A Review of FPGA-Based Custom Computing Architecture for Convolutional Neural Network Inference

Convolutional neural network (CNN) has been widely adopted in many tasks. Its inference process is usually applied on edge devices where the computing resources and power consumption are limited. At present, the performance of general processors cannot meet the requirement for CNN models with high computation complexity and large number of pa-rameters. Field-programmable gate array (FPGA)-based custom computing architecture is a promising solution to further enhance the CNN inference performance. The software/hardware co-design can effectively reduce the computing overhead, and improve the inference performance while ensuring accuracy. In this paper, the mainstream methods of CNN structure design, hardware-oriented model compression and FPGA-based custom architecture design are summarized, and the improvement of CNN inference performance is demonstrated through an example. Challenges and possible research directions in the future are concluded to foster research efforts in this domain.     

电场力显微镜不同探针对表面电荷注入的影响

在电场力显微镜(EFM)下利用不同金属镀层微探针,在微纳米尺度下对聚酰亚胺薄膜的表面电荷生成特性进行研究。采用电场力显微镜导电探针在聚酰亚胺薄膜表面注入电荷,并对微纳米区域产生的电荷进行表征,结果表明不同金属镀层的微探针对聚酰亚胺薄膜上电荷注入效果不同。铂铱合金镀层具有比钴铬合金镀层更高的功函数fm,因此前者在金属-电介质接触中产生更大肖特基势垒,进而降低了电荷的注入程度。该研究为微纳米尺度下探索聚合物绝缘材料表面电荷生成、发展机理提供了一个新的研究方法和途径。     

一种低压太阳能LED草坪灯驱动控制集成电路设计

为了减少太阳能LED草坪灯控制系统的成本,降低对太阳能电池的要求,设计了一种能工作在0.8V的低压太阳能LED草坪灯驱动控制器.该电路的输出电流可以在5mA～100mA范围内调节,能够满足各种不同LED对电流的需求.控制充电回路通断的开关管集成在芯片内部,减少了外围器件,使得整个草坪灯系统只需外接一个电感、二极管、电容...     

一种最大长度序列四相码信号脉压性能的研究

介绍了四相编码信号产生及数字脉压实现的方法，分析了采样起始时刻和多普勒频率对脉压结果的影响，比较了两个信号部分叠加时线性脉压和硬限幅脉压的结果。     

Open‐Shell Donor–Acceptor Conjugated Polymers with High Electrical Conductivity

Conductive polymers largely derive their electronic functionality from chemical doping, processes by which redox and charge‐transfer reactions form mobile carriers. While decades of research have demonstrated fundamentally new technologies that merge the unique functionality of these materials with the chemical versatility of macromolecules, doping and the resultant material properties are not ideal for many applications. Here, it is demonstrated that open‐shell conjugated polymers comprised of alternating cyclopentadithiophene and thiadiazoloquinoxaline units can achieve high electrical conductivities in their native “undoped” form. Spectroscopic, electrochemical, electron paramagnetic resonance, and magnetic susceptibility measurements demonstrate that this donor–acceptor architecture promotes very narrow bandgaps, strong electronic correlations, high‐spin ground states, and long‐range π‐delocalization. A comparative study of structural variants and processing methodologies demonstrates that the conductivity can be tuned up to 8.18 S cm^?1. This exceeds other neutral narrow bandgap conjugated polymers, many doped polymers, radical conductors, and is comparable to commercial grades of poly(styrene‐sulfonate)‐doped poly(3,4‐ethylenedioxythiophene). X‐ray and morphological studies trace the high conductivity to rigid backbone conformations emanating from strong π‐interactions and long‐range ordered structures formed through self‐organization that lead to a network of delocalized open‐shell sites in electronic communication. The results offer a new platform for the transport of charge in molecular systems.     

An innovative two-stage data compression scheme using adaptive block merging technique

Test data has increased enormously owing to the rising on-chip complexity of integrated circuits. It further increases the test data transportation time and tester memory. The non-correlated test bits increase the issue of the test power. This paper presents a two-stage block merging based test data minimization scheme which reduces the test bits, test time and test power. A test data is partitioned into blocks of fixed sizes which are compressed using two-stage encoding technique. In stage one, successive blocks are merged to retain a representative block. In stage two, the retained pattern block is further encoding based on the existence of ten different subcases between the sub-block formed by splitting the retained pattern block into two halves. Non-compatible blocks are also split into two sub-blocks and tried for encoded using lesser bits. Decompression architecture to retrieve the original test data is presented. Simulation results obtained corresponding to different ISCAS′89 benchmarks circuits reflect its effectiveness in achieving better compression.     

改进的基于STUMPS架构的BIST电路设计

传统的STUMPS测试方法,存在测试时间长和故障覆盖率不够高的缺点.为减少测试时间,采用Test-Per-Clock方式和向量压缩的方法处理待测电路CUT;减少了测试时间;用随机测试模式加存储测试模式,来提高故障覆盖率.经ISCAS'85标准测试电路验证,新方案取得了令人满意的结果.     

电压倍增之像素设计

为了节省面板电路驱动芯片的功率损耗以及制作成本,本研究提出一种新的像素电路设计,而在设计中将会融合电荷泵电路。利用这种电路设计的像素可有效地将像素电极上的驱动电压提高到输入电压的2～3倍以上。此像素电路设计具有两个优势:第一,可以有效降低显示面板的像素功率损耗;第二,不需高电压的面板电路驱动芯片,因此可节省芯片的成本及功率损耗。由模拟结果可知,像素电极上的驱动电压确实可由此像素电路设计而提高到输入电压的2～3倍以上;而像素的功率损耗也可有效地降低,约为传统像素的1/2。     

IQ时序故障对数字脉冲压缩的影响

雷达系统中,数字脉冲压缩可以获得较高的距离分辨率。传统的数字脉冲压缩采用正交双通道处理,即复信号的,路和Q路分别送往硬件处理器,进行合成脉压。由于IQ颠倒和错位等时序问题,给脉压带来多种消极影响,对调试和研究造成一定的误导。文中分别以测试信号和实际采集数据对上述情况进行仿真,仿真结果证明多种时序错误对应的脉压波形,对工程设计和调试有一定的借鉴意义。     

网络理论在光伏充电站智能通信网络优化中的应用

介绍未来智能电网中光伏储能电动汽车充电站主体架构,及其通信监控架构。将网络中连通度、网络流等理论应用于光伏储能、电动汽车充电站监控系统设计中,最后通过对监控网络的连通度、稳定性、经济性的比较得出最优的监控系统设计框架。