PVTVAD: PVT-Variation-Aware Fault Diagnosis in ReRAM-based CAM Arrays

Content Addressable Memories (CAMs) enable fast pattern matching and are widely used in modern systems. ReRAM-based CAMs (ReCAMs) offer high density, low power, and non-volatility; however, the interaction between intrinsic ReRAM defects and CMOS process variations introduces complex faulty behaviors that challenge conventional diagnosis. Moreover, ReCAM arrays are highly susceptible to process, voltage, and temperature (PVT) variations, whose impact is amplified in computation-in-memory (CiM) architectures due to analog sensing. Accurately localizing faulty cells in ReCAMs is particularly challenging because all cells in a row share the same match-line (ML), while only small ML voltage differences reveal location-dependent fault behavior. This paper presents a variation-aware Design-for-Testability (DfT) framework for precise ReCAM fault localization. The proposed DfT circuitry combines a tunable OpAmp and ADC with an iterative diagnosis flow that adaptively adjusts gain and sampling time in a multi-step process to improve diagnostic resolution under PVT variation. Results show that favorable voltage and temperature operating points can be exploited to accelerate diagnosis and reduce sensing overhead, while a partitioning-based diagnosis strategy reduces required gain by approximately 80% for large arrays. ... mehrAcross diverse fault scenarios and PVT corners, the proposed method localizes faults in only 2–8 diagnosis iterations and achieves approximately 2.29× lower diagnostic latency than March-based approaches, while incurring only 0.36% area overhead and up to 6.3% test-power overhead.