Abstract: A BISR mode and associated method for testing memory. All redundant elements of the memory including the ones which are not used are tested, and interaction between redundant elements of the memory and adjacent functional memory are checked. Repair information is used to repair the memory. In addition, redundant elements which are not needed to be used for repairing the memory are forced to be used, such as by faking defects to remap good elements with redundant elements.

Abstract: A pin-based memory power modeling method using arc-based characterization includes steps as follows. All power arcs of a memory model are identified and characterized. A power arc is selected from the identified and characterized power arcs. Output bus switching power is computed by removing overlapping power using the selected power arc, and a temporary value for various input ramp times and output loads is derived. Output pin power for the selected power arc is calculated using the temporary value and a ratio of a number of output bits switching over an entire bus width. Switching power for the selected power arc is calculated by a power estimation tool based on port activity and an input intrinsic power value.

Abstract: A BISR scheme which provides for on-chip assessment of the amount of repair on a given memory and for the flagging of any device as a fail when the device exceeds a pre-determined limit. Preferably, a counter is built and loaded through a test pattern during production testing, and the counter establishes the threshold for pass/fail criteria. The BISR is configured to load a repair solution and then test the memories for any additional failures and if there are any, repair them (provided enough redundant elements are available). In addition, a reliability controller for BISR designs can be provided, where the reliability controller contains a register set and a number of counters at the chip-level which can be loaded through a test pattern during production tests, where one of the counters contains the number of memories to be allowed for repair.

Abstract: A test system includes a built-in self-test (BIST) circuit and a stress applicator for use in analyzing a memory array. The stress applicator applies a selective set of stress factors to the memory array, such as temperature and voltage conditions. The BIST circuit executes a test routine on the memory array to detect the presence of any faulty memory address locations that may arise under the prevailing stress condition. A full testing cycle involves iterative repetition of the functions performed by the stress applicator and BIST circuit, with variations in the stress factors across the testing iterations. An accumulator cumulatively stores the fault information generated by the BIST circuit during each testing iteration. Following completion of the testing cycle, a repair operation is performed by a built-in self-repair (BISR) circuit to remap the faulty memory address locations indicated by the accumulator to redundant memory address locations.

Abstract: A pin-based memory power modeling method using arc-based characterization includes steps as follows. All power arcs of a memory model are identified and characterized. A power arc is selected from the identified and characterized power arcs. Output bus switching power is computed by removing overlapping power using the selected power arc, and a temporary value for various input ramp times and output loads is derived. Output pin power for the selected power arc is calculated using the temporary value and a ratio of a number of output bits switching over an entire bus width. Switching power for the selected power arc is calculated by a power estimation tool based on port activity and an input intrinsic power value.

Abstract: A pin-based memory power modeling method using arc-based characterization includes steps as follows. All power arcs of a memory model are identified and characterized. A power arc is selected from the identified and characterized power arcs. Output bus switching power is computed by removing overlapping power using the selected power arc, and a temporary value for various input ramp times and output loads is derived. Output pin power for the selected power arc is calculated using the temporary value and a ratio of a number of output bits switching over an entire bus width. Switching power for the selected power arc is calculated by a power estimation tool based on port activity and an input intrinsic power value.

Abstract: A test system includes a built-in self-test (BIST) circuit and a stress applicator for use in analyzing a memory array. The stress applicator applies a selective set of stress factors to the memory array, such as temperature and voltage conditions. The BIST circuit executes a test routine on the memory array to detect the presence of any faulty memory address locations that may arise under the prevailing stress condition. A full testing cycle involves iterative repetition of the functions performed by the stress applicator and BIST circuit, with variations in the stress factors across the testing iterations. An accumulator cumulatively stores the fault information generated by the BIST circuit during each testing iteration. Following completion of the testing cycle, a repair operation is performed by a built-in self-repair (BISR) circuit to remap the faulty memory address locations indicated by the accumulator to redundant memory address locations.

Abstract: A method for allowing native, functional, and test configurations of a memory to be independent of one another includes steps as follows. A memory is first provided. The memory has a native configuration including k words and n data output pins, k and n being positive integers. Each of the k words has a width of n bits. Then the n data output pins are connected to a programmable multiplexer for multiplexing the n data output pins into at least one group of data output pins of the programmable multiplexer. Each of the at least one group of data output pins has no more than n data output pins and is suitable for enabling the memory to have at least one of a test configuration or a functional configuration. At user's discretion, the test configuration may or may not have a width of n bits, the functional configuration may or may not have a width of n bits, and the test configuration and the functional configuration may or may not have the same width.

Abstract: A semiconductor memory testing implementation suitable for build-in self repair (BISR) memories provides, in one embodiment, a memory testing circuit configuration including an output register for receiving digital data. A plurality of shift registers serially output the digital data to be received by the output register. Each one of the plurality of shift registers includes a feedback path for enabling the digital data output by a corresponding one of the plurality of shift registers to be input back into the corresponding shift register in a same sequence as the digital data is output from the corresponding shift register.

Abstract: A circuit is configured as a splittable duplex memory cell or as a joinable single port memory pair based on the state of a programming layer. The programming layer has two states. In one state, the programming layer configures the circuit as a joinable single port memory pair. In the other state it configures the circuit as a splittable duplex memory cell. As such, the circuit can act as either a dual port memory cell or as two single port memory cells.

Abstract: A method for testing semiconductor devices advantageously increases manufacturing yields. The method includes generating memory repair data for a wafer die by writing at least one predetermined digital bit pattern into a memory on the wafer die, reading the at least one predetermined digital bit pattern back out of the memory, comparing the at least one predetermined digital bit pattern read out from the memory against the at least one predetermined digital bit pattern written into the memory, and storing results of the comparison as the memory repair data. The writing and reading are performed a plurality of times, each time with a different voltage and clock frequency combination being applied to the wafer die. The memory repair data is programmed into the wafer die, and the wafer die is assembled into a packaged semiconductor device.

Abstract: A controller for repairing a redundant memory circuit includes a fault storage matrix for mapping a repair solution, a plurality of registers for storing row and column coordinates of the repair solution, and a repair solution calculator coupled to the plurality of registers and the fault storage matrix for receiving an x-coordinate and a y-coordinate of a defective memory cell in the redundant memory circuit and for determining whether a repair solution may be found from the fault storage matrix.

Abstract: A system and method for protecting the values stored in a BISR repair block and, optionally, debugging the BISR repair logic without altering normal test flow is implemented by a circuit including a plurality of soft latches within the BISR repair block, the soft latches being coupled together to form a BISR scan chain for holding BISR repair information. A chip level scan enable signal and a scan hold control signal cooperate to control connection of the BISR scan chain to other scan chains during a scan test, so that the BSR repair information is held within the soft latches. A diagnose enable signal cooperating with the chip level scan enable signal and the scan hold control signal for enabling debugging of logic connecting the BISR scan chains.

Abstract: A BISR scheme which provides that fuse blocks are shared between memories to reduce hard-BISR implementation costs. The scheme includes a plurality of memories serially connected to a fuse controller. A plurality of fuse blocks are also serially connected to the fuse controller. There are more memory instances than there are fuse blocks, and the fuse controller is configured to allow the fuse blocks to be shared between memories. Preferably, each fuse block includes fuse elements which can be programmed with the memory instance number which needs to be repaired. The fuse block reduces routing congestion and is preferably configured to provide the flexibility of assigning any fuse block to any instance that needs repair. The programmable fuse elements are preferably loaded into a counter (which is preferably part of the fuse controller) which ensures that the correct block information gets loaded into the corresponding memory instance.

Abstract: A memory having built-in self repair with row shifting is provided. The rows in the memory are divided into smaller row groups and a bad row group is repaired with a redundant row group. Each row group receives a row select signal, which is fed into a shift circuit for the row group and a shift circuit for an adjacent row group. A shift circuit is provided for the redundant row group and the shift circuit for the redundant row group receives the row select signal for only the adjacent row group. If a bad row group is detected, then starting with the row group furthest from the redundant row group, the shift circuit for each row group before the bad row group is deactivated. The row group select signal and word line signal for the bad row group are disabled. The shift circuit for the bad row group and the shift circuit for each row group after the bad row group are activated. Therefore, the bad row group is disabled and the redundant row group fills the void.

Abstract: An apparatus comprising a controller circuit and a BISR assembly circuit. The controller circuit may be configured to present one or more control signals. The control signals may be configured to control one or more built-in self-test (BIST) and built-in self-repair (BISR) modes of operation. The BISR assembly circuit generally comprises one or more memory blocks each comprising a counter configured to generate a clock cycle count value in response to a repair solution during the BIST and BISR operations. The memory blocks may be remapped in response to the count values during one or more of the BISR operations.

Abstract: A memory having built-in self repair with row shifting is provided. The rows in the memory are divided into smaller row groups and a bad row group is repaired with a redundant row group. Each row group receives a row select signal, which is fed into a shift circuit for the row group and a shift circuit for an adjacent row group. A shift circuit is provided for the redundant row group and the shift circuit for the redundant row group receives the row select signal for only the adjacent row group. If a bad row group is detected, then starting with the row group furthest from the redundant row group, the shift circuit for each row group before the bad row group is deactivated. The row group select signal and word line signal for the bad row group are disabled. The shift circuit for the bad row group and the shift circuit for each row group after the bad row group are activated. Therefore, the bad row group is disabled and the redundant row group fills the void.

Abstract: A controller for repairing a redundant memory circuit includes a fault storage matrix for mapping a repair solution, a plurality of registers for storing row and column coordinates of the repair solution, and a repair solution calculator coupled to the plurality of registers and the fault storage matrix for receiving an x-coordinate and a y-coordinate of a defective memory cell in the redundant memory circuit and for determining whether a repair solution may be found from the fault storage matrix.

Abstract: A self-time circuit and method are presented for reducing the write cycle time in a semiconductor memory. A “dummy” memory cell having the same timing requirements as the functional cells, and associated write logic are added to the standard circuitry of the memory device. The dummy write cell receives the same control signals used to write data to the functional cells of the memory, and is configured to issue a completion signal when a write access is concluded, causing the write cycle to be terminated. The circuit and method permits write cycle time to be reduced to the lowest practical value, independently of the read cycle time. This potentially increases the overall operating speed of the memory device. The circuit and method disclosed herein are adaptable to the most common types of memory devices, such as SRAM, DRAM and CAM.

Abstract: A method is presented for self-test and self-repair of a semiconductor memory device. Prior to the self-repair stage, both redundant and regular memory portions are comprehensively tested, preferably using a checkerboard bit pattern. Faulty rows identified in each memory portion during testing are recorded. Known-bad rows in regular memory are then replaced by known-good redundant rows in the self-repair stage, and the resulting repaired memory is retested for verification. Compared to existing methods, the new method is believed to provide improved test coverage, making it both more effective in identifying non-repairable memory devices and less prone to fail repairable ones.