Method and apparatus for calibrating data-dependent noise prediction
View Patent ↗Disclosed herein is an apparatus and method of calibrating the parameters of a Viterbi detector 138 in which each branch metric is calculated based on noise statistics that depend on the signal hypothesis corresponding to the branch. An offline algorithm for calculating the parameters of data-dependent noise predictive filters 304 A-D is presented which has two phases: a noise statistics estimation or training phase, and a filter calculation phase. During the training phase, products of pairs of noise samples are accumulated in order to estimate the noise correlations. Further, the results of the training phase are used to estimate how wide (in bits) the noise correlation accumulation registers need to be. The taps [ t 2 [k] , t 1 [k] , t 0 [k] ] of each FIR filter are calculated based on estimates of the entries of a 3-by-3 conditional noise correlation matrix C [k] defined by C ij [k] =E ( n i−3 n j−3 |NRZ condition k ).
1. A method of calibrating a Viterbi detector, said Viterbi detector comprising at least one noise predictive filter, said method comprising:
(a) obtaining noise samples in a training phase;
(b) averaging said noise samples;
(c) estimating entries of a 3-by-3 conditional noise matrix C [k] defined by
C
ij
[
k
]
=
E
(
n
i
-
3
n
j
-
3
|
NRZ
condition
k
)
;
and
(d) calculating at least one tap of said at least one noise predictive filter based on said estimated entries.
2. The method of claim 1 , further comprising:
(e) increasing an amount of said noise samples obtained to increase accuracy of said at least one noise predictive filter.
3. The method of claim 1 , further comprising:
(e) determining an expected mis-prediction cost;
(f) estimating an amount of training samples required to achieve said expected mis-prediction cost;
(g) estimating an expected noise product magnitude;
(h) calculating a bit width of a noise product accumulator based on said expected noise product magnitude;
(i) generating a pseudo random bit sequence (“PRBS”) of non-return-to-zero (“NRZ”) bits synchronized with said noise samples;
(j) matching said PRBS with a NRZ condition block;
(k) accumulating said matching of said PRBS into said noise product accumulator; and
(l) counting how many of said matching of said PRBS are accumulated.
4. An apparatus for calibrating a Viterbi detector comprising at least one noise predictive filter, said apparatus comprising:
a tap generator operative to generate at least one tap coefficient for said at least one noise predictive filter based on data samples obtained during off line training, said tap coefficient representative of a noise correlation estimate;
said tap generator further comprising a tap calculator operative to compute said at least one tap coefficient based on a 3-by-3 conditional noise matrix C [k] defined by
C
ij
[
k
]
=
E
(
n
i
-
3
n
j
-
3
|
NRZ
condition
k
)
.
5. The apparatus of claim 4 , further comprising:
a noise sampler operative to obtain said data samples;
a linear feedback shift register (“LFSR”) operative to generate a periodic pseudo random bit sequence (“PRBS”) of non-return to zero (“NRZ”) bits synchronized with said data samples;
a condition selector coupled with said noise sampler and said LFSR and operative to select one of said data samples which matches said PRBS NRZ bits;
a statistic accumulator coupled with said noise sampler and said condition selector and operative to receive and accumulate said selected one of said data samples;
a statistic selector coupled with said condition selector and said statistic accumulator and operative to enable said selected data sample to be accumulated by said statistic accumulator;
a statistic counter coupled with said statistic accumulator and operative to count a number of said one of said data samples accumulated; and
wherein said tap calculator further bases said at least one tap coefficient computation on said accumulated of said one of said data samples.
6. An apparatus for calibrating a Viterbi detector, said Viterbi detector comprising at least one noise predictive filter, said apparatus comprising:
means for obtaining noise samples in a training phase;
means for averaging said noise samples;
means for estimating entries of a 3-by-3 conditional noise matrix C [k] defined by
C
ij
[
k
]
=
E
(
n
i
-
3
n
j
-
3
|
NRZ
condition
k
)
;
and
means for calculating at least one tap of said at least one noise predictive filter based on said estimated entries.