Search for: All records

Circuit linearity calibration can represent a set of high-dimensional search problems if the observability is limited. For example, linearity calibration of digital-to-time converters (DTC), an essential building block of modern digital phaselocked loops (DPLLs), is an example of a high-dimensional search problem as difficulty of measuring ps delays hinders prior methods that calibrate stage by stage. And, a calibrated DTC can become nonlinear again due to changes in temperature (T) and power supply voltage (V). Prior work reports a deep reinforcement learning framework that is capable of performing DTC linearity calibration with nonlinear calibration banks; however, this prior work does not address maintaining calibration in the face of temperature and supply voltage variations. In this paper, we present a meta-reinforcement learning (RL) method that can enable the RL agent to quickly adapt to a new environment when the temperature and/or voltage change. Inspired by the Style Generative Adversarial Networks (StyleGANs), we propose to treat temperature and voltage changes as the styles of the circuits. In contrast to traditional methods employing circuit sensors to detect changes in T and V, we utilize a machine learning (ML) sensor, to implicitly infer a wide range of environmental changes. The style information from the ML sensor is subsequently injected into a small portion of the policy network, modulating its weights. As a proof of concept, we first designed a 5-bit DTC at the normal voltage (1V) and normal temperature (27℃) corner (NVNT) as the environment. The RL agent begins its training in the NVNT environment. Following this initial phase, the agent is then tasked with adapting to environments with different temperature and supply voltages. Our results show that the proposed technique can reduce the Integral Non-Linearity (INL) to less than 0.5 LSB within 10, 000 search steps in a changed environment. Compared to starting learning from a random initialized policy and a trained policy, the proposed meta-RL approach takes 63% and 47% fewer steps to complete the linearity calibration, respectively. Our method is also applicable to the calibration of many other kinds of analog and RF circuits. 

This paper presents Millimetro, an ultra-low-power tag that can be localized at high accuracy over extended distances. We develop Mil-limetro in the context of autonomous driving to efficiently localize roadside infrastructure such as lane markers and road signs, even if obscured from view, where visual sensing fails. While RF-based localization offers a natural solution, current ultra-low-power local-ization systems struggle to operate accurately at extended ranges under strict latency requirements. Millimetro addresses this challenge by re-using existing automotive radars that operate at mmWave fre-quency where plentiful bandwidth is available to ensure high accuracy and low latency. We address the crucial free space path loss problem experienced by signals from the tag at mmWave bands by building upon Van Atta Arrays that retro-reflect incident energy back towards the transmitting radar with minimal loss and low power consumption. Our experimental results indoors and outdoors demonstrate a scal-able system that operates at a desirable range (over 100 m), accuracy (centimeter-level), and ultra-low-power (< 3 uW). 

Abstract—Robotic geo-fencing and surveillance systems require accurate monitoring of objects if/when they violate perimeter restrictions. In this paper, we seek a solution for depth imaging of such objects of interest at high accuracy (few tens of cm) over extended ranges (up to 300 meters) from a single vantage point, such as a pole mounted platform. Unfortunately, the rich literature in depth imaging using camera, lidar and radar in isolation struggles to meet these tight requirements in real-world conditions. This paper proposes Metamoran, a solution that explores long-range depth imaging of objects of interest by fusing the strengths of two complementary technologies: mmWave radar and camera. Unlike cameras, mmWave radars offer excellent cm-scale depth resolution even at very long ranges. However, their angular resolution is at least 10× worse than camera systems. Fusing these two modalities is natural, but in scenes with high clutter and at long ranges, radar reflections are weak and experience spurious artifacts. Metamoran’s core contribution is to leverage image segmentation and monocular depth estimation on camera images to help declutter radar and discover true object reflections.We perform a detailed evaluation of Metamoran’s depth imaging capabilities in 400 diverse scenarios. Our evaluation shows that Metamoran estimates the depth of static objects up to 90 m away and moving objects up to 305 m away and with a median error of 28 cm, an improvement of 13× over a naive radar+camera baseline and 23× compared to monocular depth estimation. 

This paper presents a two-layer RF/analog weighting MIMO transceiver that comprises fully-connected (FC) multi-stream beamforming tiles in the RF-domain first layer, followed by a fully connected analog- or digital-domain baseband layer. The architecture mitigates the complexity versus spectral-efficiency tradeoffs of existing hybrid MIMO architectures and enables MIMO stream/user scalability, superior energy-efficiency, and spatial-processing flexibility. Moreover, multi-layer architectures with FC tiles inherently enable the co-existence of MIMO with carrier-aggregation and full-duplex beamforming. A compact, reconfigurable bidirectional circuit architecture is introduced, including a new Cartesian-combining/splitting beamforming receiver/transmitter, dual-band bidirectional beamforming network, dual-band frequency translation chains, and baseband Cartesian beamforming with an improved programmable gain amplifier design. A 28/37 GHz band, two-layer, eight-element, four-stream (with two FC-tiles) hybrid MIMO transceiver prototype is designed in 65-nm CMOS to demonstrate the above features. The prototype achieves accurate beam/null-steering capability, excellent area/power efficiency, and state-of-the-art TX/RX mode performance in two simultaneous bands while demonstrating multi-antenna (up to eight) multi-stream (up to four) over-the-air spatial multiplexing operation using proposed energy-efficient two-layer hybrid beamforming scheme. 

Tire wear is a leading cause of automobile accidents globally. Beyond safety, tire wear affects performance and is an important metric that decides tire replacement, one of the biggest maintenance expense of the global trucking industry. We believe that it is important to measure and monitor tire wear in all automobiles. The current approach to measure tire wear is manual and extremely tedious. Embedding sensor electronics in tires to measure tire wear is challenging, given the inhospitable temperature, pressure, and dynamics of the tire. Further, off-tire sensors placed in the well such as laser range-finders are vulnerable to road debris that may settle in tire grooves. This paper presents Osprey, the first on-automobile, mmWave sensing system that can measure accurate tire wear continuously and is robust to road debris. Osprey’s key innovation is to leverage existing, high-volume, automobile mmWave radar, place it in the tire well of automobiles, and observe reflections of the radar’s signal from the tire surface and grooves to measure tire wear, even in the presence of debris. We achieve this through a super-resolution Inverse Synthetic Aperture Radar algorithm that exploits the natural rotation of the tire and improves range resolution to sub-mm. We show how our system can eliminate debris by attaching specialized metallic structures in the grooves that behave as spatial codes and offer a unique signature, when coupled with the rotation of the tire. In addition to tire wear sensing, we demonstrate the ability to detect and locate unsafe, metallic foreign objects such as nails lodged in the tire. We evaluate Osprey on commercial tires mounted on a mechanical, tire-rotation rig and a passenger car.We test Osprey at different speeds, in the presence of different types of debris, different levels of debris, on different terrains, and different levels of automobile vibration. We achieve a median absolute tire wear error of 0.68 mm across all our experiments. Osprey also locates foreign objects lodged in the tire with an error of 1.7 cm and detects metallic foreign objects with an accuracy of 92%. 

Digital phase-locked loops (DPLL) are finding new applications in highly demanding contexts such as frequency synthesis for millimeter-wave (mm-wave) communications and clock generation for ultra-high-speed wireline transceivers. In a typical DPLL, however, a time-to-digital converter (TDC) with fine time resolution, high linearity and high dynamic range is required to meet stringent noise and spur performance requirements, which negatively impacts the power consumption in a DPLL. A bang-bang phase-detector (BBPD) outperforms a multi-bit TDC in terms of its’ jitter-power tradeoff, but its’ highly non-linear phase detection characteristic limits the locking speed of the loop. This research explores the design and of a 60 GHz digital sub-sampling phase-locked loop that uses a BBPD loop for frequency tracking and a coarse TDC loop for fast frequency acquisition. A prototype of the DPLL is designed in a 28-nm CMOS technology with supporting evidence through extensive simulations.