Active biofluid management is central to the realization of wearable bioanalytical platforms that are poised to autonomously provide frequent, real-time, and accurate measures of biomarkers in epidermally-retrievable biofluids (e.g., sweat). Accordingly, here, a programmable epidermal microfluidic valving system is devised, which is capable of biofluid sampling, routing, and compartmentalization for biomarker analysis. At its core, the system is a network of individually-addressable microheater-controlled thermo-responsive hydrogel valves, augmented with a pressure regulation mechanism to accommodate pressure built-up, when interfacing sweat glands. The active biofluid control achieved by this system is harnessed to create unprecedented wearable bioanalytical capabilities at both the sensor level (decoupling the confounding influence of flow rate variability on sensor response) and the system level (facilitating context-based sensor selection/protection). Through integration with a wireless flexible printed circuit board and seamless bilateral communication with consumer electronics (e.g., smartwatch), contextually-relevant (scheduled/on-demand) on-body biomarker data acquisition/display was achieved.
In recent years, "edge intelligent computing power deployment" has become a key and difficult problem in key areas such as system structure and artificial intelligence. Some independent decision-making scenario requires the edge devices to have the ability to handle high energy efficient and high throughput tasks. However, restricted by the traditional accurate computing mode and memory-computation separation structure, the edge devices often fail to provide the computing power guarantee of high throughput and high energy efficiency, which seriously hinders the edge intelligent deployment. This paper summarizes the new computing model integrating memory and computation of deploying stochastic circuits around memory, and constructs an edge intelligent system taking stochastic computing as a core from three levels of microstructure of circuit, architecture of near-memory stochastic computing and system-level software/hardware cooptimization. It strives to break through the power consumption and performant bottleneck of traditional computing model, and greatly improves the deploying ability of complex intelligent computing on edge equipment.
16 April 2024 in IEEE Journal of Solid-State Circuits
This work presents a high-efficiency ten-level series-capacitor 24-to-1-V buck converter with three techniques to improve its efficiency and reliability. First, we propose a fast in situ efficiency tracking (FIT) technique to maximize efficiency across load conditions and process, voltage, and temperature (PVT) variations. The second technique is the power-FET code roaming technique. It distributes the stress on power-FET segments at light load conditions to avoid excessive aging on specific segments. Last but not least, we directly supply the gate drive voltage from converter switch nodes to save power and area. The proposed converter was fabricated in a 180-nm BCD process. It achieves a peak power efficiency of 93.89% at 405-mA load current. The proposed fast in situ tracking technique demonstrates an efficiency improvement of up to 34.74% compared to the baseline without efficiency tracking. It also increases the high-efficiency ( $>$ 85%) load range from 13.51 $\times$ to 76.92 $\times$ . Besides, the power-FET roaming technique mitigates the development of local hotspots, thus slowing down the on-time resistance degradation by 5.8 $\times$ .
23 October 2024 in IEEE Transactions on Very Large Scale Integration (VLSI) Systems
This article presents a Li-ion battery-compatible single-inductor-multiple-output (SIMO) buck converter that fulfills the power management need of an integrated sub-mW RISC-V microprocessor. The proposed converter can directly take a 4.2-V battery voltage and produce four power rails ranging from 1.8 V for I/O to 0.5 V for the processor core. The three-level input stage is chosen to reduce the inductor ripple size and switching loss, thus increasing power conversion efficiency (PCE). In addition, the fully digital implementation using novel domino flash analog–digital converters (ADCs) enables low static current. Also, pulse frequency modulation (PFM) results in a wide dynamic range. The proposed three-level SIMO converter has been prototyped in a 65-nm CMOS technology with the 32-bit RISC-V processor. Measurement results show that the converter achieves a 1000 $\times$ load current range (0.8 $\mu$ A–0.8 mA) to support the active or sleep modes of the processor. The converter marks the PCE of 56.2%–72.8%. Compared to the ideal buck-low-dropout voltage regulator (LDO) architecture (LDO-only), it improves the PCE by 23.8% (46.4%).