claim 6

A transversal filter method comprising: receiving successive input signal samples at a first of K=N+n-1 successive shift register stages, where N and n are integers greater than one and m=N/n is an integer greater than zero; applying first and second ordered sets of K coefficients as weights to the contents of said K stages in respective first and second half cycles per input signal sample, the last n-1 coefficients of the first set and the first N-1 coefficients of the second set being zero; adding the thereby weighted contents of said K stages to produce a filter output signal in said first half cycle and to produce a coefficient adjusting output signal in said second half cycle; forming a difference between the filter output signal and a reference signal; updating the second set of coefficients with said difference such that the last n coefficients of said second set comprises the differences formed in the last n successive second half cycles; and adjusting m of the first N coefficients of the first set as a function of the coefficient adjusting output signal in a manner that each of the first N coefficients of the first set is adjusted only once for each n successive input samples received.

Claim 7

The method of claim 4 wherein m=1.

Claim 8

The method of claim 4 where m=2.

Claim 9

A decision feedback filter apparatus: a first group of K1=X+n-1 successive shift register stages and a second group of K2=X+1 successive shift register stages, where X, Y and n are integers greater than one, a first stage of the first group having an input for receiving successive input signal samples, a first stage of the second group having an input for receiving a decision feedback signal; register means for storing a first subset of X coefficients, a second subset of Y coefficients and a third subset of n coefficients; multiplier means responsive to said register means for applying first and second ordered sets of K1 coefficients as weights to the contents of said K1 stages of said first group and first and second ordered sets of K2 coefficients as weights to the contents of said K2 stages of said second group in respective first and second half cycles per input signal sample, the first X coefficients of the first set of K1 coefficients being said first subset, the first Y coefficients of the first set of K2 coefficients being said second subset and the last n-1 coefficients of both the first set of K1 coefficients and the second set of K2 coefficients being zero, the first X-1 coefficients of said second set of K1 coefficients and the first Y-1 coefficients of said second set of K2 coefficients being zero and the last n coefficients of both the second set of K1 coefficients and the second set of K2 coefficients being said third subset; means for adding the thereby weighted contents of said first and second groups of stages, to produce a common output signal in said first half cycle and to produce respective first and second coefficient adjusting output signals in said second half cycle; a decision element having an input fed by said common output signal and having an output at which is formed a decision output signal; means between the output of said decision element and the input of the second group of stages for forming said decision feedback signal from said decision output signal; means for forming a difference between the common output signal and the decision output signal; means coupled to said register means for updating the stored third subset of coefficients with said difference such that the third subset comprises the differences formed in the last n successive second half cycles; and means coupled to the register means for adjusting at least one of the stored first subset of coefficients as a function of the first coefficient adjusting output signal and for adjusting at least one of the stored second subset of coefficients as a function of the second coefficient adjusting output signal.

Claim 10

A decision feedback filter apparatus: a first group of K1=X+n-1 successive shift register stages and a second group of K2=X+1 successive shift register stages, where X, Y and n are integers greater than one, a first stage of the first group having an input for receiving successive input signal samples, a first stage of the second group having an input for receiving a decision feedback signal; register means for storing a first subset of X coefficients, a second subset of Y coefficients and a third subset of n coefficients; multiplier means responsive to said register means for applying first and second ordered sets of K1 coefficients as weights to the contents of said K1 stages of said first group and first and second ordered sets of K2 coefficients as weights to the contents of said K2 stages of said second group in respective first and second half cycles per input signal sample, the first X coefficients of the first set of K1 coefficients being said first subset, the first Y coefficients of the first set of K2 coefficients being said second subset and the last n-1 coefficients of both the first set of K1 coefficients and the second set of K2 coefficients being zero, the first X-1 coefficients of said second set of K1 coefficients and the first Y-1 coefficients of said second set of K2 coefficients being zero and the last n coefficients of both the second set of K1 coefficients and the second set of K2 coefficients being said third subset; means for adding the thereby weighted contents of said first and second groups of stages, to produce a common output signal in said first half cycle and to produce respective first and second coefficient adjusting output signals in said second half cycle; a decision element having an input fed by said common output signal and having an output at which is formed a decision output signal; means between the output of said decision element and the input of the second group of stages for forming said decision feedback signal from said decision output signal; means for forming a difference between the common output signal and the decision output signal; means coupled to said register means for updating the stored third subset of coefficients with said difference such that the third subset comprises the differences formed in the last n successive second half cycles; and means coupled to the register means for adjusting at least one of the stored first subset of coefficients as a function of the first coefficient adjusting output signal and for adjusting at least one of the stored second subset of coefficients as a function of the second coefficient adjusting output signal.

Claim 11

A decision feedback filter method comprising: receiving successive input signal samples at a first stage of a first group of K1=X+n-1 successive shift register stages, where X and n are integers greater than one; inputting a decision feedback signal at a first stage of a second group of K2=Y+n-1 successive shift register stages, where Y is an integer greater than one; storing a first subset of X coefficients, a second subset of Y coefficients and a third subset of n coefficients; applying first and second ordered sets of K1 coefficients as weights to the contents of said K1 stages of said first group and first and second ordered sets of K2 coefficients as weights to the contents of said K2 stages of said second group in respective first and second half cycles per input signal sample, the first X coefficients of the first set of K1 coefficients being said first subset, the first Y coefficients of the first set of K2 coefficients being said second subset and the last n-1 coefficients of both the first set of K1 coefficients and the second set of K2 coefficients being zero, the first X-1 coefficients of said second set of K1 coefficients and the first Y-1 coefficients of said second set of K2 coefficients being zero and the last n coefficients of both the second set of K1 coefficients and the second set of K2 coefficients being said third subset; adding the thereby weighted contents of said first and second groups of stages, to produce a common output signal in said first half cycle and to produce respective first and second coefficient adjusting output signals in said second half cycle; forming a decision output signal in response to the common output signal; forming said decision feedback signal in response to said decision output signal; forming a difference between the common output signal and the decision output signal; updating the stored third subset of coefficients with said difference such that the third subset comprises the differences formed in the last n successive second half cycles; adjusting at least one of the stored first subset of coefficients as a function of the first coefficient adjusting output signal; and adjusting at least one of the stored second subset of coefficients as a function of the second coefficient adjusting output signal.

Claim 12

A decision feedback filter method comprising: receiving successive input signal samples at a first stage of a first group of K1=X+n-1 successive shift register stages, where X and n are integers greater than one; inputting a decision feedback signal at a first stage of a second group of K2=Y+n-1 successive shift register stages, where Y is an integer greater than one; storing a first subset of X coefficients, a second subset of Y coefficients and a third subset of n coefficients; applying first and second ordered sets of K1 coefficients as weights to the contents of said K1 stages of said first group and first and second ordered sets of K2 coefficients as weights to the contents of said K2 stages of said second group in respective first and second half cycles per input signal sample, the first X coefficients of the first set of K1 coefficients being said first subset, the first Y coefficients of the first set of K2 coefficients being said second subset and the last n-1 coefficients of both the first set of K1 coefficients and the second set of K2 coefficients being zero, the first X-1 coefficients of said second set of K1 coefficients and the first Y-1 coefficients of said second set of K2 coefficients being zero and the last n coefficients of both the second set of K1 coefficients and the second set of K2 coefficients being said third subset; adding the thereby weighted contents of said first and second groups of stages, to produce a common output signal in said first half cycle and to produce respective first and second coefficient adjusting output signals in said second half cycle; forming a decision output signal in response to the common output signal; forming said decision feedback signal in response to said decision output signal; forming a difference between the common output signal and the decision output signal; updating the stored third subset of coefficients with said difference such that the third subset comprises the differences formed in the last n successive second half cycles; adjusting at least one of the stored first subset of coefficients as a function of the first coefficient adjusting output signal; and adjusting at least one of the stored second subset of coefficients as a function of the second coefficient adjusting output signal.