Medical industry increasingly using convolutional neural networks (CNNs) for image processing. Nowadays, computing facilities based on Von Neumann architecture aredevoted to accelarate CNNs, yet rapidly hitting a bottlenneck in performance and energy efficiency. The computing-in-memory (CIM) architecture based on random-access memory (ReRAM) emerged as a method to overcome the issue. This work proposes a charge-domain one-transistor-one-resistor-one-capacitor (1T1R1C) CIM macro using energy-efficient charge calculation and capacitive coupling for CNNs acceleration in medical semantic segmentation. The multiplication-and-accumulation (MAC) is realized by charge distribution with a cell and capacitive coupling across different cells on a plate line. The configurable output resolution is achieved by on-chip ReRAM-based charge integral, which is energy efficient and flexible to change the output resolution. By evaluation in the 180nm technology, the proposed macro with a 64×64 array achieves a peak energy efficiency of 142.2 GOPS/W, ∼1.3X higher than previous work. The inference dice coefficient of UNet reaches 89.7%.

In order to reduce the performance deterioration of the control system of surface-mounted permanent magnet synchronous motor (SPMSM) under the influence of external disturbances, parameter changes and other factors, a new active disturbance rejection sliding mode control strategy (ADR-SMC) based on active disturbance rejection control (ADRC) is proposed. Firstly, based on the speed-current compound loop mathematical model of SPMSM, a nonlinear active disturbance rejection controller (NLADRC) is established to observe internal and external disturbances. Secondly, based on the independent characteristics of each module of NLADRC, a sliding mode error feedback control law with fast convergence speed and strong robustness is designed to replace the NLSEF module by using the characteristics of sliding mode control (SMC) with low requirements and strong robustness for the system model. Then, the sliding mode error feedback control law is updated in real time according to the internal disturbance estimated by the extended state observer (ESO). The resulting ADR-SMC improves the dynamic response capability of the system and improves the robustness of the system to internal and external disturbances. The stability of ADR-SMC compound closed-loop SPMSM control system is proved by Lyapunov theory. Finally, the effectiveness and superiority of the proposed control strategy are verified by experimental results.

Aiming at the increasing demand for the load power of more electric aircraft, this paper proposes an energy storage system based on battery. Through the modeling of the battery and the analysis of the charging and discharging characteristics, combined with the various working conditions of the aircraft, a complete energy management strategy is designed to realize the coordinated operation of the energy storage system and the generator. Finally, a semi-physical simulation platform is built based on Plecs RT Box, and the feasibility of the proposed scheme is verified by experiments.

A spiking neural network of ergodic sequential logic neuron models is presented. It is shown that the presented network is capable of reproducing various biologically plausible spatio-temporal phenomena (e.g., basic synchronization and complicated chimera phenomenon) observed in the brain. Moreover, it is revealed that the presented network is able to operate with lower power compared to a standard digital-processor-based spiking neural network. It is then discussed that the presented network will be a useful building block in a brain prosthetic device.

A metasurface which can realize switching between transmission and reflection modes by changing temperature is proposed. The metasurface consists of a dielectric layer and two metal layers with an extremely low profile. At low temperature, it can convert the incident circularly polarized wave into an orthogonal polarization wave; while at high temperature, the incident circularly polarized wave will be completely reflected and keep the same polarization. By appropriately arranging the unit cell structure, a full space focusing lens is realized at 1.0THz. This work provides a new idea for designing full-space metasurfaces, which has great potential value in the fields of reconfigurable and dynamically controllable device design.

This letter presents a 6-bit passive vector-modulated phase shifter (VMPS) using nonlinear complementary voltage control, which is implemented in a 45-nm silicon-on-insulator (SOI) technology. The proposed passive VMPS uses a simplified, compact, low-loss transformer-based coupler to generate highly balanced orthogonal signals. The amplitude controller is implemented with only four transistors and is controlled by nonlinear complementary voltage, which greatly improves the available state rate and the accuracy of the phase shifter. The measured VMPS achieves a rms phase/gain error of 1.6°/0.35dB with only 0.55mm² chip area in 20-24GHz and no power consumption.

At present, the grounding protection of flexible grounding system mostly directly adopts the zero-sequence time-limit overcurrent protection of low-resistance grounding system, which has problems such as insufficient protection sensitivity when grounding via high-resistance. The changes in zero-sequence voltage (ZSV) and zero-sequence current (ZSC) upstream and downstream of the fault point before and after the small resistor is put into operation are analyzed. The fault information collected by the synchronous phasor measurement unit (PMU) are orthogonalized. Finally, the correlation coefficient method is utilized to compare the numerical values and polarity of projection amounts between adjacent PMUs. Based on this, a fault location method based on steady-state projection is proposed in this paper. Simulation and on-site test data show that the proposed method can accurately identify the fault section under different fault conditions.

In this paper, a complementary dithering technique is proposed which utilizes LMS digital background calibration method to correct the interstage gain errors in pipeline analog-to-digital converters (ADCs). It not only has a better scattering effect on spectral spurs but also eliminates the increment of residual amplitude caused by dither injection, which will greatly alleviate the design requirements of residual amplifier. Simultaneously, the comparator resolving time nature is utilized to construct calibration windows, which averts the use of duplicate comparators and its digital logic is simple. Behavioral simulation results of 12-bit, 1.25GS/s pipelined ADC manifest that the proposed calibration technique enhances SNDR and SFDR from 44.27dB and 49.43dB to 70.8dB and 115.3dB respectively.

With the development of wind power generation, improving the accuracy of fault diagnosis in permanent magnet synchronous motor (PMSM) is crucial for reducing maintenance costs. This paper integrates signal analysis and machine learning algorithms, proposing an approach that combines the improved empirical wavelet transformation (IEWT) with the CatBoost algorithm to diagnose open-phase faults in PMSM. IEWT effectively suppresses the modal aliasing phenomenon caused by noise and spectral leakage in traditional empirical wavelet transformation, and the decomposition results can demonstrate the time-frequency characteristics of the fault signal used for training the CatBoost classification model. Experimental results indicate that the proposed algorithm achieves high overall accuracy in diagnosing open-phase faults in PMSM, with balanced performance across each category.