Memory system quality margin analysis and configuration
A memory quality engine can improve the operation of a memory system by setting more effective operating parameters, disabling or removing memory devices unable to meet performance requirements, and providing evaluations between memory populations. These improvements can be accomplished by converting quality measurements of a memory population into CDF-based data, formulating comparisons of the CDF-based data to metrics for quality analysis, and applying the quality analysis. In some implementations, the metrics for quality analysis can use one or more thresholds, such as a system trigger threshold or an uncorrectable error correction condition threshold, which are set based on the error correction capabilities of a memory system. Formulating the comparison to these metrics can include determining a margin between the CDF-based data at a particular codeword frequency and one of the thresholds.
1. A method comprising:
obtaining quality measures for a memory population of a memory system, wherein the quality measures comprise a histogram indicating counts of errors per memory unit in the memory population;
converting the quality measures into cumulative distribution function (CDF)-based data by computing a cumulative distribution function of the histogram;
formulating one or more comparisons of the CDF-based data to metrics for a quality analysis by:
obtaining an error amount threshold condition;
obtaining an acceptable memory unit failure rate;
computing a specific quality measure by determining an error amount indicated by the CDF-based data at the acceptable memory unit failure rate; and
computing a margin based on a comparison of the specific quality measure and the error amount threshold condition; and
applying results of the quality analysis to make improvements in the memory system by causing, based on the computed margin, one or more actions in relation to the memory population, the one or more actions comprising a modification to operating parameters for the memory population.
2. The method of claim 1 , wherein the memory population comprises: a plurality of memory cells, a plurality of memory pages, a plurality of memory word lines, a plurality of memory blocks, a plurality of memory planes, a plurality of memory dies, or a plurality of memory devices, or a combination thereof.
3. The method of claim 1 , wherein the errors per memory unit in the quality measures for the memory population comprise bit errors per codeword.
4. The method of claim 1 , wherein converting the quality measures into CDF-based data comprises computing the CDF-based data as one minus the cumulative distribution function of the histogram.
5. The method of claim 1 , wherein the error amount threshold condition either correlates to data recovery speed or specifies an error amount above which errors are not recoverable by a controller of the memory population, and wherein the acceptable memory unit failure rate specifies a rate at which memory units meeting or exceeding the error amount threshold condition cause system performance to fall below a predetermined performance threshold.
6. The method of claim 5 , wherein the error amount threshold condition is a system trigger threshold correlated to data recovery speed that specifies an error rate above which the memory system has to invoke secondary methods to perform data recovery by a controller of the memory system.
7. The method of claim 5 , wherein the error amount threshold condition is an uncorrectable error correction condition (UECC) threshold specifying an error rate above which errors are not recoverable by the controller.
8. The method of claim 1 , wherein the margin is computed based on producing a dividend from the error amount threshold condition divided by the specific quality measure and taking a logarithm of the dividend.
9. The method of claim 1 , wherein the acceptable failure rate is based on measures of error recovery functions performed by one or more devices containing the memory population.
10. The method of claim 1 , wherein the modification to the operating parameters comprises adjusting a voltage, a clock rate, or a latency, or a combination thereof for the memory population based on the computed margin.
11. A non-transitory computer-readable storage medium storing instructions that, when executed by a computing system, cause the computing system to perform operations comprising:
obtaining quality measures for a memory population of a memory system;
converting the quality measures into cumulative distribution function (CDF)-based data;
formulating one or more comparisons of the CDF-based data to metrics for a quality analysis, by:
obtaining an error amount threshold condition;
obtaining an acceptable memory unit failure rate;
computing a specific quality measure by determining an error amount indicated by the CDF-based data at the acceptable memory unit failure rate; and
computing a margin based on a comparison of the specific quality measure and the error amount threshold condition; and
applying results of the quality analysis to make improvements in the memory system by causing, based on the computed margin, one or more actions in relation to the memory population.
12. The non-transitory computer-readable storage medium of claim 11 ,
wherein the quality measures comprise a histogram indicating counts of errors per memory unit in the memory population; and
wherein converting the quality measures into CDF-based data is performed by computing a cumulative distribution function of the histogram.
13. The non-transitory computer-readable storage medium of claim 11 , wherein the error amount threshold condition either specifies an error amount correlated to data recovery speed or specifies an error amount above which errors are not recoverable by a controller of the memory population.
14. The non-transitory computer-readable storage medium of claim 13 , wherein the acceptable memory unit failure rate specifies a rate at which memory units having at least the error amount threshold condition error count will cause system performance to fall below a predetermined performance threshold.
15. The non-transitory computer-readable storage medium of claim 11 , wherein causing the one or more actions comprises causing a modification to operating parameters for the memory population.
16. The non-transitory computer-readable storage medium of claim 11 , wherein causing the one or more actions comprises generating a report indicating the computed margin.
17. The non-transitory computer-readable storage medium of claim 11 , wherein the margin is computed on a logarithmic scale and is based on a ratio of the threshold condition to the specific quality measure.
18. The non-transitory computer-readable storage medium of claim 11 , wherein the acceptable memory unit failure rate is based on measures of error recovery functions performed by devices containing the memory population.
19. A system comprising:
a memory;
one or more processors;
an interface configured to, using the one or more processors:
obtain quality measures for a memory population of a memory system;
obtain an error amount threshold condition; and
obtain an acceptable memory unit failure rate;
a cumulative distribution function (CDF) converter configured to, using the one or more processors, convert the quality measures into CDF-based data;
a margin analysis engine configured to, using the one or more processors, formulate one or more comparisons of the CDF-based data to metrics for a quality analysis, by:
computing a specific quality measure by determining an error amount indicated by the CDF-based data at the acceptable memory unit failure rate; and
computing a margin based on a comparison of the specific quality measure and the error amount threshold condition; and
an analysis application engine configured to, using the one or more processors, apply results of the quality analysis to make improvements in the memory system by causing, based on the computed margin, one or more actions in relation to the memory population.
20. The system of claim 19 ,
wherein the quality measures comprise a histogram indicating counts of errors per memory unit in the memory population; and
wherein converting the quality measures into CDF-based data is performed by computing a cumulative distribution function of the histogram.