Scholarlink

This work presents a breakthrough energy-efficient Dual-Data-Aware 8T Static Random Access Memory (SRAM) architecture that integrates multi-bit Compute-assisted Organic Thin-Film Transistor (OTFT) technology with a Generative Adversarial Network (GAN)-tuned refinement system for superior image fidelity. The innovative design employs a three-stage GAN model—Generator, Discriminator, and Refinement Network—to restore degraded image features including low contrast and color distortion, specifically optimized for flexible display applications. We introduce GAN Truncation in the OTFT-based 8T SRAM structure (GAN Trunc-O-SRAM) to achieve enhanced picture quality. The SRAM core leverages MoS₂/WS₂ OTFTs with h-BN encapsulation, significantly improving charge mobility while reducing leakage power and ensuring environmental stability. Performance evaluation reveals exceptional results: 66% leakage reduction, 62% faster read/write access, and 58% improved energy efficiency compared to conventional SRAM designs. Image quality metrics demonstrate remarkable improvements with 5×, 3.6×, and 28× enhancements in structural similarity (SSIM) indices, Peak signal-to-noise ratio (PSNR), and power efficiency over 6T, 8T, and truncated 8T SRAMs, respectively. The design also achieves superior signal-to-noise margins with 7.6%, 2.5%, and 3.9% improvements across different configurations. These results establish this architecture as an ideal solution for next-generation low-power, high-fidelity flexible image processing systems, marking a significant advancement in organic memory technology for display applications. © 2017 Elsevier Inc. All rights reserved.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A High Speed Energy Efficient GAN-Tuned Trunc-O-SRAM for Robust Flexible Image Intelligence Systems

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Surbhi Bharti,
  • Ashwni Kumar

摘要

This work presents a breakthrough energy-efficient Dual-Data-Aware 8T Static Random Access Memory (SRAM) architecture that integrates multi-bit Compute-assisted Organic Thin-Film Transistor (OTFT) technology with a Generative Adversarial Network (GAN)-tuned refinement system for superior image fidelity. The innovative design employs a three-stage GAN model—Generator, Discriminator, and Refinement Network—to restore degraded image features including low contrast and color distortion, specifically optimized for flexible display applications. We introduce GAN Truncation in the OTFT-based 8T SRAM structure (GAN Trunc-O-SRAM) to achieve enhanced picture quality. The SRAM core leverages MoS₂/WS₂ OTFTs with h-BN encapsulation, significantly improving charge mobility while reducing leakage power and ensuring environmental stability. Performance evaluation reveals exceptional results: 66% leakage reduction, 62% faster read/write access, and 58% improved energy efficiency compared to conventional SRAM designs. Image quality metrics demonstrate remarkable improvements with 5×, 3.6×, and 28× enhancements in structural similarity (SSIM) indices, Peak signal-to-noise ratio (PSNR), and power efficiency over 6T, 8T, and truncated 8T SRAMs, respectively. The design also achieves superior signal-to-noise margins with 7.6%, 2.5%, and 3.9% improvements across different configurations. These results establish this architecture as an ideal solution for next-generation low-power, high-fidelity flexible image processing systems, marking a significant advancement in organic memory technology for display applications. © 2017 Elsevier Inc. All rights reserved.