telemedicine, mobile health, prosthetics, athletic training, human-machine interface (HMI) and so on. From "skin-like" electronics (a.k.a. e-skins) [1] to "epidermal electronics" (a.k.a. e-tattoos), [2] people are hopeful that the emerging flexible/ stretchable electronics technologies will disrupt the wearable industry. Specifically, e-tattoos are ultrathin, ultrasoft membranes that can well conform to the skin to monitor a variety of biomarkers including electrophysiology, [3] mechanoacoustic signals, [4] skin temperature, [5] skin hydration, [6] skin stiffness, [7] blood pressure, [8] blood oxygen saturation, [9] and even sweat analytes. [10] Wireless communication enabled by near field communication (NFC) [9a,11] and Bluetooth [12] have also been demonstrated in a few recent e-tattoos. However, which biomarker to measure is highly personal and may vary from time to time for the same individual. Moreover, different biomarkers should be measured at different locations using different types of sensors and read-out circuits. Even if one can build a multimodal e-tattoo, excessive recordings not personalized to the user may cause unnecessary power and bandwidth consumption, which is a major concern for wireless wearables. Although it is possible to build specific e-tattoos for specific sensing tasks, it would be a big waste of the wireless transmission and read-out circuits if the whole e-tattoo has to be disposed after just one use.Herein, we propose a possible remedy for all the aforementioned challenges-the modular and reconfigurable e-tattoos, where layers of distinct functionalities (e.g., NFC layer, analog front end (AFE) layer, electrode layer, etc.) can be pre-fabricated as building blocks that can be picked and assembled into customized e-tattoos and can also be swapped out to form new e-tattoos. Electrical connections between the layers can be achieved through aligned vias. Compared with existing monolayer, fully pre-defined e-tattoos, the newly proposed modular and reconfigurable e-tattoos would have the following appealing advantages. First, the multilayer stack can effectively shrink the footprint of the e-tattoo on the skin, especially when numerous components and complex circuits are needed for signal read-out and wireless transmission. Second, when the measurement is done, only the passive electrode/sensor layer that has been in direct contact with the skin needs to be peeled off from the e-tattoo and disposed. As a result, the leftover NFC In the past few years, ultrathin and ultrasoft epidermal electronics (a.k.a. e-tattoos) emerged as the next-generation wearables for telemedicine, mobile health, performance tracking, human-machine interface (HMI), and so on. However, it is not possible to build an all-purpose e-tattoo that can accommodate such a wide range of applications. Thus, the design, fabrication, and validation of modular and reconfigurable wireless e-tattoos for personalized sensing are reported. Such e-tattoos feature a multilayer stack of stretchable layers of distinct function...

No abstract

No abstract

A high energy-efficiency switching scheme for a successive approximation register (SAR) analogue-to-digital converter (ADC) is presented. The proposed method can achieve 98.4% savings in switching energy when compared to a conventional SAR. The proposed technique also achieves a 4 × reduction in total capacitance used in the digital-to-analogue converter (DAC) compared to the conventional DAC.

Automatic transistor sizing is a challenging problem in circuit design due to the large design space, complex performance tradeoffs, and fast technological advancements. Although there has been plenty of work on transistor sizing targeting on one circuit, limited research has been done on transferring the knowledge from one circuit to another to reduce the re-design overhead. In this paper, we present GCN-RL Circuit Designer, leveraging reinforcement learning (RL) to transfer the knowledge between different technology nodes and topologies. Moreover, inspired by the simple fact that circuit is a graph, we learn on the circuit topology representation with graph convolutional neural networks (GCN). The GCN-RL agent extracts features of the topology graph whose vertices are transistors, edges are wires. Our learning-based optimization consistently achieves the highest Figures of Merit (FoM) on four different circuits compared with conventional black-box optimization methods (Bayesian Optimization, Evolutionary Algorithms), random search, and human expert designs. Experiments on transfer learning between five technology nodes and two circuit topologies demonstrate that RL with transfer learning can achieve much higher FoMs than methods without knowledge transfer. Our transferable optimization method makes transistor sizing and design porting more effective and efficient.