Abstract: A dual-port memory system is implemented using single-port memory cells. An access arbiter having a synchronization circuit is used to prioritize and synchronize the access requests associated with the two ports. The access arbiter can also prioritize and synchronize refresh requests, in the case where the single-port memory cells require refresh. Access requests on the two ports and the refresh requests can be asynchronous. The access arbiter synchronizes the various requests by latching the requests into first-stage registers when a row access signal (RAS) is activated, and subsequently latching the contents of the first-stage registers into second-stage registers after a selected delay.

Abstract: A non-volatile memory system including an array of cells, each having an access transistor and a capacitor sharing a floating gate. The access transistors in each row are fabricated in separate well regions, which are independently biased. Within each row, the source of each access transistor is coupled to a corresponding virtual ground line, and each capacitor structure is coupled to a corresponding word line. Alternately, the source of each access transistor in a column is coupled to a corresponding virtual ground line. Within each column, the drain of each access transistor is coupled to a corresponding bit line. Select memory cells in each row are programmed by band-to-band tunneling. Bit line biasing prevents programming of non-selected cells of the row. Programming is prevented in non-selected rows by controlling the well region voltages of these rows. Sector erase operations are implemented by Fowler-Nordheim tunneling.

Abstract: A non-volatile memory (NVM) cell fabricated on a semiconductor substrate, and including a floating gate electrode (which extends at least partially over all active regions of the NVM cell). The NVM cell also includes a PMOS access transistor located in a first n-type region, a PMOS control capacitor located in a second n-type region (separate from the first n-type region), and an NMOS programming transistor located in a p-type region. The floating gate electrode is a continuous electrode which extends over the active regions of the PMOS access transistor, the PMOS control capacitor and the NMOS programming transistor. Various array connections are provided for implementing arrays using this NVM cell. The PMOS access transistor and NMOS programming transistor can be replaced with an NMOS access transistor and a PMOS erase transistor, respectively, in an alternate embodiment.

Abstract: A high-speed partial-word write operation is performed in a memory system. First, an error correction code (ECC) word, which includes a data word and associated error correction bits, is read from a memory array. A word line and a plurality of sense amplifiers are enabled during this read operation. The read data word is corrected in response to the associated error correction bits, thereby creating a corrected data word. The corrected data word is merged with a write data word, thereby creating a merged write data word. Write error correction bits are generated in response to the merged write data word, and the merged write data word and write error correction bits are written to the memory array. The word line and the plurality of sense amplifiers remain enabled from the reading operation through the write operation, thereby speeding up the partial-word write operation.

Abstract: A word line driver is provided for accessing a DRAM cell embedded in a conventional logic process. The DRAM cell includes a p-channel access transistor coupled to a cell capacitor. The word line driver includes an n-channel transistor located in a p-well, wherein the p-well is located in a deep n-well. The deep n-well is located in a p-type substrate. A word line couples the drain of the n-channel transistor to the gate of the p-channel access transistor. A negative boosted voltage supply applies a negative boosted voltage to the p-well and the source of the n-channel transistor. The negative boosted voltage is less than ground by an amount equal to or greater than the threshold voltage of the p-channel access transistor. The deep n-well and the p-type substrate are coupled to ground. The various polarities can be reversed in another embodiment.

Abstract: A word line driver is provided for accessing a DRAM cell embedded in a conventional logic process. The DRAM cell includes a p-channel access transistor coupled to a cell capacitor. The word line driver includes an n-channel transistor located in a p-well, wherein the p-well is located in a deep n-well. The deep n-well is located in a p-type substrate. A word line couples the drain of the n-channel transistor to the gate of the p-channel access transistor. A negative boosted voltage supply applies a negative boosted voltage to the p-well and the source of the n-channel transistor. The negative boosted voltage is less than ground by an amount equal to or greater than the threshold voltage of the p-channel access transistor. The deep n-well and the p-type substrate are coupled to ground. The various polarities can be reversed in another embodiment.

Abstract: A method, apparatus, and system in which an embedded memory fabricated in accordance with a conventional logic process includes one or more electrically-alterable non-volatile memory cells, each having a programming transistor, a read transistor and a control capacitor, which share a common floating gate electrode. The under-diffusion of the source/drain regions of the programming transistor and control capacitor are maximized. In one embodiment, the source/drain regions of the programming transistor are electrically shored by transistor punch-through (or direct contact).

Abstract: A method and structure for implementing a DRAM memory array as a second level cache memory in a computer system. The computer system includes a central processing unit (CPU), a first level SRAM cache memory, a CPU bus coupled to the CPU, and second level cache memory which includes a DRAM array coupled to the CPU bus. When accessing the DRAM array, row access and column decoding operations are performed in a self-timed asynchronous manner. Predetermined sequences of column select operations are then performed in a synchronous manner with respect to a clock signal. A widened data path is provided to the DRAM array, effectively increasing the data rate of the DRAM array. By operating the DRAM array at a higher data rate than the CPU bus, additional time is provided for precharging the DRAM array. As a result, precharging of the DRAM array is transparent to the CPU bus.

Abstract: A method and apparatus for scaling an embedded DRAM array from a first process to a second process, wherein the scaling involves reducing the linear dimensions of features by a constant scale factor. From the first process to the second process, DRAM cell capacitor layout area is reduced by the square of the scale factor, while cell capacitance is reduced by the scale factor. The voltage used to supply the logic transistors is scaled down from the first process to the second process. However, the voltage used to supply the sense amplifiers remains constant in both processes. Thus, in an embedded DRAM array of the second process, sense amplifiers are supplied by a greater voltage than the logic transistors. This allows the sensing voltage of DRAM cells to be maintained from one process generation to another, while allowing memory size to scale with the square of the process scale factor.

Abstract: A word line driver is provided for accessing a DRAM cell embedded in a conventional logic process. The DRAM cell includes a p-channel access transistor coupled to a cell capacitor. The word line driver includes an n-channel transistor located in a p-well, wherein the p-well is located in a deep n-well. The deep n-well is located in a p-type substrate. A word line couples the drain of the n-channel transistor to the gate of the p-channel access transistor. A negative boosted voltage supply applies a negative boosted voltage to the p-well and the source of the n-channel transistor. The negative boosted voltage is less than ground by an amount equal to or greater than the threshold voltage of the p-channel access transistor. The deep n-well and the p-type substrate are coupled to ground. The various polarities can be reversed in another embodiment.

Abstract: An embedded memory system includes an array of dynamic random access memory (DRAM) cells, which are isolated with deep trench isolation, and logic transistors, which are isolated with shallow trench isolation. Each DRAM cell includes an access transistor and a capacitor structure. The capacitor structure is fabricated by forming a metal-dielectric-semiconductor (MOS) capacitor in a deep trench isolation region. A cavity is formed in the deep trench isolation, thereby exposing a sidewall region of the substrate. The sidewall region is doped, thereby forming one electrode of the cell capacitor. A gate dielectric layer is formed over the exposed sidewall, and a polysilicon layer is deposited over the resulting structure, thereby filling the cavity. The polysilicon layer is patterned to form the gate electrode of the access transistor and a capacitor electrode, which extends over the sidewall region and upper surface of the substrate.

Abstract: Write check bits are generated in a predictive manner for partial-word write transactions in a memory system implementing error code correction. A read data word and associated read check bits are read from an address of the memory. If an error exists in a byte of the read data word, this byte is identified. At the same time, one or more bytes of the uncorrected read data word are merged with one or more bytes of a write data word, thereby creating a merged data word. Write check bits are generated in response to the merged data word. If the merged data word includes a byte of the read data word, which contains an error, the write check bits are modified to reflect this error. The merged data word and the modified (or unmodified) write check bits are then written to the address of the memory.

Abstract: A non-volatile memory cell is fabricated using a conventional logic process, with minor modifications. The cell is fabricated by forming a shallow trench isolation (STI) region in a well region of a semiconductor substrate. A recessed region is formed in the STI region, wherein the recessed region extends into the STI region and exposes a sidewall region in the well region. A capacitor region is formed in the sidewall region. A dielectric layer is formed over the well region, including the sidewall region. A gate electrode is then formed over the dielectric layer, wherein a portion of the gate electrode extends into the recessed region. An access transistor of the cell is then formed in a self-aligned manner with respect to the gate electrode. A capacitor structure is formed by the gate electrode (in the recessed region), the dielectric layer on the sidewall region, and the capacitor region.

Abstract: A high-speed partial-word write operation is performed in a memory system. First, an error correction code (ECC) word, which includes a data word and associated error correction bits, is read from a memory array. A word line and a plurality of sense amplifiers are enabled during this read operation. The read data word is corrected in response to the associated error correction bits, thereby creating a corrected data word. The corrected data word is merged with a write data word, thereby creating a merged write data word. Write error correction bits are generated in response to the merged write data word, and the merged write data word and write error correction bits are written to the memory array. The word line and the plurality of sense amplifiers remain enabled from the reading operation through the write operation, thereby speeding up the partial-word write operation.

Abstract: A memory device that uses error correction code (ECC) circuitry to improve the reliability of the memory device in view of single-bit errors caused by hard failure or soft error. A write buffer is used to post write data, so that ECC generation and memory write array operation can be carried out in parallel. As a result there is no penalty in write latency or memory cycle time due to ECC generation. A write-back buffer is used to post corrected ECC words during read operations, so that write-back of corrected ECC words does not need to take place during the same cycle that data is read. Instead, write-back operations are performed during idle cycles when no external memory access is requested, such that the write back operation does not impose a penalty on memory cycle time or affect memory access latency.

Abstract: A vertical one-transistor, floating-body DRAM cell is fabricated by forming an isolation region in a semiconductor substrate, thereby defining a semiconductor island in the substrate. A buried source region is formed in the substrate, wherein the top/bottom interfaces of the buried source region are located above/below the bottom of the isolation region, respectively. A recessed region is etched into the isolation region, thereby exposing sidewalls of the semiconductor island, which extend below the top interface of the buried source region. A gate dielectric is formed over the exposed sidewalls, and a gate electrode is formed in the recessed region, over the gate dielectric. A drain region is formed at the upper surface of the semiconductor island region, thereby forming a floating body region between the drain region and the buried source region. Dielectric spacers are formed adjacent to the gate electrode, thereby covering exposed edges of the gate dielectric.

Abstract: A one-transistor, floating-body (1T/FB) dynamic random access memory (DRAM) cell is provided that includes a field-effect transistor fabricated using a process compatible with a standard CMOS process. The field-effect transistor includes a source region and a drain region of a first conductivity type and a floating body region of a second conductivity type, opposite the first conductivity type, located between the source region and the drain region. A buried region of the first conductivity type is located under the source region, drain region and floating body region. The buried region helps to form a depletion region, which is located between the buried region and the source region, the drain region and the floating body region. The floating body region is thereby isolated by the depletion region. A bias voltage can be applied to the buried region, thereby controlling leakage currents in the 1T/FB DRAM cell.

Abstract: A memory system is provided that includes an array of memory cells that require periodic refresh, and a temperature-adaptive refresh controller. Data retention time of the memory cells decreases exponentially as temperature increases. The temperature-adaptive refresh controller selects the refresh period of the memory cells in response to the subthreshold current of a reference transistor. The subthreshold current of the reference transistor increases exponentially as temperature increases As a result, the refresh period is empirically tied to the data retention time. Consequently, the power required for refresh operations decreases as temperature decreases. Power is therefore conserved in applications that operate predominantly at room temperature.

Abstract: A non-volatile memory cell is fabricated using a conventional logic process, with minor modifications. The cell is fabricated by forming a shallow trench isolation (STI) region in a well region of a semiconductor substrate. A recessed region is formed in the STI region, wherein the recessed region extends into the STI region and exposes a sidewall region in the well region. A capacitor region is formed in the sidewall region. A dielectric layer is formed over the well region, including the sidewall region. A gate electrode is then formed over the dielectric layer, wherein a portion of the gate electrode extends into the recessed region. An access transistor of the cell is then formed in a self-aligned manner with respect to the gate electrode. A capacitor structure is formed by the gate electrode (in the recessed region), the dielectric layer on the sidewall region, and the capacitor region.

Abstract: A method and structure for implementing a DRAM memory array as a second level cache memory in a computer system. The computer system includes a central processing unit (CPU), a first level SRAM cache memory, a CPU bus coupled to the CPU, and second level cache memory which includes a DRAM array coupled to the CPU bus. When accessing the DRAM array, row access and column decoding operations are performed in a self-timed asynchronous manner. Predetermined sequences of column select operations are then performed in a synchronous manner with respect to a clock signal. A widened data path is provided to the DRAM array, effectively increasing the data rate of the DRAM array. By operating the DRAM array at a higher data rate than the CPU bus, additional time is provided for precharging the DRAM array. As a result, precharging of the DRAM array is transparent to the CPU bus.