Search for: All records

Abstract This study discusses the feasibility of Ferroelectric Capacitors (FeCaps) and Ferroelectric Field-Effect Transistors (FeFETs) as In-Memory Computing (IMC) elements to accelerate machine learning (ML) workloads. We conducted an exploration of device fabrication and proposed system-algorithm co-design to boost performance. A novel FeCap device, incorporating an interfacial layer (IL) and$$\text {Hf}_{0.5}\text {Zr}_{0.5}\text {O}_2$$ ${\text{Hf}}_{0.5}{\text{Zr}}_{0.5}{\text{O}}_2$(HZO), ensures a reduction in operating voltage and enhances HZO scaling while being compatible with CMOS circuits. The IL also enriches ferroelectricity and retention properties. When integrated into crossbar arrays, FeCaps and FeFETs demonstrate their effectiveness as IMC components, eliminating sneak paths and enabling selector-less operation, leading to notable improvements in energy efficiency and area utilization. However, it is worth noting that limited capacitance ratios in FeCaps introduced errors in multiply-and-accumulate (MAC) computations. The proposed co-design approach helps in mitigating these errors and achieves high accuracy in classifying the CIFAR-10 dataset, elevating it from a baseline of 10% to 81.7%. FeFETs in crossbars, with a higher on-off ratio, outperform FeCaps, and our proposed charge-based sensing scheme achieved at least an order of magnitude reduction in power consumption, compared to prevalent current-based methods. 

As solar photovoltaic (PV) has emerged as a dominant player in the energy market, there has been an exponential surge in solar deployment and investment within this sector. With the rapid growth of solar energy adoption, accurate and efficient detection of PV panels has become crucial for effective solar energy mapping and planning. This paper presents the application of the Mask2Former model for segmenting PV panels from a diverse, multi-resolution dataset of satellite and aerial imagery. Our primary objective is to harness Mask2Former’s deep learning capabilities to achieve precise segmentation of PV panels in real-world scenarios. We fine-tune the pre-existing Mask2Former model on a carefully curated multi-resolution dataset and a crowdsourced dataset of satellite and aerial images, showcasing its superiority over other deep learning models like U-Net and DeepLabv3+. Most notably, Mask2Former establishes a new state-of-the-art in semantic segmentation by achieving over 95% IoU scores. Our research contributes significantly to the advancement solar energy mapping and sets a benchmark for future studies in this field. 

With the expansion of sensor nodes to newer avenues of technologies, such as the Internet of things (IoT), internet of bodies (IoB), augmented reality (AR), and mixed reality, the demand to support high-speed operations, such as audio and video, with a minimal increase in power consumption is gaining much traction. In this work, we focus on these nodes operating in audio-based AR (AAR) and explore the opportunity of supporting audio at a low power budget. For sensor nodes, communicating one bit of data usually consumes significantly higher power than the power associated with sensing and processing/computing one data bit. Compressing the number of communication bits at the expense of a few computation cycles considerably reduces the overall power consumption of the nodes. Audio codecs such as AAC and LDAC that currently perform compression and decompression of audio streams burn significant power and create a floor to the minimum power possible in these applications. Compressive sensing (CS), a powerful mathematical tool for compression, is often used in physiological signal sensing, such as EEG and ECG, and it can offer a promising low-power alternative to audio codecs. We introduce a new paradigm of using the CS-based approach to realize audio compression that can function as a new independent technique or augment the existing codecs for a higher level of compression. This work, CS-Audio, fabricated in TSMC 65-nm CMOS technology, presents the first CS-based compression, equipped with an ON-chip DWT sparsifier for non-sparse audio signals. The CS design, realized in a pipelined architecture, achieves high data rates and enables a wake-up implementation to bypass computation for insignificant input samples, reducing the power consumption of the hardware. The measurement results demonstrate a 3X-15X reduction in transmitted audio data without a perceivable degradation of audio quality, as indicated by the perceptual evaluation of audio quality mean opinion score (PEAQ MOS) >1.5. The hardware consumes 238 μW power at 0.65 V and 15 Mbps, which is (~20X-40X) lower than audio codecs. 

Recent advances in audio-visual augmented reality (AR) and virtual reality (VR) demands 1) high speed (>10Mbps) data transfer among wearable devices around the human body with 2) low transceiver (TRX) power consumption for longer lifetime, especially as communication energy/b is often orders of magnitude higher than computation energy/switching. While WiFi can transmit compressed video (HD 30fps, compressed @6-12Mbps), it consumes 50-to-400mW power. Bluetooth, on the other hand, is not designed for video transfer. New mm-Wave links can support the required bandwidth but do not support ultra-low-power (<1mW). In recent years, Human-Body Communication (HBC) [1]–[6] has emerged as a promising low-power alternative to traditional wireless communication. However, previous implementations of HBC transmitters (Tx) suffer from a large plate-to-plate capacitance (C p , between signal electrode and local ground of the transmitter) which results in a power consumption of aC p V2f (Fig. 16.6.1) in voltage-mode (VM) HBC. The recently proposed Resonant HBC [6] tries to overcome this problem by resonating C p with a parallel inductor (L). However, the operating frequency is usually < a few 10's of MHz for low-power Electro-Quasistatic (EQS) operation, resulting in a large/bulky inductor. Moreover, the resonant LC p circuit has a large settling time (≈5Q 2 RC P , where R is the effective series resistance of the inductor) for EQS frequencies which will limit the maximum symbol rate to <1MSps for a 21MHz carrier (the IEEE 802.15.6 standard for HBC), making resonant HBC infeasible for> 10Mb/s applications. 

The emergence of Audio-based Augmented Reality has been calling for increasing data-rates for audio signals, with significant reduction in power to enable extremely energy-constrained sensor nodes. Typically, the communication power dominates sensing and computing power in a node [1]. For highly energy constrained scenarios, compressive sensing (CS) have been demonstrated (Fig. 1), where samples are first compressed at the sensor to contain the same information in a smaller number of samples, before transmitting to a receiver, where the signal is reconstructed. Previous CS works [2]-[5] have focused entirely on “sparse” physiological signals, operating in low speed regime. This work illustrates the first CS design, enabled with a discrete wavelet transform (DWT) sparsifier for catering to non-sparse signals such as high definition audio. Audio recording and playback are quite sensitive to quality, thereby requiring audio codecs, such as. aac, for efficient compression and decompression of audio streams, which usually consume power in the order of mW [6]. Audio inferencing operated in intelligent assistants are more tolerant to input quality, functioning effectively when the Perceptual Evaluation of Audio Quality Mean Opinion Score (PAEQ MOS) [7], an ITU-R standard objective metric for characterizing perceived audio quality, exceeds 1.5. CS presents an opportunity to achieve >10X reduction in transmitted audio data with orders of magnitude lower power, as compared to codecs. The design is implemented in 65 nm CMOS and consumes 238 uW power at 0.65 V and 15 Mbps. 

To solve the challenge of powering and communication in a brain implant with low end-end energy loss, we present Bi-Phasic Quasi-static Brain Communication (BP-QBC), achieving < 60dB worst-case channel loss, and ~41X lower power w.r.t. traditional Galvanic body channel communication (G-BCC) at a carrier frequency of 1MHz (~6X lower power than G-BCC at 10MHz) by blocking DC current paths through the brain tissue. An additional 16X improvement in net energy-efficiency (pJ/b) is achieved through compressive sensing (CS), allowing a scalable (6kbps-10Mbps) duty-cycled uplink (UL) from the implant to an external wearable, while reducing the active power consumption to 0.52μW at 10Mbps, i.e. within the range of harvested body-coupled power in the downlink (DL), with externally applied electric currents < 1/5th of ICNIRP safety limits. BP-QBC eliminates the need for sub-cranial interrogators, utilizing quasi-static electrical signals for end-to-end BCC, avoiding transduction losses. 

Applications like Connected Healthcare through physiological signal monitoring and Secure Authentication using wearable keys can benefit greatly from battery-less operation. Low power communication along with energy harvesting is critical to sustain such perpetual battery-less operation. Previous studies have used techniques such as Tribo-Electric, Piezo-Electric, RF energy harvesting for Body Area Network devices, but they are restricted to on-body node placements. Human body channel is known to be a promising alternative to wireless radio wave communication for low power operation [1-4], through Human Body Communication, as well as very recently as a medium for power transfer through body coupled power transfer [5]. However, channel length (L) dependency of the received power makes it inefficient for L>40cm. There have also been a few studies for low power communication through the human body, but none of them could provide sustainable battery-less operation. In this paper, we utilize Resonant Electro Quasi-Static Human Body Communication (Res-EQS HBC) with Maximum Resonance Power Tracking (MRPT) to enable channel length independent whole-body communication and powering (Fig.1). We design the first system to simultaneously transfer Power and Data between a HUB device and a wearable through the human body to enable battery-less operation. Measurement results show 240uW, 28uW and 5uW power transfer through the body in a MachineMachine (large devices with strong ground connection) Tabletop (small devices kept on a table, as in [5]) and Wearable-Wearable (small form factor battery operated devices) scenario respectively, independent of body channel length, while enabling communication with a power consumption of only 2.19uW. This enables >25x more power transfer with >100x more efficiency compared to [5] for Tabletop and 100cm Body distance by utilizing the benefits of EQS. The MRPT loop automatically tracks device and posture dependent resonance point changes to maximize power transfer in all cases.