Analog-to-Digital Converter (ADC) is the bridge connecting the analog world and the digital world. With the development of circuit techniques and manufacturing process, its performance indicators have made great progress. The classification and performance characteristics of ADCs are firstly introduced, and then a description is made on the basic principles and technological development of ADCs of different structures including directions: high-speed and high-resolution ADCs. For high-speed ADCs, this article focuses on performance optimization techniques for SAR ADCs and Pipelined-SAR ADCs, such as CDAC controlling methods and comparator design, non-binary redundancy, loop-unrolled, and inter-stage redundancy. For high-resolution ADCs, the various types of Delta-Sigma ADCs and their advantages are analyzed, and the technical characteristics of Zoom ADCs and NS-SAR ADCs introduced. Some new types of hybrid-architecture ADCs are also summarized, describing their composition and research progress.

In pipelined-successive approximation register analog-to-digital converter (pipelined-SAR ADC), it is necessary to use large-open-loop gain operational amplifiers to improve the gain accuracy of closed-loop residual amplifications. The proposed ring amplifier uses a gain-enhanced output stage to improve the open-loop gain and stability, achieving an open-loop gain of over 90 dB and significantly reducing the residue gain errors without any calibration techniques, meeting the accuracy requirement of a 16 bit ADC. The ADC is implemented in the 65 nm CMOS process with an active area of 0.256 mm². At a sampling rate of 25 MS/s and with Nyquist-rate input, the proposed ADC achieves simulated signal-to-noise distortion ratio (SNDR) and spurious free dynamic range (SFDR) of 77.8 dB and 96.8 dB, respectively, with a power consumption of 2.8 mW. The proposed ADC achieves Walden and Schreier figure-of-merit (FoM) of 18.0 fJ/conversion-step and 174.3 dB, respectively.

Based on a 40 nm CMOS process, a high-gain low noise amplifier (LNA) chip was designed. The topology architecture of the chip adopted transformer input matching technology and positive feedback same-phase amplification technology to improve the input matching degree and gain. By introducing an active biasing network and a transformer matching network into the input stage of the traditional common-source common-gate structure, the chip can not only work stably at room temperature, but also shows excellent performance within the temperature range of –40 ℃ to 125 ℃ in simulation. Therefore, this design can be used for transceiver receiving ports in millimeter wave frequency bands under different temperature environments, and has certain temperature stability characteristics. The chip’s layout size is 0.383 mm×0.694 mm. The post-layout simulation results show that the LNA achieves a noise figure of less than 4.96 dB, a maximum gain of 18.11 dB, an input return loss of less than –16.08 dB, and an output return loss of less than –11.54 dB within the working frequency range of 24~27 GHz at room temperature. In addition, the LNA design has excellent performance indicators such as an input P_1dB of –20.36 dBm and a DC power dissipation of 12.8 mW.

A circularly polarized active integrated transmitter antenna based on a switched capacitor power amplifier (SCPA) is designed to meet the increasing demand for high performance in modern wireless communication systems. In the active integrated transmitter antenna, the antenna not only provides the optimal load impedance for the SCPA, but also compensates for the parasitic inductance introduced by the bonding wires, reducing circuit losses and achieving a compact and integrated designing result. Due to the optimal load impedance provided by the antenna, the SCPA die removes the output matching network, and the die is designed with polar and quadrature multiplexing to support more complex modulation while maintaining system compactness. The planar slot antenna introduces a section of elongated stub in the annular slot structure so that the base mode of the annular slot structure is decomposed into two orthogonal degenerate resonant modes, and both right-hand circular polarization and left-hand circular polarization are achieved in the same frequency band. The experimental results show that the designed SCPA operates at 1.92 GHz with a maximum output power of 10.25 dBm. The equivalent isotropic radiated power (EIRP) of the active integrated antenna at the operating frequency reaches 12.5 dBm.

Leakage power consumption is a key issue in integrated circuit applications, and body bias adjustment technology is one of the most commonly used power consumption adjustment technologies. The traditional body bias adjustment circuit has problems such as small bias voltage range and multiple power supply voltages, which not only increases the cost of the entire system, but also limits the optimization effect of body bias adjustment technology. Based on the 22 nm FDSOI (Fully Depleted Silicon on Insulator) RVT (Regular Voltage Threshold) process, a wide-range body bias adjustment circuit suitable for 22 nm FDSOI RVT digital integrated circuits is proposed. This circuit has a programmable (0 V, ±2 V) wide voltage output range, can achieve 50 mV bias voltage resolution, and does not require additional power input. The test circuit was implemented based on the 22 nm FDSOI process. The simulation results show that the body bias adjustment circuit proposed in this design can reduce the standby leakage of the test circuit by 34% to 92% and has a wide performance tracking range.

Physical verification is a critical concern in chip manufacturing, ensuring chip yield. Detecting potential hotspots in the chip layout before actual manufacturing is a critical step, which ensures manufacturing feasibility and enhances production efficiency. Traditional hotspot detection techniques suffer from long detection cycles and high computational resource consumption, resulting in increased time costs throughout the production cycle and limited detection of hotspot patterns. Based on active learning techniques and visual state space models, this paper proposes a new hotspot detection model. A memory-based sampling strategy is employed for query evaluation to mitigate the impact of the imbalance between hotspot and non-hotspot data on the model. Furthermore, the resolution constraints of the CNN structure and the secondary complexity of the ViT network architecture are optimized, leading to linear complexity for the hotspot detector. Testing results on the ICCAD-2012 competition dataset show that the proposed hotspot detector significantly reduces the false positive rate, achieving a rate of only 1.47%, while the recall reaches an impressive 98.89%.

Solid State Disks based on 3D triple cell NAND flash has becoming a dominant storage medium in mass storage systems due to their high storage density and low cost per bit. With the rapid development of technologies, 3D NAND flash chips are becoming less reliable with high storage densities. Reduced reliability and overly conservative manufacturers' formulation of lifetime nominal values result in flash chips being prematurely phased out before reaching their actual lifespan with unnecessary waste. Lifetime prediction of flash chips through machine learning-based prediction models can optimize storage strategies to effectively extend lifetime and reduce losses. However, due to the differences in production processes, the error characteristics of flash memory chips are somewhat different from each other, which affects the accuracy of the life prediction of flash memory chips. In this paper, we experimentally find that the bit error rate of data retention errors can be used to characterize the number of program/erase cycles times, and propose to stimulate the interference between word-lines by writing specific contents to adjacent word-lines, which can effectively reduce the elapsed time and improve the accuracy of the life time prediction. Experimental results show that the elapsed time can be shorten by about 90.9%, and the prediction accuracy can be improved by 33.3 percentage points.

Accurate segmentation of the left ventricular endocardium from cardiac magnetic resonance imaging to obtain the left ventricular region is an important step in the analysis of cardiac function. It is noted that reinforcement learning is prone to localization deviation in left ventricular endocardium segmentation by locating the left ventricular endocardial edges, leading to a performance decrease in the segmentation. To address this, this paper proposes a direction-constrained reinforcement learning method for left ventricular endocardium segmentation, which divides the segmentation task into two stages. In the first stage, the proposed method extracts global edge features of the endocardium, and in the second stage, reinforcement learning is used to iteratively locate the endocardial edge points to obtain the edge, obtaining the segmented left ventricular endocardium. The proposed method constrains the direction of agent positioning, which can reduce the localization deviation and overlap, such that the segmentation accuracy can be improved. Finally, the experimental results on two public datasets, including the Automated Cardiac Diagnosis Challenge (ACDC) and Sunnybrook Cardiac MR Left Ventricle Segmentation Challenge (Sunnybrook) , show that the proposed method has higher accuracy than the compared methods. Specifically, the F1-score of the proposed method are 0.9482 and 0.9387, and the Average perpendicular distance (APD) are 3.5863 and 4.9447, which can effectively segment the left ventricular endocardium.

The existing expensive constrained multi-objective optimization algorithms based on surrogate models face two main issues. Firstly, the use of regression models to fit constraints introduces errors that affect the algorithm's search direction. Secondly, when the objective function is non-fittable, the performance of the regression model for fitting is poor. To address these issues, a collaborative expensive constrained multi-objective evolutionary optimization algorithm is proposed, which combines a classification model with a regression model. This method employs the classification model to roughly divide the search space, guiding the algorithm to quickly enter the feasible region and reducing the impact of constraint fitting errors. The regression model is then used to optimize the objective function within the feasible region. The collaboration of the two models allows the classification model to provide a general search direction while the regression model performs detailed modeling. This fusion of models not only considers the impact of constraint errors on the algorithm but also comprehensively addresses the fittability of the objective function, enabling a more comprehensive and accurate depiction of the characteristics of complex problems. As a result, it enhances the efficiency and effectiveness of the algorithm, providing an effective approach for further improving expensive constrained multi-objective optimization based on surrogate models.

Active domain adaptation (ADA) aims to train an effective model under the context of domain adaptation with as few queried instances as possible. However, existing algorithms tend to select instances that are either uninformative, redundant, or outliers due to domain shift. To address this issue, a novel approach called neighbor environment perception sample selection (NEPS) for active domain adaptation is proposed. NEPS explores the target sample informativeness in a neighbor environment-aware manner to select instances that are potentially most valuable under domain shift. Specifically, from informativeness perspective, NEPS aims to acquire knowledge not only from individual data points but also from their neighboring samples. This is achieved by measuring neighbor awareness informativeness score (NAIS) , which ensures the selected samples have both high individual informativeness score and environment informativeness score. Additionally, NEPS ranks and selects samples based on their similarity scores with labeled samples to ensure diversity among the chosen instances. Furthermore, NEPS makes effective use of all labeled samples as well as a large amount of unlabeled data from the target domain to enhance the model's performance. Experimental results demonstrate that NEPS exhibits strong sample selection capability and outperforms existing models in terms of classification performance on various benchmark datasets.

A factor-level feature and attribute preference joint learning session-based recommendation model is proposed to address the problem of low recommendation accuracy caused by short sequences, sparse data, and difficulty in generalization. The model first learns user global-level session item embeddings by constructing a global level session item dependency perception graph. Then, using the method of disentanglement representation learning, the items in the conversation are decomposed into multiple relatively independent factor-level features to learn user factor-level interest preferences. Then, using contextualized self-attention graph neural networks, user preferences for session item attributes are captured. Finally, factor-level interest preferences and the project attribute preferences are jointly learned to obtain the user's final interest preferences, which in turn completes the session recommendation. Multiple experiments on two publicly available datasets, Diginetica and Cosmetics, have shown that our model outperforms the baseline model in comparison, verifying the recommendation performance and design rationality of our model.

The radioactive tracers used in Single-Photon Emission computerized Tomography (SPECT) scans can cause radiation exposure to the human body. Therefore, low-dose SPECT has attracted widespread attention in nuclear medicine imaging. Under low-dose imaging conditions, projection data is heavily contaminated by severe noise. There has been a significant amount of studies that explore fully supervised deep learning reconstruction methods to suppress image noise. The quality of images obtained by fully supervised methods depend on the quantity and quality of the labels. However, it is challenging to obtain the normal-dose images with supervised labels in clinical practice. To overcome the challenge, we propose a pre-trained mean teacher method with GAN loss to achieve low-dose SPECT reconstruction. The proposed method introduces a Swin-Conv-Unet-based pre-trained model into the mean teacher model to enhance the reliability of unlabeled training data. The teacher model supervises the student model through consistency regularization; the pre-trained model is trained with a small amount of labeled data and enhances the supervision reliability through GAN loss. Numerical experiments validate the performance of the proposed method in noise suppression and feature preservation. When compared with the mean teacher method, the SSIM of the reconstructed images is increased by 2%, the RMSE is reduced by 9%, and the PSNR is increased by 0.77 dB. The dataset is generated by the SIMIND simulation software using XCAT digital phantoms.

The existing single-photon LiDAR, such as ICESat-2/ATLAS, are with high observation sensitivity and significant background noise, which usually require effective filtering methods for removing noise. This paper proposes an improved adaptive photon point cloud signal extraction method based on the principle of point cloud density denoising. The proposed method first adaptively determines grid width and height according to the point cloud distribution characteristics to partition point cloud data into grids. Then, it conducts grid continuity tests as units for point cloud filtering. Finally, it employs K-means clustering and cloth simulation filter algorithms to accurately extract reliable vertical control points. Experimental results on ATL03 photon point clouds show that the proposed method achieves promising effectiveness for point cloud data with different terrain variations by achieving approximately 99.0%, 99.9%, and 99.5% in terms of signal point recall rate (Recall) , precision rate (Precision) , and F-measure, respectively. From the registration experimental results with reference point clouds, the elevation errors of photon elevation points are 0.960 m, 0.957 m, and 0.872 m, respectively, outperforming the official ATL08 control group provided by the authorities.

The mobile ad-hoc network (MANET) under port yards relies heavily on the channel information between small communication nodes. However, it remains a challenging problem to deploy traditional large channel measurement equipment on small nodes. In addition, the wireless channel between nodes exhibits strong spatial and temporal variation characteristics. Consequently, conducting channel measurement and modeling in this scenario presents dual challenges of the equipment portability and channel modeling accuracy. To address the portability of measurement equipment, this paper investigates the highly portable and high-precision channel measurement method based on the universal software radio peripheral (USRP) and analyzes the influence of metal structures on the large-scale and small-scale parameters. To address the problem of low accuracy of the ray-tracing channel model due to the high complexity of the scenario, we propose an optimal parameter search method based on the channel measurement data by identifying the key simulation parameters of the ray-tracing channel model. By searching for the optimal values of the key simulation parameters, the channel model is iteratively calibrated. The experimental results show that our proposed model is suitable for fitting the path loss with a double-slope model in the MANET scenario, where the multipath effect caused by the metal structure results in a multi-peak distribution of the average power delay profile. The results also show that the proposed optimal parameter search method significantly improves the agreement between the simulation values of path loss and RMS delay spread with the measured values, such that the accuracy and universality of the channel model can be improved.

In recent years, the Multi-layer convolutional sparse coding (ML-CSC) model has been regarded as a theoretical explanation for convolutional neural networks (CNN). While the ML-CSC model performs well on datasets with high feature contrast, its performance is not satisfactory on datasets with low feature contrast. To address this issue, this paper introduces a multi-scale technique to design a multi-scale multi-layer convolutional sparse coding network (MSMCSCNet), which not only achieves better image classification results in scenarios with weak feature contrast, but also provides the model with a solid theoretical foundation and higher interpretability. Experimental results demonstrate that, without increasing the parameter count, MSMCSCNet achieves accuracy improvements of 5.75, 9.75, and 9.8 percentage points on the Cifar10, Cifar100 datasets, and the Imagenet32 subset, respectively, compared to existing ML-CSC models. Furthermore, ablation experiments further validate the effectiveness of the model's multi-scale design and feature selection mechanism.

A polarization-insensitive plasmonic cavity photodetector is presented using a two-dimensional nanoscale gold cylindrical array as a plasma electrode. The two-dimensional symmetry of the cylinder makes the plasma electrode insensitive to the polarization angle of the incident light. By forming a metal-semiconductor-metal cavity between a subwavelength metal cylinder and a metal reflector, effective enhancement of light absorption in the ultrathin region can be achieved. The optical and electrical responses of the photodetector were calculated using the time-domain finite difference method and the finite element method, and the effects of the geometric parameters on the performance of the gold nanocylinders were analyzed. The results show that after optimizing the parameters, the overall light absorption rate of the device is 94.7%, and the light absorption rate in GaAs semiconductor can reach 81.1%, and the responsivity of the device can reach 0.37 A/W at 5 V bias voltage and 10 mW incident light power. The device exhibits extremely low polarization-dependent characteristics, where changes in polarization angle do not cause changes in absorption in the device semiconductor, and the absorption response can be effectively preserved without changes in the position of the absorption peak when the angle of incidence changes.