Numerous embodiments of a precision tuning algorithm and apparatus are disclosed for precisely and quickly depositing the correct amount of charge on the floating gate of a non-volatile memory cell within a vector-by-matrix multiplication (VMM) array in an artificial neural network. Selected cells thereby can be programmed with extreme precision to hold one of N different values.

A nonvolatile memory device is provided. A nonvolatile memory device comprises a word line, a bit line, a memory cell array including a first memory cell at an intersection region between the word line and the bit line, a word line voltage generating circuitry configured to generate a program voltage, the program voltage to be provided to the word line, a row decoder circuitry configured to receive the program voltage from the word line voltage generating circuitry and configured to provide the program voltage to the word line, a verification circuitry configured to generate a verification signal in response to verifying a success or a failure of programming of the first memory cell, and a control circuitry configured to apply the program voltage to the first memory cell in response to the verification signal, and configured to cut off the program voltage in response to the verification signal.

For a non-volatile memory that uses hard bit and soft bit data in error correction operations, to reduce the amount of soft bit data that needs to be transferred from a memory to the controller and improve memory system performance, the soft bit data can be compressed before transfer. After the soft bit data is read and stored into the internal data latches associated with the sense amplifiers, it is compressed within these internal data latches. The compressed soft bit data can then be transferred to the transfer data latches of a cache buffer, where the compressed soft bit data can be consolidated and transferred out over an input-output interface. Within the input-output interface, the compressed data can be reshuffled to put into logical user data order if needed.

For a non-volatile memory that uses hard bit and a soft bit data in error correction operations, architectures are introduced for the compression of the soft bit data to reduce the amount of data transferred over the memory's input-output interface. For a memory device with multiple planes of memory cells, the internal global data bus is segmented and a data compression circuit associated with each segment. This allows soft bit data from a cache buffer of a plane using one segment to transfer data between the cache buffer and the associated compression circuit concurrently with transferring data from a cache buffer of another plane using another segment, either for compression or transfer to the input-output interface.

For a non-volatile memory that uses hard bit and a soft bit data in error correction operations, an on-the-fly compression scheme is used for the soft bit data. As soft bit data is transferred to a memory's input-output interface, the soft bit data is compressed prior to transmission to the an ECC engine memory controller, while hard bit data is transferred in un-compressed form.

A nonvolatile memory device includes: a peripheral circuit for repeatedly performing program loops each including a program operation including a setup operation on the plurality of bit lines and an application operation of applying a program pulse to a selected word line and the verification operation, and a control logic circuit for controlling the peripheral circuit, wherein the peripheral circuit performs a first program loop of the program loops by: applying each a first and a second program pulses in each a first and a second section of the application operation, setting a first bit line to a first level and a second bit line to a second level lower than the first level from a start of the setup operation until an end of the first section, and resetting the first and the second bit line to the second level in the second section.

A memory device includes a memory array of memory cells. A page buffer is to apply, to a bit line, a first voltage or a second voltage that is higher than the first voltage during a program verify operation. Control logic operatively coupled with the page buffer is to perform operations including: causing a plurality of memory cells to be programmed with a first program pulse; measuring a threshold voltage for the memory cells; forming a threshold voltage distribution from the measured threshold voltages; classifying, based on the threshold voltage distribution, a first subset of the memory cells as having a faster quick charge loss than that of a second subset of the memory cells; and causing, in response to the classifying, the page buffer to apply the second voltage to the bit line during a program verify operation performed on any of the first subset of memory cells.

A memory device includes: a plurality of memory cells; a peripheral circuit for performing a program operation including a plurality of loops each including a program voltage apply step and a verify step by using a plurality of verify voltages; and a program operation controller for controlling the peripheral circuit to perform the program operation. The program operation controller includes: a verify voltage controller for changing a verify voltage interval as an interval between the plurality of verify voltages from a predetermined target loop among the plurality of loops; and a bit line voltage controller to control bit line voltages applied to bit lines connected to first memory cells and second memory cells in the program voltage apply steps of an (n+1)th loop and an (n+2)th loop, based on a verify result in the verify step of an nth loop among the plurality of loops.

A semiconductor memory device includes a memory string and a control circuit. The memory string includes a first memory cell connected to a first word line and a second memory cell adjacent to the first memory cell and connected to a second word line. The control circuit is configured to perform a multi-bit-data writing with respect to each of the first and second memory cells. The multi-bit-data writing includes, in order, a first programming to program the first memory cell, the first programming with respect to the second memory cell, a reading of first data from the first memory cell, a second programming to program the second memory cell, and a verification of data programmed in the second memory cell. The control circuit is configured to set a verify voltage to be applied to the second word line during the verification based on the first data.

A method of operating a semiconductor memory device programming selected memory cells to store bits of data in each of the selected memory cells includes foggy programming and fine programming.