27 October 2020 in IEEE Journal of Solid-State Circuits
Circuit and design techniques are presented for enhancing the performance and reliability of a 3-D-stacked high bandwidth memory-2 extension (HBM2E). A data-bus window extension technique is implemented to cope with reduced clock cycle time ranging from data-path architecture, through-silicon via (TSV) placement, and TSV-PHY alignment. A power TSV placement in the middle of array and at the chip edge along with a dedicated top metal for power mesh improves power IR drop by 62%. An on-die ECC (OD-ECC) scheme featuring a self-scrubbing function is designed to be orthogonal to system ECC. An uncorrectable bit error rate (UBER) is improved by 10⁵ times with the proposed OD-ECC and scrubbing scheme. A memory built-in self-test (MBIST) block supports low-frequency cell and core test in a parallel manner and all channel at-speed operation with adjustable ac parameters. The proposed parallel-bit MBIST reduces test time by 66%. A 16-GB HBM2E fabricated in the second generation of 10-nm class DRAM process achieves a bandwidth up to 640 GB/s (5 Gb/s/pin) and provides a stable bit-cell operation at a high temperature (e.g., 105 ° C).
23 February 2021 in IEEE Access
As the technology node of the dynamic random-access memory (DRAM) continues to decrease below the 10-nm-class, bit-cell failures due to the external environments have increased. As a result, DRAM vendors perform post package inspections to provide fault-free DRAMs to the end customers. However, post package inspections require considerable test costs. To overcome this issue, an in-DRAM built-in self-test (BIST) mechanism is implemented in this study as an alternative solution. Herein, we propose compact and high test-coverage features for the in-DRAM BIST that to resolve the area problem when applied to a commodity DRAM. The proposed BIST secures the same test coverage with a shorter time than the conventional BIST. The proposed BIST reduces the test time by 52% of the DDR BIST in functions with the same test coverages. Further, the implemented BIST can achieved an area overhead of 0.051% based on a 16Gb DDR4 DRAM in the second generation of the 10-nm-class DRAM process.
23 January 2026 in IEEE Transactions on Nuclear Science
Multi-bit flip-flops (MBFFs) are widely adopted in low-power digital designs due to their ability to reduce area and clock switching power by sharing clock buffers across multiple individual flip-flops. However, this clock resource sharing introduces new bit-upsets vulnerabilities in the form of single-event transients (SETs) within the shared clock buffer, distinct from standard storage-node single-event upset (SEU). This work presents neutron-induced soft error characterization results for MBFFs implemented with both standard and radiation-hardened dual interlocked storage cell (DICE) latches in a 0.9V, 28nm CMOS technology. Neutron irradiation experiment was conducted at the ChipIR facility in the Rutherford Appleton Laboratory, UK. A total of 84 test chips containing over 800,000 flip-flops were tested under multiple voltage conditions and data patterns. Upset mechanisms were classified into storage-node upsets, local clock buffer SETs, and global scan clock SETs. The results confirm that MBFFs, despite the use of DICE for hardened storage, exhibit significantly increased susceptibility to SETs in shared clock buffers. This study highlights an important trade-off between power efficiency and radiation resilience in multi-bit flip-flop designs, calling attention to the need for targeted mitigation strategies to ensure robust system operation in radiation prone environments.