As the technology node of the traditional semiconductor integrated circuit entering the era of sub-10 nm, its computing speed and power dissipation are approaching their physical limits. Therefore, without a major overhaul, such kind of technology may face serious issues in fulfilling the ever-growing demand of computational capacity in big data era. Superconducting integrated digital circuit, i.e. superconducting single flux quantum (SFQ) circuit, a proven technology with high processing speed and low power dissipation, can be leveraged to overcome the bottleneck encountered by semiconductor technology. To achieve power efficiency goal set for “green” supercomputer in big data era, the power dissipation of superconducting SFQ circuit needs to be further reduced even though it is already orders of magnitude smaller than that of traditional semiconductor circuit. Most of the energy in the traditional SFQ circuit, invented in the 80s of the last century, is dissipated in maintaining its stable state of the elementary unit of digital circuit and in the delete operation during information processing. In this proposal, we plan to conduct a comprehensive investigation of the physical mechanism involved in the energy dissipation process in basic elements of SFQ circuit and by developing new SFQ basic elements we hope to find a stable state with much lower energy dissipation. Meanwhile, we will introduce new computation architecture, i.e., the reversible logic, which, in principle, can reduce the power dissipation below thermodynamics threshold. The combination of new basic elements and logic can be used to create an adiabatic SFQ system with dramatically improved systematic energy efficiency. The likely findings of this proposal help to build the theoretical and technological foundation for large-scale low-power-dissipation superconducting digital circuit in future.
随着传统微电子集成电路的技术节点跨入10纳米时代,其速度和功耗已经逼近物理极限而难以满足大数据时代快速增长的应用需求。超导数字电路技术,或称单磁通量子(SFQ)电路技术,凭借其高速、低功耗的优势,成为一项近期最有可能解决传统微电子技术面临之能效瓶颈的电路技术。为满足“绿色环保”超级计算机的能效要求,超导SFQ电路的功耗在已比传统微电子电路呈数量级优势的基础上仍需进一步的降低。而发明于上世纪80年代的传统SFQ电路中大部分功耗来自于维持计算基本元件稳态所需能耗以及在运算中删除信息步骤产生的能耗。本项目拟深入研究SFQ电路基本单元能耗的物理机制,从降低计算单元器件稳态能耗入手,研究开发新型低功耗计算单元。同时,在SFQ电路中引入可实现物理极限低功耗的可逆运算逻辑,避免删除信息步骤,实现整个SFQ计算系统的绝热变化,从而大幅提高SFQ电路能效,为大规模低功耗超导计算技术奠定基本理论和技术基础。
本研究首先针对多种新型超导单磁通量子(single flux quantum, SFQ)电路计算单元,基于半定制超导专用电子设计和自动化(electronic design automation, EDA)工具建立了超导SFQ电路仿真平台,实现了对超导集成电路的逻辑和数字仿真。确立了我国第一代超导SFQ集成电路的标准工艺流程和设计规则,构建了大规模电路设计所需逻辑单元门。从而完成了超导集成电路仿真设计,工艺制备以及测试表征的全套自主研究程序,填补了我国在超导集成电路研究领域的长期空白。同时研究了提高超导SFQ集成电路能率的电路设计方法,探索研究多种新型超导集成电路的计算单元并深入研究了新型低功耗超导ERSFQ电路偏置网络稳态功耗的机理,取得了多项创新成果,为加速我国在超导集成电路研究领域赶超国际先进水平奠定了重要的基石。
{{i.achievement_title}}
数据更新时间:2023-05-31
基于 Kronecker 压缩感知的宽带 MIMO 雷达高分辨三维成像
五轴联动机床几何误差一次装卡测量方法
气力式包衣杂交稻单粒排种器研制
基于相似日理论和CSO-WGPR的短期光伏发电功率预测
倒装SRAM 型FPGA 单粒子效应防护设计验证
新型复数算术运算单元设计与VLSI实现
基于单电子效应的纳米硅逻辑运算器
新型动态逻辑计算及芯片研究
光学二维信号并行数子处理和光逻辑运算