Segmented LLR Computation for Distorted Flash Memory Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage and communication systems face challenges in computing probability values for received values with arbitrary probability density functions, which are more complex than Gaussian distributions, due to distortions like back pattern dependency and intercell interference in flash memory devices.
Innovation Solution
The method involves computing soft data or log likelihood ratios by identifying segments with associated parameters in a piecewise linear function, allowing for the use of Gaussian approximations within each segment to calculate probabilities for arbitrary distributions, thereby simplifying the computation process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If probability values are computed for arbitrary probability density functions, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent divides the probability density function into multiple segments, where each segment is approximated by a Gaussian distribution. This segmentation allows the system to handle arbitrary PDFs by breaking them down into simpler, computationally manageable pieces, each with its own mean and variance parameters that can be efficiently processed.
Solution Approach 2:
The patent transforms the arbitrary probability density function into a set of Gaussian distributions by changing the parameters (mean and variance) for different segments. This parameter transformation enables the use of well-established Gaussian computation methods while accurately representing the original arbitrary distribution, thus improving precision without requiring entirely new computational approaches.
2Device complexity
If Gaussian approximations are used within segments, then computation complexity is reduced, but measurement precision may be compromised
Solution Approach 1:
By dividing the arbitrary PDF into multiple segments and applying Gaussian approximation to each segment individually, the patent maintains higher overall precision compared to using a single Gaussian approximation for the entire distribution. Each segment's Gaussian parameters are optimized to match the local characteristics of the arbitrary PDF, reducing approximation errors.
Solution Approach 2:
The patent applies different Gaussian parameters (mean and variance) to different segments of the probability density function, allowing each segment to be locally optimized for accuracy. This local quality approach ensures that the Gaussian approximation closely matches the arbitrary PDF in each specific region, thereby maintaining measurement precision while benefiting from the computational simplicity of Gaussian distributions.
Data Source
AI summary
Methods and apparatus are provided for computing soft data or log likelihood ratios for received values in communication or storage systems. Soft data values or log likelihood ratios are computed for received values in a communication system or a memory device by obtaining at least one received value; identifying a segment of a function corresponding to the received value, wherein the function is defined over a plurality of segments, wherein each of the segments has an associated set of parameters; and calculating the soft data value or log likelihood ratio using the set of parameters associated with the identified segment. The computed soft data values or log likelihood ratios are optionally provided to a decoder.