Freescale Semiconductor,inc.

Nxp Usa, Inc.

Richard Eguchi

Chen He

Jon Scott Choy

Jon W Weilemann, Ii

Thomas Jew

Yanzhuo Wang

Anirban Roy

Ronald J Syzdek

Peter J Kuhn

Daniel Hadad

Katrina M Prosperi

Craig T Swift

Fuchen Mu

Jacob Williams

Thomas S Harp

Chu-Chung Lee

Peter R Harper

Tu-Anh N Tran

Kerry A Ilgenstein

Padmaraj Sanjeevarao

Timothy J Strauss

William M Williams

Horacio P Gasquet

Kelly K Taylor

Craig D Gunderson

Lois E Yong

Richard Kenneth Glaeser

Benjamin A Schmid

Sung-Wei Lin

Jon W Weilmann, Ii

Larry J Grieve

Melvin Guison Mangibin

Nathan I Moon

A memory includes an array of resistive memory cells and circuitry for setting a write parameter for improving write effectiveness to the cells of the memory array. The circuitry performs a write parameter setting routine that determines a midpoint resistance of a memory state of cells of the array and determines a write efficiency of a weak write operation to cells of the array. Based on the determined midpoint resistance and the determined write efficiency, the circuit sets a write parameter level for subsequent writes to cells of the array.

Resistive memory with adjustable write parameter

As disclosed herein, a non-volatile memory circuit includes an array of memory cells. The non-volatile memory circuit also includes circuitry for performing a hard write to selective bits of addressed cells simultaneously with a normal write to the other bits of the addressed cells during a write operation to the addressed cells.

Non-volatile memory with selectable hard write

Embodiments of a magnetoresistive random access memory (MRAM) diagnostic system are provided, which includes: preconditioning all bit cells in an MRAM cell array to a data value of one during a diagnostic mode, wherein the MRAM cell array is implemented in an active side of a semiconductor substrate; applying a first magnetic disturb field having a predetermined field strength to the MRAM cell array, subsequent to the preconditioning, wherein the first magnetic disturb field is generated by an antenna implemented in a number of layers of conductive and dielectric material over the active side of the semiconductor substrate; performing a first error correcting code (ECC) read operation to read the MRAM cell array, subsequent to the applying the first magnetic disturb field; and in response to detecting at least one uncorrectable read during the first ECC read operation, setting a fail state and exiting the diagnostic mode.

Magnetic disturb diagnostic system for MRAM

A method and memory for stressing a plurality of non-volatile memory cells is provided. The method includes entering a memory cell stressing mode and providing one or more erase stress pulses to the plurality of non-volatile memory cells; determining that a threshold voltage of at least a subset of the plurality of non-volatile memory cells has a first relationship that is either greater than or less than a first predetermined voltage; providing one or more program stress pulses to the plurality of memory cells; and determining that the threshold voltage of at least a subset of the plurality of memory cells has a second relationship to a second predetermined voltage that is different than the first relationship.

Method and apparatus for stressing a non-volatile memory

An operating voltage and reference current are adjusted in a memory device. At least a portion of an array of memory cells is preconditioned to an erased state using an erase verify voltage on word lines coupled to the memory cells and a first reference current in sense amplifiers coupled to bit lines for the array. A test reference current is set for the sense amplifiers. A bitcell gate voltage is set on the word lines to a present overdrive voltage. The at least a portion of the array is read. If any of the memory cells in the at least a portion of the array are read as being programmed, the present overdrive voltage is increased until none of the memory cells in the at least a portion of the array are read as being programmed.

Non-volatile memory (NVM) with endurance control

Methods and systems are disclosed for imminent read failure detection based upon changes in error voltage windows for non-volatile memory (NVM) cells. In certain embodiments, data stored within an array of NVM cells is checked at a first time using a diagnostic mode and high/low read voltage sweeps to determine a first error voltage window where high/low uncorrectable errors are detected. Stored data is then checked at a second time using the diagnostic mode and high/low read voltage sweeps to determine a second error voltage window where high/low uncorrectable errors are detected. The difference between the error voltage windows are then compared against a voltage difference threshold value to determine whether or not to indicate an imminent read failure condition. An address sequencer, error correction code (ECC) logic, and a bias generator can be used to implement the imminent failure detection.

Imminent read failure detection based upon changes in error voltage windows for NVM cells

Methods and systems are disclosed for imminent read failure detection based upon unacceptable wear for non-volatile memory (NVM) cells. In certain embodiments, a first failure time is recorded when a first diagnostic mode detects an uncorrectable error within the NVM cell array using a first set of read voltage levels below and above a normal read voltage level. A second failure time is recorded when a second diagnostic mode detects an uncorrectable error within the NVM cell array using a second set of read voltage levels below and above a normal read voltage level. The first and second failure times are then compared against a threshold wear time value to determine whether or not an imminent read failure is indicated. The diagnostic modes can be run separately for erased NVM cell distributions and programmed NVM cell distributions to provide separate wear rate determinations.

Imminent read failure detection based upon unacceptable wear for NVM cells

Methods and systems are disclosed for imminent read failure detection using high/low read voltage levels. In certain embodiments, data stored within an array of non-volatile memory (NVM) cells is checked using read voltage levels below and above a normal read voltage level. An imminent read failure is then indicated if errors are detected within the same address for both voltage checks. Further, data stored can be checked using read voltage levels that are incrementally decreased below and incrementally increased above a normal read voltage level. An imminent read failure is then indicated if read errors are detected within the same address for both voltage sweeps and if high/low read voltage levels triggering faults differ by less than a predetermined threshold value. An address sequencer, error correction code (ECC) logic, and a bias generator can be used to implement the imminent failure detection.

Imminent read failure detection using high/low read voltage levels

In accordance with at least one embodiment, an onboard analog-to-digital converter (ADC) on a system-on-a-chip (SOC) is utilized to determine whether a charge pump output for a non-volatile memory (NVM) is correct or not. The SOC is directed to wait until the output is within an expected range before moving to the next step in a start-up procedure. If the maximum allowed start-up time is exceeded, an error signal is sent to the SOC such that the application can react to it.

Non-volatile memory robust start-up using analog-to-digital converter

In accordance with at least one embodiment, a clock counter on a system (for example, a system-on-a-chip (SOC) or other system) is utilized to count a number of a clock edges of a memory clock within a predefined time based on a predetermined system clock frequency and, therefore, to determine whether the memory clock for a memory array (for example, a non-volatile memory (NVM) array or other memory array) is correct or not. The system is directed to wait until the count is within an expected range before moving to the next step in a start-up procedure. If the maximum allowed start-up time is exceeded, an error signal is sent to the system such that the application can react to it.

Robust memory start-up using clock counter

A method includes, in one implementation, performing a memory operation to place memory cells of a memory array to a first logic state using a voltage of a charge pump. A portion of the operation is performed on the memory cells using the voltage of the charge pump. A temperature of the memory array is compared to a threshold. If the temperature is above a reference level, a load on the charge pump is reduced by providing the voltage to only a reduced number of memory cells.

Temperature-based adaptive erase or program parallelism

A method of performing a write operation on memory cells of a memory array includes applying a first plurality of pulses the write operation on the memory cells in accordance with a first predetermined ramp rate, wherein the first plurality of pulses is a predetermined number of pulses; performing a comparison of a threshold voltage of a subset of the memory cells with an interim verify voltage; and if a threshold voltage of any of the subset of memory cells fails the comparison with the interim verify voltage, continuing the write operation by applying a second plurality of pulses on the memory cells in accordance with a second predetermined ramp rate which has an increased ramp rate as compared to the first predetermined ramp rate.

Non-volatile memory (NVM) with adaptive write operations

A non-volatile memory built-in self-trim mechanism is provided by which product reliability can be improved by minimizing drift of reference current used for accessing the non-volatile memory and for performing initial trimming of the reference current. Embodiments perform these tasks by using an analog-to-digital converter to provide a digital representation of the reference current (Iref) and then comparing that digital representation to a stored target range value for Iref and then adjusting a source of Iref accordingly. For a reference current generated by a NVM reference bitcell, program or erase pulses are applied to the reference cell as part of the trimming procedure. For a reference current generated by a bandgap-based circuit, the comparison results can be used to adjust the reference current circuit. In addition, environmental factors, such as temperature, can be used to adjust the measured value for the reference current or the target range value.

Built-in self trim for non-volatile memory reference current

A technique for detecting an imminent read failure in a non-volatile memory array includes applying a bulk read stress to a plurality of cells of the non-volatile memory array and determining whether the plurality of cells exhibit an uncorrectable error correcting code (ECC) read during an array integrity check at a margin read verify voltage level subsequent to the bulk read stress. The technique also includes providing an indication of an imminent read failure for the plurality of cells when the plurality of cells exhibit the uncorrectable ECC read during the array integrity check. In this case, the margin read verify voltage level is different from a normal read verify voltage level.

Stress-based techniques for detecting an imminent read failure in a non-volatile memory array

Non-volatile memory (NVM) systems and related methods adjust program/erase bias conditions for non-volatile memory (NVM) cells to improve performance and product lifetime of NVM systems. System embodiments include integrated NVM systems having an NVM controller, a bias voltage generator, and an NVM cell array. Further, the NVM systems can store performance degradation information and program/erase bias condition information within storage circuitry. The disclosed embodiments adjust program/erase bias conditions for the NVM cells based upon performance degradation determinations, for example, temperature-based performance degradation determinations and interim verify based performance degradation determinations.

Growing community of inventors

Austin, TX, United States of America

Richard Eguchi