摘要:
In a memory system comprising a plurality of memory units each of which possesses unit-level error correction capabilities and each of which are tied to a system level error correction function, memory reliability is enhanced by providing means for disabling the unit-level error correction capability, for example, in response to the occurrence of an uncorrectable error in one of the memory units. This counter-intuitive approach which disables an error correction function nonetheless enhances overall memory system reliability since it enables the employment of the complement/recomplement algorithm which depends upon the presence of reproducible errors for proper operation. Thus, chip level error correction systems, which are increasingly desirable at high packaging densities, are employed in a way which does not interfere with system level error correction methods.
摘要:
In a memory system comprising a plurality of memory units each of which possesses unit-level error correction capabilities and each of which are tied to a system level error correction function, memory reliability is enhanced by providing a mechanism for disabling the unit-level error correction capability, for example, in response to the occurrence of an uncorrectable error in one of the memory units. This counter-intuitive approach which disables an error correction function nonetheless enhances overall memory system reliability since it enables the employment of the complement/recomplement algorithm which depends upon the presence of reproducible errors for proper operation. Thus, chip level error correction systems, which are increasingly desirable at high packaging densities, are employed in a way which does not interfere with system level error correction methods.
摘要:
In a memory system comprising a plurality of memory units each of which possesses unit-level error correction capabilities and each of which are tied to a system level error correction function, memory reliability is enhanced by providing means for fixing the output of one of the memory units at a fixed value in response to the occurrence of an uncorrectable error in one of the memory units. This counter-intuitive approach to the generation of forced hard errors nonetheless enhances overall memory system reliability since it enables the employment of the complement/recomplement algorithm which depends upon the presence of reproducible errors for proper operation. Thus, clip level error correction systems, which are increasingly desirable at high packaging densities, are employed in a way which does not interfere with system level error correction methods.
摘要:
A memory system that provides extra data bits without utilizing storage capacity. A first data word is fetched from memory and corrected to remove any single-bit errors. A second data word (which is a subset of the first data word as corrected) is then fetched, and new data correction bits (parity or ECC check bits) is generated for the second data word. Both the second data word and the newly-generated data correction bits are output. This structure amortizes the expense of in-system data correction over a greater data output, and over a smaller storage capacity relative to the data output.
摘要:
A semiconductor memory device is described in which wordline redundancy is implemented without impacting the access time. A redundant decoder circuit generates a wordline drive inhibit signal which inhibits the generation of a normal wordline signal. Deselection also deselects the normally accessed reference cells, requiring that the redundant cells provide their own reference signal. This last requirement is accomplished by utilization of twin cells for the redundant memory. Placing the redundant memory cells on the sense node side of the bit line isolators enables the effective doubling of the number of redundant cells available to each of a plurality of sub-arrays of normal memory.
摘要:
A DRAM having on-chip ECC and both bit and word redundancy that have been optimized to support the on-chip ECC. The bit line redundancy features a switching network that provides an any-for-any substitution for the bit lines in the associated memory array. The word line redundancy is provided in a separate array section, and has been optimized to maximize signal while reducing soft errors. The array stores data in the form of error correction words (ECWs) on each word line. A first set of data lines (formed in a zig-zag pattern to minimize unequal capacitive loading on the underlying bit lines) are coupled to read out an ECW as well as the redundant bit lines. A second set of data lines receive the ECW as corrected by bit line redundancy, and a third set of data lines receive the ECW as corrected by the word line redundancy. The third set of data lines are coupled to the ECC block, which corrects errors encountered in the ECW. The ECC circuitry is optimized to reduce the access delays introduced by carrying out on-chip error correction. The ECC block provides both the corrected data bits and the check bits to an SRAM. Thus, the check bits can be externally accessed. At the same time, having a set of interrelated bits in the SRAM compensates for whatever access delays are introduced by the ECC. To maximize the efficiency of switching from mode to mode, the modes are set as a function of received address signals.
摘要:
This invention relates to semiconductor memories and includes a sense amplifier architecture in which sensed data bit or column lines are electrically isolated and shielded from their immediately adjacent active neighbors by utilization of non-selected bit lines as an AC ground bus. In its simplest embodiment, shielded bit line (SBL) architecture includes two pairs of opposed bit lines associated with a common sense amplifier. One of each of the bit line pairs is multiplexed into the sense amplifier and the other unselected bit line pair is clamped to AC ground to shield the selected bit line pair from all dynamic line-to-line coupling.
摘要:
A first, “known good” reference off-chip driver circuit actuated by an initial logic program (IPL) input signal has an output lead connected as one of the inputs to a comparator circuit for providing a reference off-chip driver output signal. A second off-chip driver circuit including a plurality of “n” separate driver circuit paths connected to input signal and produces output signals connected to a common node to provide output driver signals to the common node. The common node is connected to the second input of the comparator circuit for comparison with the reference off-chip driver output signal from the first off-chip driver circuit to determine the operating state of the second off-chip driver circuit with respect to the operating state of the first off-chip driver circuit.
摘要:
An integrated memory cube structure and method of fabrication wherein stacked semiconductor memory chips are integrated by a controlling logic chip such that a more powerful memory architecture is defined with the functional appearance of a single, higher level memory chip. A memory/logic cube is formed having N memory chips and at least one logic chip, with each memory chip of the cube having M memory devices. The controlling logic chip coordinates external communication with the N memory chips such that a single memory chip architecture with N.times.M memory devices appears at the cube's I/O pins. A corresponding fabrication technique includes an approach for facilitating metallization patterning on the side surface of the memory subunit.
摘要:
A read only memory array of stacked IGFET devices composed of first and second sub-arrays of field effect transistors. The first sub-array of first field effect transistors is formed in a substrate. Each of the first field effect transistor devices is responsive to a polysilicon gate electrode. The second sub-array of second field effect transistors is formed in a layer of laser annealed polysilicon material which overlies the first sub-array. The gate electrodes of the first field effect transistors act as the gate electrodes of the second field effect transistors.