This paper presents a low power 60 GHz transceiver that includes RF, LO, PLL and BB signal paths integrated into single chip. The has been fabricated in standard 90 nm CMOS process specially designed ESD protection on all mm-wave pads. With 1.2 V supply the chip consumes 170 mW while transmitting 10 dBm 138 receiving. Data transmission up to 5 Gb/s each of I Q channels measured, as data reception over 1 m wireless link at 4 QPSK with less than <sup...

This paper describes a low power and element-scalable 60 GHz 4-element phased array transceiver implemented in standard 65 nm CMOS process. Using 1.2 V supply, the consumes <;34 mW/element including LO synthesis distribution. Energy area efficiency are achieved by utilizing baseband phase shifting architecture, holistic impedance optimization, lumped-element based design. Each receiver (RX) element provides 24 dB of gain with an average noise figure (NF) 6.8 while total saturated output...

A 65-nm digitally modulated polar TX for WLAN 802.11g is fully integrated along with baseband digital filtering. The employs dynamic impedance modulation to improve efficiency at back-off powers. High-bandwidth phase achieved efficiently an open-loop architecture. Operating from 1.2-V/1-V supplies, the delivers 16.8 dBm average power -28-dB EVM 24.5% drain efficiency, 22% PAE, 19.3% system and 247-mW consumption including entire on-chip chain; noise floor -125 dBm/Hz 200-MHz offset without external

Uncertainty quantification is a fundamental yet unsolved problem for deep learning. The Bayesian framework provides principled way of uncertainty estimation but often not scalable to modern neural nets (DNNs) that have large number parameters. Non-Bayesian methods are simple implement conflate different sources uncertainties and require huge computing resources. We propose new method quantifying DNNs from dynamical system perspective. core our view DNN transformations as state evolution...

We introduce BAG, the Berkeley Analog Generator, an integrated framework for development of generators and Mixed Signal (AMS) circuits. Such are parameterized design procedures that produce sized schematics correct layouts optimized to meet a set input specifications. BAG extends previous work by implementing interfaces integrate all steps flow into single environment providing helper classes - both at schematic layout level aid designer in developing truly technology-independent circuit...

Fine-tuned pre-trained language models can suffer from severe miscalibration for both in-distribution and out-of-distribution (OOD) data due to over-parameterization. To mitigate this issue, we propose a regularized fine-tuning method. Our method introduces two types of regularization better calibration: (1) On-manifold regularization, which generates pseudo on-manifold samples through interpolation within the manifold. Augmented training with these imposes smoothness improve calibration....

Yue Yu, Lingkai Kong, Jieyu Zhang, Rongzhi Chao Zhang. Proceedings of the 2022 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2022.

This paper presents key design techniques and challenges in implementing one of the first integrated, energy-efficient 60 GHz transceivers including baseband circuitry. The 90 nm CMOS direct-conversion operates from a 1.2 V supply has been optimized for 5-to-10 Gb/s QPSK modulation centered at GHz. To achieve power consumption 170 mW transmit mode 138 receive mode, this leverages co-integration optimization mm-wave mixed-signal circuits.

This paper presents a low-power mixed-signal adaptive 60 GHz baseband in 65 nm CMOS. The design integrates variable gain amplifiers, analog phase rotator, 40-coefficient I/Q decision feedback equalizers (DFEs), clock generation and data recovery circuits, adaptation hardware. achieves 10 Gb/s operation with BER <; <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-12</sup> while consuming 53 mW (adaptation on)/45 off), of which the core signal...

In order to support higher throughputs, the power consumption of 2-to-5GHz Wi-Fi transmitters (TXs) has been continuously rising, and hence become increasingly problematic for mobile devices. To extend battery life, TX must be efficient not only at peak but also backoff, due use high Peak-to-Average-Power-Ratio (PAPR) OFDM modulation. Many recent works have aimed enhance PA efficiency back-off powers [1-4], relatively few integrated these techniques into a complete system. For example,...

Nickel-catalyzed asymmetric hydrogenation of α-benzamidocinnamic acid esters was studied to establish an effective approach for synthesizing methylsulfonyl-substituted phenylalanine. Using the nickel acetate and simple R-BINAP as catalyst, under the...

A novel tandem reaction starting from alkynones and cyclic ketones has been developed, which can efficiently synthesis pyranoindolones fused with a medium-sized ring. This process involves base induced C−C σ-bond cleavage of ketone subsequent I2 mediated 1,2-acyl migration/cyclization reaction. Transition metal-free easily accessible materials make this 1, 2-acyl methodology attractive.

The 60GHz band has received significant attention as an enabler for multi-Gb/s wireless communication. Practical mm-Wave systems will require relatively large phased arrays in order to robustly overcome path-loss and fading issues. Despite progress, CMOS implementations of have so far been area power hungry. This paper therefore presents a 4-element 65nm phased-array transceiver consuming <;34mW/element (including LO synthesis distribution) that utilizes baseband (BB) phase shifting holistic...

The 60GHz band has gained great interest as an enabler for multi-Gb/s wireless links. Recent efforts [1-4] have focused on reducing transceiver power to drive adoption of in mobile devices. To further accelerate this adoption, the cost current mm-Wave solutions should also be reduced. Especially short range designs (<;1m), overall may dominated by packaging and testing. This paper therefore presents a low-power CMOS 4-element phased-array QPSK with integrated slot-loop antennas. Utilizing...

This paper presents simulation of thermal stratification heat storage water tank with an inside cylinder openings by three-dimensional Computational Fluid Dynamics (CFD) methods. Specifically, this focuses both on investigating the influence inner cylinder's design and operating parameters flow characteristic, overall performance hot during charging operation process installed in energy systems. Validation CFD results experimental data found literature has a good consistency. In order to...

Clusters play a critical role in the hydrodynamics of gas–solid flow. In this work, probe method, DBSCAN-based and Voronoï-based method are used to identify clusters suspensions, their structural dynamic properties then quantified. The obtained by including cluster fraction, solid concentration inside cluster, size, comparable those using other two particle-based methods, but methods could offer more insights into fractal dimension, shape, topology clusters. However, boundary particles...

Large language models (LLMs) have recently demonstrated the potential in acting as autonomous agents for sequential decision-making tasks. However, most existing methods either take actions greedily without planning or rely on static plans that are not adaptable to environmental feedback. Consequently, performance of LLM degenerates with problem complexity and plan horizons increase. We propose a closed-loop approach, AdaPlanner, which allows agent refine its self-generated adaptively...

We introduce BAG, the Berkeley Analog Generator, an integrated framework for development of generators and Mixed Signal (AMS) circuits. Such are parameterized design procedures that produce sized schematics correct layouts optimized to meet a set input specifications. BAG extends previous work by implementing interfaces integrate all steps flow into single environment providing helper classes -- both at schematic layout level aid designer in developing truly technology-independent circuit...

Kinetic theory is a common choice for closing the solid phase stress in continuum dilute and moderate dense gas–solid flows. In this article, methods are proposed postprocessing data obtained from discrete particle simulations, results then used to critically assess fundamental assumptions of kinetic theory. It shown that (i) predictions are, respectively, valid accurate homogeneous granular flows, as expected. Those prove effective; (ii) case nonequilibrium heterogeneous nearly all get...

Learning from noisy labels is a challenge that arises in many real-world applications where training data can contain incorrect or corrupted labels. When fine-tuning language models with labels, easily overfit the label noise, leading to decreased performance. Most existing methods for learning use static input features denoising, but these are limited by information they provide on true distributions and result biased predictions. In this work, we propose Dynamics-Enhanced Generative Model...