Abstract: The present disclosure provides a method of training artificial intelligence to estimate a lifetime of a storage device, which includes steps of: determining whether an operating parameter of the storage device executing a processing program on bit value values is smaller than an operational threshold parameter or not, if yes, decoding the bit value values stored in the storage device by a decoder; determining whether the bit value values stored in the storage device are successfully decoded by the decoder or not, if yes, classifying the memory unit of the storage device into a strong correct region, a weak correct region, a strong error region or a weak error region; determining whether the number of the memory units falls within an allowable number range or not, if not, initiating an artificial intelligence neural network system to use machine learning to estimate the lifetime of the storage device.

Abstract: A method for garbage collecting for non-volatile memories are disclosed. The method includes steps: a) providing a SSD, connected to a host, containing a plurality of TLC blocks and SLC blocks; b) reading 3M TLC pages in a TLC block having data; c) moving valid data in the TLC blocks to at least one clean TLC block; d) sending a host program command of 1 page to the host; e) repeating step b) to step d) until valid data in 8 TLC blocks are moved; f) reading 1 SLC page in a SLC block having data; g) moving valid data in the SLC block to the at least one clean TLC block; h) sending a host program command of ? page to the host; and i) repeating step f) to step h) until valid data in the SLC block having data are moved.

Abstract: A method and a system for determining bit values in a non-volatile memory having a number of cells each for storing a bit value are disclosed. The method includes the steps of: a) providing a first test sensing voltage to the cells and calculating a cell count; b) providing another test sensing voltage to the cells and calculating a difference of the cell counts between this step and previous step; c) providing still another test sensing voltage and calculating another difference of the cell counts between this step and previous step; d) processing step c) for N times; e) calculating differential amounts of cell counts and assigning an index number to each group of cells; f) choosing a voltage as an updated sensing voltage.

Abstract: A method and a controller for recovering data in event of a program failure and a storage system using the method and the controller are disclosed. The controller includes main units of a parity generator, a volatile memory module and a processor. With a parity in the volatile memory module and successfully programmed sub-data, a program failed write data can be recovered and correctly programmed. The method of the present invention has advantages of saving use of storage resources and extending lifetime of the storage system than other methods for recovering data in event of a program failure.

Abstract: A SSD and a method for reducing use of DRAM in the SSD are disclosed. The method includes the steps of: A. providing a referring table in a DRAM module of a SSD; B. providing a logical-to-physical address table in the DRAM module; C. receiving a command for accessing a target data in a target logical address of the SSD; D. checking if one physical address is stored in the logical-to-physical address table; E. executing the command by using the mapping data in the subgroup or copying a corresponding subgroup including one mapping data for the target logical address from the mapping table to the DRAM module via the referring table; and; and F. adding a target physical address of the DRAM module where the mapping data for the target logical address is stored to the logical-to-physical address table so that the target logical address is able to correspond thereto.

Abstract: A decoding algorithm with an enhanced parity check matrix and a re-encoding scheme for LDPC codes is disclosed. The decoding algorithm includes the steps of: providing the enhanced parity check matrix; receiving a message part of an original codeword encoded by a generator matrix from the enhanced parity check matrix; setting a LLR for each bit node of the enhanced parity check matrix; processing hard decision on the message part of the original codeword; encoding the message part of the original codeword by the generator matrix to generate a new codeword having a generated parity part; comparing the original parity part with the generated parity part to find out bits of difference; voting candidate error bits to choose the most probably erratic bits; modifying LLR of the chosen bits to have a modified codeword; and processing a conventional iterative decoding procedure on the modified codeword to have a processed codeword.

Abstract: A field firmware upgrading method is adapted in a field firmware upgrading system having a computing device and a storage device. First, the computing device is boot up to a real-time operating system (RTOS), wherein the RTOS and a firmware of the storage device are communicated with each other via a driver. The RTOS sends reading commands to the firmware, such that the firmware performs a reading operation meeting a prefix proceeding, wherein the prefix proceeding defines an order which specific logic block addresses (LBAs) are read. After the firmware performs the reading operation meeting the prefix proceeding, the RTOS sends writing commands to the firmware, such that the firmware performs a writing operation meeting a specific writing proceeding to write an firmware image file into the storage device beginning at a target LBA defined by the specific writing proceeding.

Abstract: A method for transferring a command from a host to a device controller and a system using the method are disclosed. The method includes the steps of: A. determining a segment size; B. dividing a command into a number of sections each having a size the same as the segment size; C. sequentially distributing the sections to n groups; D. changing distributing order to a reverse order or keeping the same distributing order in step C if a cycle of distribution is finished while there are sections left for distributing; E. restructuring the section(s) in each group as a sub-command after all sections are distributed; and F. providing the sub-commands to a device controller synchronously.

Abstract: A SSD and a method for reducing use of DRAM in the SSD are disclosed. The method includes the steps of: A. providing a referring table in a DRAM module of a SSD; B. providing a logical-to-physical address table in the DRAM module; C. receiving a command for accessing a target data in a target logical address of the SSD; D. checking if one physical address is stored in the logical-to-physical address table; E. executing the command by using the mapping data in the subgroup or copying a corresponding subgroup including one mapping data for the target logical address from the mapping table to the DRAM module via the referring table; and; and F. adding a target physical address of the DRAM module where the mapping data for the target logical address is stored to the logical-to-physical address table so that the target logical address is able to correspond thereto.

Abstract: A method and a controller for recovering data in event of a program failure and a storage system using the method and the controller are disclosed. The controller includes main units of a parity generator, a volatile memory module and a processor. With a parity in the volatile memory module and successfully programmed sub-data, a program failed write data can be recovered and correctly programmed. The method of the present invention has advantages of saving use of storage resources and extending lifetime of the storage system than other methods for recovering data in event of a program failure.

Abstract: An adaptive compression data storing method for non-volatile memories and a system using the method are disclosed. The system includes a host interface unit, a data compressor, a padding unit, a buffer, a combining unit, and a mapping table unit. By combining some compressed data in one page, the present invention can settle the problem that space for storing a compressed data that can not be utilized. Further, lifetime of non-volatile memories can be extended.

Abstract: A method for improving access performance of a non-volatile storage device when programming data of a size smaller than a fixed minimum program number (FMPN) is disclosed. The method includes the steps of: predetermining a size of a blank data section for combining with a first data section and a second data section, the total size of the first data section, the second data section and the blank data section equals the FMPN; reading out data located at the second data section; updating a new data to the first data section; combining the new data with the data at the second data section; and incorporating the combined data with the blank data of the blank data section to become a final data, and programming the final data.

Abstract: A method for encoding multi-modes of BCH codes and an associated encoder is disclosed. The method has the steps of: building a number of encoding matrices; combining the encoding matrices with one side aligned to form a combined matrix; seeking common sub-expressions (CSEs) in the combined matrix, and encoding a message using the combined matrix.

Abstract: An early termination method with a re-encoding scheme for decoding of error correction codes is disclosed. The method includes the steps of: A. receiving soft values; B. processing hard decision on the soft value to determine a codeword; C. separating the codeword into a data part and a first parity part; D. re-encoding the data part to get a second parity part; E. checking if the first parity part and the second parity part are equivalent; and F. if a result of step E is yes, stopping decoding the codeword; if the result of step E is no, processing a decoding algorithm on the codeword. By this method, the received codeword still can be correctly decoded if there are many errors in the parity region and its decoding performance can be improved.

Abstract: A decoding algorithm with an enhanced parity check matrix and a re-encoding scheme for LDPC codes is disclosed. The decoding algorithm includes the steps of: providing the enhanced parity check matrix; receiving a message part of an original codeword encoded by a generator matrix from the enhanced parity check matrix; setting a LLR for each bit node of the enhanced parity check matrix; processing hard decision on the message part of the original codeword; encoding the message part of the original codeword by the generator matrix to generate a new codeword having a generated parity part; comparing the original parity part with the generated parity part to find out bits of difference; voting candidate error bits to choose the most probably erratic bits; modifying LLR of the chosen bits to have a modified codeword; and processing a conventional iterative decoding procedure on the modified codeword to have a processed codeword.

Abstract: A method for detecting problem cells of a SATA SSD and a SATA SSD having self-detecting function looking for problem cells are disclosed. The method includes the steps of: providing a detecting program used to detect aged and died cells in a SATA SSD; writing the detecting program to a MCU (Micro Control Unit) in the SATA SSD; pulling high electric potential of a communicating pin of a SATA connector of the SATA SSD to initiate the detecting program; collecting location data of aged and died cells in the SATA SSD by the detecting program; and storing the location data in a storage area in the SSD. The present invention utilizes the DAS/DSS pin as a channel to initiate detecting program. It has advantages of using current interface of SSD, no effort on taking apart hardware and automatically running the detecting program without human control.

Abstract: The present disclosure illustrates a low density parity-check code decoder adapted for decoding coding data having bit nodes and check nodes. The decoder includes a calculation module and a memory. The calculation module includes k calculation units and n shift units, and the memory includes n memory units. The memory is coupled to the calculation module. Each shift unit is one-to-many coupled to the k calculation units. The n memory units are coupled to the n shift units. The calculation module operatively divides the coding data into n first-bit-strings. The ith calculation unit operatively generates a second-bit-string by calculating ith bits of the n first-bit-strings. The jth shift unit operatively generates a third-bit-string upon receiving jth bits of the k second-bit-strings, and shifts the third-bit-string. The memory units are configured for storing the n shifted third-bit-strings respectively.

Abstract: A method and decoder for early terminating decoding processes of serial concatenated coding are disclosed. The method includes the steps of providing a codeword, encoded by a first coding and a second coding sequentially; setting a maximum syndrome weight; decoding the second coding for the codeword by iterative calculations for syndromes; terminating decoding of the second coding if a number of the iterative calculations reaches a preset number or a syndrome weight of one iterative calculation is equal to or smaller than the maximum syndrome weight, otherwise repeating the decoding step and the terminating step; and decoding the first coding for the codeword.

Abstract: The present invention discloses a multi-code Chien's search circuit for BCH codes with various values of m in GF(2m). The circuit includes: a combined matrix unit, a number of first multiplexers, a number of registers and a number of second multiplexers. By designing the Chien's search circuit having several Chien's search matrices, with peripheral components, it is able to achieve applications for different code rates, different code lengths and even different m in GF(2m).

Abstract: A simplified inversionless Berlekamp-Massey algorithm for binary BCH codes and circuit implementing the method are disclosed. The circuit includes a first register group, a second register group, a control element, an input element and a processing element. By breaking the completeness of math structure of the existing simplified inversionless Berlekamp-Massey algorithm, the amount of registers used can be reduced by two compared with conventional algorithm. Hardware complexity and operation time can be reduced.