Multi-dimensional optimization of read channel
Variations of the Nelder-Mead direct search method are employed to find read channel parameter settings in a discrete field having three or more dimensions. The three or more dimensions correspond to read channel parameters, at least some of which are highly correlated. The steps of the Nelder-Mead method are executed according to a methodology to arrive at substantially optimal parameter settings for a read channel, even where a discrete function defining parameter outcomes is noisy. In some embodiments, dimensional collapse, considered inefficient in a two-dimensional field, is allowed in order to reach an optimal solution in a greater-than-two-dimensional field.
1. A method for determining read channel parameters in a data storage device, comprising:
receiving a read channel performance function corresponding to at least three read channel performance parameters;
selecting a plurality of vertexes in a discrete, continuous field defined by the read channel performance function;
determining a worst vertex in the plurality of vertexes having a worst performance;
determining if replacing the worst vertex will cause dimensional collapse; and
determining if dimensional collapse is desirable.
2. The method of claim 1 , wherein determining if dimensional collapse is desirable comprises determining if the vertexes satisfy strict dimensional constraints.
3. The method of claim 1 , further comprising:
determining a best vertex in the plurality of vertexes; and
expanding the best vertex according to a central weight of all vertexes in the plurality of vertexes.
4. The method of claim 1 , further comprising:
determining a best vertex in the plurality of vertexes; and
iteratively contracting each vertex in the plurality of vertexes with respect to the best vertex.
5. The method of claim 4 , further comprising:
determining that a pre-determined number of iterations have been executed; and
halting further iterations.
6. The method of claim 4 , further comprising:
logging each vertex used during each iterations; and
rejecting any new vertex previously logged.
7. The method of claim 6 , further comprising selecting a neighboring vertex to the rejected vertex.
8. A data storage device comprising:
a processor;
memory connected to the processor;
a data storage element connected to the processor; and
computer executable program code to configured the processor to:
receive a read channel performance function corresponding to at least three read channel performance parameters;
select a plurality of vertexes in a discrete, continuous field defined by the read channel performance function;
determine a worst vertex in the plurality of vertexes having a worst performance;
determine if replacing the worst vertex will cause dimensional collapse; and
determine if dimensional collapse is desirable.
9. The data storage device of claim 8 , wherein determining if dimensional collapse is desirable comprises determining if the vertexes satisfy strict dimensional constraints.
10. The data storage device of claim 8 , wherein the computer executable program code further configures the processor to:
determine a best vertex in the plurality of vertexes; and
expand the best vertex according to a central weight of all vertexes in the plurality of vertexes.
11. The data storage device of claim 8 , wherein the computer executable program code further configures the processor to:
determine a best vertex in the plurality of vertexes; and
iteratively contract each vertex in the plurality of vertexes with respect to the best vertex.
12. The data storage device of claim 11 , wherein the computer executable program code further configures the processor to:
determine that a pre-determined number of iterations have been executed; and
halt further iterations.
13. The data storage device of claim 11 , wherein the computer executable program code further configures the processor to:
log each vertex used during each iterations; and
reject any new vertex previously logged.
14. The data storage device of claim 13 , wherein the computer executable program code further configures the processor to select a neighboring vertex to the rejected vertex.
15. A computer apparatus comprising:
a processor;
memory connected to the processor;
a data storage element connected to the processor; and
computer executable program code to configured the processor to:
receive a read channel performance function corresponding to at least three read channel performance parameters;
select a plurality of vertexes in a discrete, continuous field defined by the read channel performance function;
determine a worst vertex in the plurality of vertexes having a worst performance;
determine if replacing the worst vertex will cause dimensional collapse; and
determine if dimensional collapse is desirable.
16. The apparatus of claim 15 , wherein determining if dimensional collapse is desirable comprises determining if the vertexes satisfy strict dimensional constraints.
17. The apparatus of claim 15 , wherein the computer executable program code further configures the processor to:
determine a best vertex in the plurality of vertexes; and
expand the best vertex according to a central weight of all vertexes in the plurality of vertexes.
18. The apparatus of claim 15 , wherein the computer executable program code further configures the processor to:
determine a best vertex in the plurality of vertexes; and
iteratively contract each vertex in the plurality of vertexes with respect to the best vertex.
19. The apparatus of claim 18 , wherein the computer executable program code further configures the processor to:
determine that a pre-determined number of iterations have been executed; and
halt further iterations.
20. The apparatus of claim 18 , wherein the computer executable program code further configures the processor to:
log each vertex used during each iterations; and
reject any new vertex previously logged.