Search

This paper presents a four-level pulse amplitude modulation (PAM-4) receiver that incorporates a continuous time linear equalizer, a variable gain amplifier, a phase interpolator-based clock and data recovery, and a 4-tap direct decision feedback equalizer (DFE) for moderate channel loss applications in wireline communication. A dynamic current-mode logic comparator (DCMLC) is proposed and employed in the DFE. The DCMLC, which adopts dynamic logic, breaks the trade-off between the bandwidth and the clock to Q delay in the traditional current-mode logic comparator (CMLC). Compared with the traditional CMLC, the DCMLC reduces the clock to Q delay by 36%, which allows the implementation of a 4-tap direct DFE. Moreover, the first tap feedback signals are directly tapped from the output of the DCMLC, allowing the first tap feedback current to initiate 0.5UI before the decision clock. The PAM-4 receiver prototype is fabricated in a 65nm CMOS process. At a data rate of 56-Gbps, it can compensate for up to 20.17dB loss and achieve a bit error rate $< 1\text{E}$ -10 with a power efficiency of 4.75 pJ/bit. [ABSTRACT FROM AUTHOR]

The Ring-AllReduce framework is currently the most popular solution to deploy industry-level distributed machine learning tasks. However, only about half of the maximum bandwidth can be achieved in the optimal condition. In recent years, several in-network aggregation frameworks have been proposed to overcome the drawback, but limited hardware information have been disclosed. In this paper, we propose a scalable fully-pipelined architecture that handles tasks like forwarding, aggregation and retransmission with no bandwidth loss. The architecture is implemented on a Xilinx Ultrascale FPGA that connects to 8 working servers with 10 Gb/s network adapters, and it is able to scale to more complicated scenarios involving more workers. Compared with Ring-AllReduce, using AllReduce-Switch improves the efficient bandwidth of AllReduce communication with a ratio of $1.75\times $. In image training tasks, the proposed hardware architecture helps to achieve up to $1.67\times $ speedup to the training process. For computing-intensive models, the speedup from communication may be partially hidden by computing. In particular, for ResNet-50, AllReduce-Switch improves the training process with MPI and NCCL by $1.30\times $ and $1.04\times $ respectively. [ABSTRACT FROM AUTHOR]

This paper investigates an ultra-broadband sampling technique based on charge sampling using an Integrate-and-Hold Circuit (IHC) and ultra-short integration times. The charge sampling technique is mathematically analyzed in detail and compared to conventional switched-capacitor sampling. The mathematical analysis allows to predict the sampler bandwidth as well as the degradation of sampling precision due to analog circuit impairments such as integrator gain error, integration capacitor leakage, hold-mode droop, thermal noise, and clock jitter. Furthermore, design, simulation, and measurement results of an ultra-broadband charge sampler IC in SiGe BiCMOS technology are presented. The charge sampler IC achieves a 1dB bandwidth of 70 GHz. A resolution of better than 5.9 effective number of bits (ENOB) is measured from 0 to 70 GHz at a sampling rate of 5 GS/s. The results suggest that charge sampling using an IHC is a viable concept for ultra-broadband sampling. [ABSTRACT FROM AUTHOR]

This paper addresses the problem of simultaneous fault detection and control (SFDC) for Markov jump systems (MJSs) over sensor networks. In order to save the bandwidth of the network, round-Robin protocol is adopted for the communication between the sensors and fault detection filter. First, to fully utilize the data of sensor networks, a new distributed system is developed especially for SFDC, and a residual system, which integrates the MJSs with the distributed SFDC system, is also established. Then, by constructing a mode-dependent Lyapunov functional, the conditions for distributed SFDC are derived, which can make the residual system asymptotically mean-square stable and meanwhile satisfy the prescribed H∞ performance requirements. Finally, comparison analyses are made by examples to illustrate the effectiveness and superiority of the proposed new design approach. [ABSTRACT FROM AUTHOR]

The continuous-time pipelined (CTP) ADC is a promising emerging high-speed analog-to-digital conversion technique that achieves anti-alias filtering and analog-to-digital conversion in one step. Driving such a converter is easy, thanks to its resistive input impedance. RC time-constant shifts, which will occur in practice due to a change in ambient temperature, degrade the performance of such converters. The aim of this work is to understand this phenomenon, quantify the resulting SNDR degradation, and thereby derive design tradeoffs. The theory is compared with measurements from a three-stage CTP that targets 70dB SNDR in a 100MHz bandwidth while sampling at 800MS/s. [ABSTRACT FROM AUTHOR]