Search

Read-disturb Modeled Artificial Neural Network Coupled LDPC ECC (RDNN-LDPC) is proposed to correct errors of read-hot data for 3D-TLC NAND flash. Conventional ANN-LDPC is optimized to correct errors of read-cold data. However, ANN-LDPC does not correct errors of read-hot data. To correct errors of read-hot data, this paper analyzes how input parameter and model change. As a result, measured results of proposed RDNN-LDPC extend acceptable read cycle of 3D-TLC NAND flash by 10-times.

International audience; Highly moisture-​resistant optical adhesives need to be developed for optical devices to ensure that they can be used reliably in optical devices for outside environment use. Further, a principal aim is that these adhesives have less toxicity impact on the environment and also on humans, so they need to be eco-​friendly. A Vegetable Oil Based Epoxidized Thermoset (VOBET) derived from linseed, was tested as an optical adhesive after being cured with different hardeners. A conventional optical adhesive, the petroleum-​based, Epotek 353ND was used as a ref. material. With increasing humidity the ref. material's shear strength decreased significantly and interface peeling was obsd. for all samples, while VOBET samples maintained a certain level of strength. One particular VOBET formulation showed exceptional reliability. A comparison of shear strength before and after a pressure cooker test of thermally cured VOBET samples and Epotek 353ND samples showed that the strength of the latter decreased significantly while the former maintained a certain ratio of the initial strength. Furthermore, one particular VOBET formulation that was successfully cured at room temp. showed ∼7.2 MPa shear strength, which is comparable to that of Epotek 353ND.