Patent Grant
7982985
1. Field of the Invention
This disclosure relates generally to data storage systems, and more specifically to a read channel in a hard disk drive (HDD).
2. Description of Related Art
Storage devices like hard disk drives are widely used in electronic devices, such as computers, MP3 players, video recorders, digital cameras and set-top boxes which need to store a large amount of data.
r=c*q (1)
wherein r is a number of equalized samples at the output of the FIR equalizer 204 and c is the FIR response.
The FIR response c may be any length. In one example, the length of c is 10, i.e., c has 10 taps and c=(c0, c1, c2, c3, c4, c5, c6, c7, c8, c9).
The Viterbi detector 205 detects data x from the disk 110 and outputs binary data at its output. To enable the Viterbi detector 205 to detect data x, the FIR equalizer 204 needs to make equation (2) roughly true:
r=x*target polynomial (2)
The target polynomial is chosen for a particular HDD and is programmable. One example of the target polynomial is [1, 0, −1]. Other target polynomials are, of course, possible.
The FIR response c may be initialized, but initial values of the taps of c may not be good enough to realize equation (2). A feedback loop including the reconstruction filter 206, the subtractor 207 and the adaptive logic 208 may be used to adapt the response of the FIR equalizer 204. Based on detected data {circumflex over (x)} at the output of the Viterbi detector 205, the reconstruction filter 206 may reconstruct the equalized samples. If there is no noise and mis-equalization, reconstructed samples at the output of the reconstruction filter 206 would be the same as the equalized samples r at the output of the FIR equalizer 204.
The subtractor 207 may receive the reconstructed samples from the reconstruction filter 206 and the equalized samples r from the FIR equalizer 204 and obtain their differences, which are error samples.
The adaptive logic 208 may receive the error signals from the subtractor 207 and the ADC samples from the ADC 203, and adapt the response of the FIR equalizer 204 to minimize the error samples.
One problem with the read channel 200 in
Conventional approaches typically use a constraint at the adaptive logic 208 to restrict the interactions during the adaptation process. One prior art approach is called Fix2 constraint, which fixes values of certain taps of the FIR response c, e.g., c3 and c4. Another prior art approach is called Rank2 constraint, which fixes two quantities of the FIR response, e.g., c0-c2+c4-c6+c8, and c1-c3+c5-c7+c9, while allowing individual tap values to change.
One problem with these just-mentioned constraints is that the constraints are not based on tap values of the FIR response c. In addition, when the constraints are 0 or very small, the constraints may become useless.
Embodiments of the present invention are described herein with reference to the accompanying drawings, similar reference numbers being used to indicate functionally similar elements.
Each of
Each of
An apparatus for (i) adapting a finite impulse response (FIR) equalizer in a hard disk drive read channel and (ii) compensating for an error signal, the apparatus comprises: a first memory to store a first direction of an FIR response of the FIR equalizer; and adaptive logic to (i) receive the first direction from the first memory and (ii) calculate a constraint to adapt the FIR response to reduce the error signal, wherein the constraint prevents adaptation of the FIR response going into the first direction.
According to one aspect of the present invention, the first memory is a gain direction memory; and the first direction is a gain direction of the FIR response, wherein the gain direction of the FIR response is a change of direction of the FIR response responding to a change in a gain of the FIR response.
According to one aspect of the present invention, the first memory is a phase direction memory; and the first direction is a phase direction of the FIR response, wherein the phase direction of the FIR response is a change of a direction of the FIR response responding to a change in a phase of the FIR response.
According to one aspect of the present invention, the apparatus further comprises: a second memory to store a second direction of the FIR response of the FIR equalizer, wherein the second direction is a phase direction corresponding to a change of a direction of the FIR response responding to a change in a phase of the FIR response.
According to one aspect of the present invention, the adaptive logic determines the constraint as follows:
P=I−V(VTV)−IVT, and
V=[cw]
wherein P is the constraint, I is the identity matrix, c is the gain direction and w is the phase direction.
According to one aspect of the present invention, the apparatus further comprises: a first determining unit to (i) determine whether a first predetermined condition is met and (ii) enable the adaptive logic to calculate the constraint only in response to the first predetermined condition having been met.
According to one aspect of the present invention, the first predetermined condition is:
cTcprev/(∥c∥•∥cprev∥)<T1
wherein c and cprev are current and previous values of the gain direction respectively, T1 is a threshold, and 0<T1<1.
According to one aspect of the present invention, the apparatus further comprises: a second determining unit to (i) determine whether a second predetermined condition is met and (ii) enable the adaptive logic to calculate the constraint only in response to the second predetermined condition having been met, wherein the second predetermined condition is:
wTwprev/(∥w∥•∥wprev∥)<T2
wherein w and wprev are current and previous values of the phase direction, T2 is a threshold, and 0<T2<1.
According to one aspect of the present invention, the apparatus further comprises: a third determining unit to (i) determine whether a third predetermined condition is met and (ii) enable the adaptive logic to calculate the constraint only in response to the third predetermined condition having been met, wherein the third predetermined condition is:
wTwprev/(∥w∥•∥wprev∥)<T2
wherein w and wprev are current and previous values of the phase direction, T2 is a threshold, and 0<T2<1.
According to one aspect of the present invention, the constraint limits adaptation of the FIR response to stay generally in a direction perpendicular to the first direction.
A method and apparatus for adapting an FIR equalizer in a hard disk drive read channel which determines a gain direction and/or a phase direction of an FIR response c of the FIR equalizer, and calculates a constraint so that the update does not change the FIR response in its gain and/or phase direction.
At 301, the gain direction of an FIR response is determined. The gain direction refers to the change of the direction of the FIR response c responding to a change in the gain. Assuming that a base FIR response is c=[c0 c1 c2 . . . c9]T, and its gain and phase are 1 and 0 respectively, then, as shown in
(1+α)c=c+cα (3)
Thus, the direction of change is cα. Since α is a constant, the gain direction is the FIR response c itself.
At 302, the phase direction of the FIR response is determined. The phase direction refers to the change of the direction of the FIR response c responding to a change in the phase.
c(τ)≈c+τ·w (4)
wherein w is the phase direction.
The continuous function c(t) of the curve 500 is:
c(t)=c0 sin c(t)+c1 sin c(t−1)+ . . . +c9 sin c(t−9) (5)
The base FIR response may also be written as:
c=[c(0),c(1),c(2), . . . c(9)]T (6)
wherein c(0) represents c(t) at time 0, c(1) represents c(t) at time 1, . . . and c(9) represents c(t) at time 9.
The value of c(0), c(1), . . . and c(9) may be obtained using function (5). When t=0, c(0)=c0; when t=1, c(1)=c1, . . . and when t=9, c(9)=c9. In other words:
c=[c(0),c(1),c(2), . . . c(9)]T=[c0,c1, . . . , c9]T (7)
When c is perturbed by τ, the function (6) may change to:
c perturbed by τ=[c(τ)c(τ+1) . . . c(τ+9)]T (8)
When τ is small,
c(τ)≈c(0)+τc′(0)
c(τ+1)≈c(1)+τc′(1)
. . .
c(τ+9)≈c(9)+τc′(9) (9)
wherein c′(0) is the derivative of c at time 0, c′(1) is the derivative of c at time 1, . . . and c′(9) is the derivate of c at time (9).
From (8) and (9), it can be seen:
From (4) and (10), it can be seen that the phase direction w is:
Following the properties of the sinc function,
Thus, the phase direction w is a matrix A times the FIR response c.
At 303, a constrained FIR update d is calculated such that the adaptation will not go into the gain direction and the phase direction of the FIR response. To decouple the adaptation of the FIR response c and the gain and phase direction of the read channel, (12) and (13) need to be satisfied:
cTd=0 (12)
wTd=0 (13)
In one embodiment, the following function can be used to calculate the constrained FIR update d:
d=Pu (14)
P=I−V(VTV)−IVT, and (15)
V=[cw] (16)
wherein u is an unconstrained update to the FIR response, P is a constraint, and I is the identity matrix.
Functions (14), (15) and (16) indicated above are only used as examples, and other functions may be used for the constrained adaptation, as long as the adaptation does not change the tap values in the gain and phase direction.
In the embodiment shown in
In addition, gain direction and phase direction may be used together with other constraints. In one embodiment, three constraints may be imposed: the gain direction, the phase direction and a fixed tap constraint, e.g., fixed tap 3. The equation (16) may change to:
V=[cws] (17)
wherein s=[0 0 0 1 0 0 0 0 0 0]T
In other words, each column of matrix V represents a constraint on the update direction. The first and second constraints may be equations (12) and (13), and the last constraint requires that tap d3=0.
Assuming that an FIR response c has two taps, CO and c1, and assuming that only the gain loop is on, and ignoring the phase loop,
cTd=[10]=0 (18)
The constraint may be updated, e.g., at B, before the tap value reaches the global solution. Since the gain direction is along the OB line, the same as the tap, the update direction should be along the MN line, which is generally perpendicular to the gain direction line OB. The tap value may reach the global solution at D, with the converged value (0.8, 0.8). When the tap value is perturbed by noise and fluctuates from D to E, the gain direction may change to OE, and the update direction may change to JK. Consequently, the tap value may move to F, instead of returning to a Embodiments of the present invention solve this problem by keeping previous values of c and w which specify the directions of constraints, defining a predetermined condition, and updating the constraint P only when the predetermined condition is satisfied, so that the tap may return to D, instead of diverging.
At 601, the gain direction of the FIR response c may be determined.
At 602, the phase direction of the FIR response c may be determined.
At 603, it is determined whether a predetermined condition is met. In one embodiment, the predetermined condition is:
cTcprev/(∥c∥•∥cprev∥)<T1 (19)
wherein c and cprev are the current and previous value of the gain direction respectively, T1 is a threshold, and 0<T1<1.
The predetermined condition (19) is used to determine whether the current c is far enough from its previous value cprev. In one embodiment, T1=0.98. The threshold T1 may be other values, as long as it can be used to check whether cprev and c are far apart. Another example of the predetermined condition may be:
∥c−cprev∥>ε, (20)
wherein ε can be any positive number.
In one embodiment, the predetermined condition is:
wTwprev/(∥w∥•∥wprev∥)<T2 (21)
wherein w and wprev are the current and previous values of the phase direction, T2 is a threshold, and 0<T2<1.
In one embodiment, T2=0.98. The threshold T2 may be other values, as long as it can be used to check whether wprev is far enough from w. Another example of the condition may be:
∥w−wprev∥>θ (22)
wherein θ can be any positive number.
If the predetermined condition is met, the procedure may proceed to 605 to calculate a new constraint P, e.g., with functions (15) and (16). At 606, the constraint P may be updated, cprev may be updated to c, and wprev may be updated to w. The procedure may then return to 601.
If the condition is not met, the procedure may proceed to 604, the constraint P is not updated, the previous value may continue to be used and the procedure may return to 601.
Although in
At the beginning of the operation, the FIFO 901 receives ADC samples, e.g., ten consecutive ADC samples from the ADC 203 in
At the same time, the logic 914 checks whether the current value of the FIR taps c from the memory 906 and an initial value from the memory 907 are far apart enough from each other, e.g., using (19) or (20), and issue a trigger signal to the OR gate 915 if they are far apart enough. The logic 913 may check whether the current value of the phase direction w from the memory 908 is sufficiently far apart from an initial value from the memory 909, e.g., using (21) or (22), and issue a trigger signal to the OR gate 915 if they are far enough apart. If the OR gate 915 receives a trigger signal, either from the logic 913 or from the logic 914, the OR gate 915 may issue a trigger signal to activate the connections in dashed lines to update the current value of FIR tap c in the memory 906, and move its previous value to the memory 907; to update the current value of the phase direction w in the memory 908, and move its previous value to the memory 909; to calculate an updated constraint P at the logic 910 with the updated c and w, and update the constraint P in the memory 911. The updated P may then be used to calculate constrained update of the FIR taps c, and the adaptation may continue.
Several features and aspects of the present invention have been illustrated and described in detail with reference to particular embodiments by way of example only, and not by way of limitation. Alternative implementations and various modifications to the disclosed embodiments are within the scope and contemplation of the present disclosure. Therefore, it is intended that the invention be considered as limited only by the scope of the appended claims.
This disclosure claims the benefit of priority to previously filed U.S. provisional patent application Ser. No. 61/170,190, filed Apr. 17, 2009, entitled A FIR WITH CONSTRAINT DEPENDENT ON TAPS. That provisional application is hereby incorporated by reference in its entirety.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5623402 | Johnson | Apr 1997 | A |
| 5689532 | Fitzpatrick | Nov 1997 | A |
| 5892632 | Behrens et al. | Apr 1999 | A |
| 6122120 | Shimoda | Sep 2000 | A |
| 6185716 | Riggle | Feb 2001 | B1 |
| 6519108 | Au et al. | Feb 2003 | B2 |
| 6778345 | Ozdemir | Aug 2004 | B1 |
| 6778601 | Ise et al. | Aug 2004 | B2 |
| 6804695 | Hsu | Oct 2004 | B1 |
| 6816328 | Rae | Nov 2004 | B2 |
| 6867941 | Ozdemir | Mar 2005 | B1 |
| 7237173 | Morita et al. | Jun 2007 | B2 |
| 7454084 | Faber et al. | Nov 2008 | B2 |
| 7460613 | Sahlman | Dec 2008 | B2 |
| 20010016926 | Riggle | Aug 2001 | A1 |
| 20030043487 | Morita et al. | Mar 2003 | A1 |
| 20070279780 | Tomita | Dec 2007 | A1 |
| 20070279781 | Tomita | Dec 2007 | A1 |
| 20070279788 | Andersen et al. | Dec 2007 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 61170190 | Apr 2009 | US |
