Abstract: A multi-bit flash memory and a reading method thereof. Multiple reference memory cells for saving reserved data are provided to operate together with multiple data memory cells. Before the data memory cells are read, data stored in the reference memory cell is sensed based on a present reference current. Then, a value of a new reference current for reading the data memory cells is determined according to a difference between the sensed data and the reserved data.

Abstract: A method for programming a MLC memory includes (a) programming the bits of the memory having a Vt level lower than the PV level of a targeted programmed state into programmed bits by using a Vd bias BL; (b) ending this method if each bit of the memory has a Vt level not lower than the PV level of the targeted programmed state, otherwise, continuing the step (c); and (c) setting BL=BL+K1 and repeating the step (a) if each of the programmed bits has a Vt level lower than the PV level, while setting BL=BL?K2, and repeating the step (a) if at least one of the programmed bits has a Vt level not lower than the PV level.

Abstract: A multi-bit flash memory and a reading method thereof. Multiple reference memory cells for saving reserved data are provided to operate together with multiple data memory cells. Before the data memory cells are read, data stored in the reference memory cell is sensed based on a present reference current. Then, a value of a new reference current for reading the data memory cells is determined according to a difference between the sensed data and the reserved data.

Abstract: A multi-bit flash memory and a reading method thereof. Multiple reference memory cells for saving reserved data are provided to operate together with multiple data memory cells. Before the data memory cells are read, data stored in the reference memory cell is sensed based on a present reference current. Then, a value of a new reference current for reading the data memory cells is determined according to a difference between the sensed data and the reserved data.

Abstract: A method for programming a MLC memory is provided. The MLC memory has a number of bits, and each bit has a number of programmed states. Each programmed state has a first PV level. The method comprises (a) programming the bits of the memory having a Vt level lower than the PV level of a targeted programmed state into programmed bits by using a Vd bias BL; (b) ending this method if each bit of the memory has a Vt level not lower than the PV level of the targeted programmed state, otherwise, continuing the step (c); and (c) setting BL=BL+K1 and repeating the step (a) if each of the programmed bits has a Vt level lower than the PV level, while setting BL=BL?K2, and repeating the step (a) if at least one of the programmed bits has a Vt level not lower than the PV level.

Abstract: Methods for programming and program verification of a flash memory are described that ease the buried drain contact induced operation and increase the retention window. In a first aspect of the invention, a program operation method provides varying program biases which are applied to different groups of memory cells. The program biases can be supplied as drain bias voltages or gate bias voltages. The program biases vary depending on which group of memory cells is programmed. In one embodiment, a first drain voltage VD1 is applied to the first group of memory cells M0 and Mn. A second drain voltage VD2 is applied to the second group of memory cells M1 and Mn-1, where VD2=VD1+?VD. In a second aspect of the invention, a plurality of program verification voltage levels are selected to verify that the memory cells pass the program voltage levels.

Abstract: A method for programming a MLC memory is provided. The MLC memory has a number of bits, and each bit has a number of programmed states. Each programmed state has a first PV level. The method includes (a) programming the bits of the memory having a Vt level lower than the PV level of a targeted programmed state into programmed bits by using a Vd bias BL; (b) ending this method if each bit of the memory has a Vt level not lower than the PV level of the targeted programmed state, otherwise, continuing the step (c); and (c) setting BL=BL+K1 and repeating the step (a) if each of the programmed bits has a Vt level lower than the PV level, while setting BL=BL?K2, and repeating the step (a) if at least one of the programmed bits has a Vt level not lower than the PV level.

Abstract: A multi-bit flash memory and a reading method thereof. Multiple reference memory cells for saving reserved data are provided to operate together with multiple data memory cells. Before the data memory cells are read, data stored in the reference memory cell is sensed based on a present reference current. Then, a value of a new reference current for reading the data memory cells is determined according to a difference between the sensed data and the reserved data.

Abstract: A method for programming a MLC memory is provided. The MLC memory has a number of bits, and each bit has a number of programmed states. Each programmed state has a first PV level. The method comprises (a) programming the bits of the memory having a Vt level lower than the PV level of a targeted programmed state into programmed bits by using a Vd bias BL; (b) ending this method if each bit of the memory has a Vt level not lower than the PV level of the targeted programmed state, otherwise, continuing the step (c); and (c) setting BL=BL+K1 and repeating the step (a) if each of the programmed bits has a Vt level lower than the PV level, while setting BL=BL?K2, and repeating the step (a) if at least one of the programmed bits has a Vt level not lower than the PV level.

Abstract: Methods for programming and program verification of a flash memory are described that ease the buried drain contact induced operation and increase the retention window. In a first aspect of the invention, a program operation method provides varying program biases which are applied to different groups of memory cells. The program biases can be supplied as drain bias voltages or gate bias voltages. The program biases vary depending on which group of memory cells is programmed. In one embodiment, a first drain voltage VD1 is applied to the first group of memory cells M0 and Mn. A second drain voltage VD2 is applied to the second group of memory cells M1 and Mn-1, where VD2=VD1+?VD. In a second aspect of the invention, a plurality of program verification voltage levels are selected to verify that the memory cells pass the program voltage levels.

Abstract: A process and a memory architecture for operating a charge trapping memory cell is provided. The method for operating the memory cell includes establishing a high threshold state in the memory cell by injecting negative charge into the charge trapping structure to set a high state threshold. The method includes using a self-converging biasing procedure to establish a low threshold state for the memory cell by reducing the negative charge in the charge trapping structure to set the threshold voltage for the cell to a low threshold state. The negative charge is reduced in the memory cell by applying a bias procedure including at least one bias pulse. The bias pulse balances charge flow into and out of the charge trapping layer to achieve self-convergence on a desired threshold level. Thereby, an over-erase condition is avoided.

Abstract: A process and a memory architecture for operating a charge trapping memory cell is provided. The method for operating the memory cell includes establishing a high threshold state in the memory cell by injecting negative charge into the charge trapping structure to set a high state threshold. The method includes using a self-converging biasing procedure to establish a low threshold state for the memory cell by reducing the negative charge in the charge trapping structure to set the threshold voltage for the cell to a low threshold state. The negative charge is reduced in the memory cell by applying a bias procedure including at least one bias pulse. The bias pulse balances charge flow into and out of the charge trapping layer to achieve self-convergence on a desired threshold level. Thereby, an over-erase condition is avoided.

Abstract: A method for fabricating a non-volatile memory having a P-type floating gate is described. A tunneling layer is formed on a substrate and then a first patterned polysilicon layer is formed on the tunneling layer. A buried drain is formed in the substrate beside the first polysilicon layer and then an insulating structure is formed on the tunneling layer on the buried drain. Thereafter, a second polysilicon layer is formed on the first polysilicon layer to constitute a floating gate together with the first polysilicon layer. A P-type ion is implanted into the second polysilicon layer and then a dielectric layer and a control gate are sequentially formed on the floating gate. A thermal process is then performed to make the P-type, ion in the second polysilicon layer diffuse into the first polysilicon layer.

Abstract: A method for operating a P-channel SONOS memory device that has a charge trapping layer located on a substrate, a gate electrode located on the trapping layer, two doped regions located in the substrate at each side of the charge trapping layer. The two doped regions are set to be a drain region and a source region. When a programming action is intended, the gate electrode and the drain region are applied with a first negative high-level bias, and the source region and the substrate are applied with a grounded voltage. When an erasing action is intended, the gate electrode is a second negative bias which is smaller than the first negative voltage in absolute value. In the mean time, the drain region is applied with the third negative bias and the substrate is applied with a grounded voltage. The third negative voltage is larger than the second negative bias in absolute value.

Abstract: A method for fabricating a non-volatile memory having a P-type floating gate is described. A tunneling layer is formed on a substrate and then a first patterned polysilicon layer is formed on the tunneling layer. A buried drain is formed in the substrate beside the first polysilicon layer and then an insulating structure is formed on the tunneling layer on the buried drain. Thereafter, a second polysilicon layer is formed on the first polysilicon layer to constitute a floating gate together with the first polysilicon layer. A P-type ion is implanted into the second polysilicon layer and then a dielectric layer and a control gate are sequentially formed on the floating gate. A thermal process is then performed to make the P-type, ion in the second polysilicon layer diffuse into the first polysilicon layer.

Abstract: A qualification test method for a non-volatile memory includes determining a relation curve between the programming voltage and the lifetime of the memory cell. A programming voltage with respect to the memory array within the expected lifetime is estimated. According to the relation curve, the accelerating test voltage and the test time period corresponding to the programming voltage operated in the expected lifetime are computed out. The test is performed for the test time period under the accelerating test voltage. All the memory cells at the programmed state are tested to see if the original programmed state still remains. If the programmed state remains, the memory array is judged to have the life period. If the programmed state does not remain, the memory array is judged to have no the life period.

Abstract: A method for operating a P-channel SONOS memory device that has a charge trapping layer located on a substrate, a gate electrode located on the trapping layer, two doped regions located in the substrate at each side of the charge trapping layer. The two doped regions are set to be a drain region and a source region. When a programming action is intended, the gate electrode and the drain region are applied with a first negative high-level bias, and the source region and the substrate are applied with a grounded voltage. When an erasing action is intended, the gate electrode is a second negative bias which is smaller than the first negative voltage in absolute value. In the mean time, the drain region is applied with the third negative bias and the substrate is applied with a grounded voltage. The third negative voltage is larger than the second negative bias in absolute value.

Abstract: A reliability test method for a non-volatile memory. A relation curve of gate voltage versus read current degradation rate is obtained. The read current degradation rate of an actual gate voltage is estimated. From the relation curve, an accelerated test gate voltage and a test time corresponding to the actual gate voltage are obtained. With the accelerated test gate voltage, the test is continuously performed within the test time. Afterward, a test result of the memory is then obtained and, by the result, it is judged whether the data is valid or not. If the data is right (retained), the memory can be guarantied to have an expected lifetime; if the data is wrong (lost), the memory is judged as fails to pass the lifetime test.

Abstract: A nitride read only memory device that includes a substrate having a source region, a drain region, and a channel region formed therebetween, a first oxide layer formed over the channel region, a nitride layer formed over the first oxide layer, a second oxide layer formed over the nitride layer, a gate structure formed over the second oxide layer, wherein a region in the substrate underneath the gate structure excludes one of the source and drain regions, a plurality of sidewall spacers formed over the nitride layer and contiguous with the gate structure, and at least one injection point for injecting electrons into the nitride layer, wherein the injection point is located at a junction between the channel region and one of the source and drain regions, and wherein electron charges are stored in portions of the nitride layer underneath the sidewall spacers.

Abstract: An accelerated test for a non-volatile memory. A threshold voltage variation standard for assessment is selected. A set of negative gate bias voltages is applied to the gate terminals of the non-volatile memory to conduct the accelerated testing and obtain a test result. A curve relating lifetime and negative gate bias voltage is derived from the test result. According to the threshold voltage variation standard, the lifetime of the non-volatile memory is found. A word line negative gate bias voltage generator is coupled to a word line driver to apply a set of negative gate bias voltages to the gate terminals of programmed memory cells and conduct an accelerated testing.