Scalable JTAG-Based 32-Bit Memory Test Architecture with MATS+ and MATS++/March-C Fault Detection

Source : Verilog HDL

Proposed Abstract:

Embedded memories are increasingly used in advanced System-on-Chip (SoC) designs for applications such as networking, automotive control, and medical imaging, where reliability and performance are critical. Ensuring fault-free operation of these memories is essential, yet memory testing remains a major challenge. Conventional MBIST architectures, while effective, often introduce significant silicon overhead, add design complexity, and lack flexibility for post-fabrication updates. In addition, existing memory test algorithms have their own drawbacks: March-C is widely applied and provides high fault coverage, but it requires long test times due to bit-oriented operations and large numbers of read–write cycles; MATS+ is simple and efficient but suffers from lower coverage, particularly for coupling and complex dynamic faults; and MATS++ improves on MATS+ with better detection capability, yet it still trades off hardware cost and scalability when applied to larger 32-bit word-oriented memories. Furthermore, most existing implementations are optimized for small SRAMs and are not easily scalable to clustered embedded memories in SoCs, nor do they fully exploit standard boundary-scan infrastructure for low-cost testing. To address these problems, this work proposes a scalable JTAG-based 32-bit memory test architecture that reuses IEEE 1149.1 boundary-scan resources to apply and compare March-C, MATS+, and MATS++ algorithms in both single-bit and multi-bit test modes. The proposed framework minimizes additional hardware cost by integrating BIST control into boundary-scan registers, while enabling algorithm programmability and flexibility for different memory clusters. The novelty lies in providing a detailed performance comparison of these algorithms under a unified boundary-scan-based architecture, focusing on trade-offs between fault coverage, test time, and silicon overhead. The design is implemented in Verilog HDL and synthesized on an FPGA using Xilinx Vivado, where parameters such as area, power, and latency are evaluated to validate efficiency and practical applicability for SoC-level memory testing.

List of the following  materials will be included with the Downloaded Backup: