Hard errors are determined for an unsuccessful decoding of codeword bits read from NAND memory cells via a read channel and input to a low-density parity check (LDPC) decoder. A bit error rate (BER) for the hard errors is estimated and BER for the read channel is estimated. Hard error regions are found using a single level cell (SLC) reading of the NAND memory cells. A log likelihood ratio (LLR) mapping of the codeword bits input to the LDPC decoder is changed based on the hard error regions, the hard error BER, and/or the read channel BER.

A method, of decoding error correction code of a memory device with dynamic bit error estimation, can include generating at least one metric corresponding to one or more syndromes associated with a code word, the code word comprising an error correction code of a memory device, decoding the code word by a first decoder integrated with the memory device, in response to a determination that the metric satisfies a threshold associated with the syndromes, the first decoder having a first execution property, and decoding the code word by a second decoder integrated with the memory device, in response to a determination that the metric does not satisfy the threshold associated with the syndromes, the second decoder having a second execution property distinct from the first execution property, or in response to a determination that the metric satisfies the threshold associated with the syndromes, and in response to a determination to perform further decoding.

A system for a fiber-optic network includes a transceiver. The transceiver includes a fiber-optic interface unit and a host unit. The host unit includes a low-complexity error correction decoder and a high-complexity error correction decoder. One or both from the low-complexity error correction decoder and the high-complexity error correction decoder are selected to decode input data from the fiber-optic interface unit, the input data including codewords.

A low-density parity-check (LDPC) decoder performs check node computations as N different segments of the check nodes which have connections only to a codeword segment of length C/N bits as well as check nodes that have connections across the entire codeword of length C. The decoder can include a controller or other compute hardware to decode the codeword, including to perform computations for separate segments of C/N bits of the codeword. The system can perform computations including adjustment of the decode computations based on an expected error rate for selected segments of the codeword.

Methods and apparatus are disclosed for managing the storage of static and dynamic neural network data within a non-volatile memory (NVM) die for use with deep neural networks (DNN). Some aspects relate to separate trim sets for separately configuring a static data NVM array for static input data and a dynamic data NVM array for dynamic synaptic weight data. For example, the static data NVM array may be configured via one trim set for data retention, whereas the dynamic data NVM array may be configured via another trim set for write performance. The trim sets may specify different configurations for error correction coding, write verification, and read threshold calibration, as well as different read/write voltage thresholds. In some examples, neural network regularization is provided within a DNN by setting trim parameters to encourage bit flips to avoid overfitting. Some examples relate to managing non-DNN data, such as stochastic gradient data.

A memory controller includes, in one embodiment, a memory interface, a plurality of decoders, and a controller circuit. The memory interface is configured to interface with a memory having a plurality of wordlines. Each decoder of the plurality of decoders is configured to determine a bit error rate (BER). The controller circuit configured to generate a plurality of bit-error-rate estimation scan (BES) hypotheses for one wordline of the plurality of wordlines, divide the plurality of BES hypotheses among the plurality of decoders, receive BER results from the plurality of decoders based on the plurality of BES hypotheses, and adjust one or more read locations of the one wordline based on the BER results from the plurality of decoders.

A low-density parity-check (LDPC) decoder performs check node computations as N different segments of the check nodes which have connections only to a codeword segment of length C/N bits as well as check nodes that have connections across the entire codeword of length C. The decoder can include a controller or other compute hardware to decode the codeword, including to perform computations for separate segments of C/N bits of the codeword. The system can perform computations including adjustment of the decode computations based on an expected error rate for selected segments of the codeword.

An error rate measuring apparatus includes: an operation unit that sets a codeword length, an FEC symbol length, and an FEC symbol error threshold; error counting means for counting FEC symbol error and an uncorrectable codeword; a display unit that performs display by setting one zone of a display area as one FEC symbol length, matching a zone length of a horizontal axis of the display area with one codeword length, and performing line feed in codeword length units according to presence or absence of the FEC symbol error in FEC symbol length units based on a counting result; search means for searching for the uncorrectable codeword starting from the cursor on the identification display; and display control means for performing display control of the cursor at a position of a head error of the searched uncorrectable codeword.

An error rate measuring apparatus includes: an operation unit that sets a codeword length, an FEC symbol length, and an FEC symbol error threshold in accordance with a communication standard of a device under test W; error counting means for counting FEC symbol error detected at one FEC symbol interval and an uncorrectable codeword; a display unit that identifies and displays bit string data according to presence or absence of the FEC symbol error in FEC symbol length units based on a counting result; and display control means for performing display control by setting one zone of a display area of identification display as one FEC symbol length, matching a zone length of a horizontal axis of the display area with one codeword length, and performing line feed in codeword length units.

Memory devices, systems including memory devices, and methods of operating memory devices are described, in which memory devices are configured to poison data based on an indication provided by a host device coupled with the memory devices. The indication may include which one or more bits to poison (invert) at which stages of performing write or read operations. In some embodiments, the memory device may invert one or more bits according to the indication and then correct one or more errors associated with inverting the one or more bit to verify its on-die ECC functionality. In some embodiments, the memory device may provide the host device with poisoned data including one or more bits inverted according to the indication such that the host device may test system-level ECC functionality using the poisoned data.