Relaxed, more optimum training for modems and the like
View Patent ↗A high-speed broadband, wireline modem including an adaptive equalizer having both a training mode and a decision-directed non-training mode. The adaptive equalizer comprising at least one of a forward path coupled to receive signal samples, the forward path including a forward filter and a decision element, and a feedback path coupled between an output of the decision element and an input of the decision element, the feedback path including a feedback filter; and means for adapting the one of said forward filter and said feedback filter based on a least squares error criterion performed substantially according to a predetermined algorithm.
1. A high-speed, broadband, wireline modem including an adaptive equalizer having both a training mode and a decision-directed non-training mode, the adaptive equalizer comprising:
at least one of: a forward path coupled to receive signal samples, the forward path including a forward filter and a decision element, and a feedback path coupled between an output of the decision element and an input of the decision element, the feedback path including a feedback filter; and
means for adapting the one of said forward filter and said feedback filter based on a least squares error criterion performed substantially according to the following computation:
e p =e (1 −K T X fast )
wherein e is a forward prediction error, K is a Kalmann Gain and X fast is an input vector.
2. The adaptive equalizer of claim 1 , further comprising a memory for storing the received signal samples.
3. The adaptive equalizer of claim 1 , wherein the means for adapting operates during said decision-directed non-training mode.
4. The adaptive equalizer of claim 1 , wherein adaptation is performed using fixed-point arithmetic operations.
5. A high-speed, broadband, wireline modem including an adaptive equalizer having both a training mode and a decision-directed non-training mode, the adaptive equalizer comprising:
at least one of: a forward path coupled to receive signal samples, the forward path including a forward filter and a decision element, and a feedback path coupled between an output of the decision element and an input of the decision element, the feedback path including a feedback filter; and
means for adapting the one of said forward filter and said feedback filter based on a least squares error criterion performed substantially according to the following computation:
F
fast
=
λ
i
F
fast
c
n
=
F
fast
e
p
1
+
e
T
F
fast
e
P
F
fast
=
F
fast
-
c
n
e
T
F
fast
b
n
=
K
fast
+
A
fast
c
n
wherein F fast , A fast and c are filter coefficients, e is a forward error prediction, λ is an error criterion, and b is a backward error predictor.
6. The adaptive equalizer of claim 5 , further comprising a memory for storing said received signal samples.
7. The adaptive equalizer of claim 5 , wherein the means for adapting operates during said decision-directed non-training mode.
8. The adaptive equalizer of claim 5 , wherein adaptation is performed using fixed-point arithmetic operations.
9. A high-speed, broadband, wireline modem including an adaptive equalizer having both a training mode and a decision-directed non-training mode, the adaptive equalizer comprising:
at least one of: a forward path coupled to receive signal samples, the forward path including a forward filter and a decision element, and a feedback path coupled between an output of the decision element and an input of the decision element, the feedback path including a feedback filter; and
means for adapting the one of said forward filter and said feedback filter based on a least squares error criterion performed substantially according to the following computation:
K fast −( m− ( D fast μ))/(1−η r μ)
D fast −D fast −K fast η r
wherein K fast is a Kalmann Gain, D fast is a backward predictor coefficient and m, μ, and η are backward prediction errors.
10. The adaptive equalizer of claim 9 , further comprising a memory for storing said received signal samples.
11. The adaptive equalizer of claim 9 , wherein the means for adapting operates during said decision-directed non-training mode.
12. The adaptive equalizer of claim 9 , wherein adaptation is performed using fixed-point arithmetic operations.