A 150 GHz high output power frequency doubler is designed and realized based on six anodes in anti-series GaAs planar Schottky varactors. The doubler is utilized by balanced construction. Co-simulation of 3-D electromagnetic simulator with nonlinear harmonic balance tool is utilized to improve the precision of simulation. According to the design results, the 150 GHz high output power frequency doubler is fabricated and tested. The measured efficiency is above 7% in 146 GHz-158 GHz. The peak efficiency is 12% at 154 GHz, and the output power is 71 mW.

In order to calculate Terahertz(THz) wave atmospheric absorption loss accurately, THz propagation segmented path model should be established firstly, and continuous changing environmental atmospheric parameters can be obtained through the whole path. Each segment path’s atmospheric absorption loss is obtained by using International Telecommunications Union-Radio communications sector(ITU-R) standard, and the whole segments are accumulated to get total atmospheric absorption loss. It is difficult to obtain complete atmospheric parameters along the propagation path due to the lack of meteorological observation equipment and methods. Besides, atmospheric refraction effect appears due to the earth’s gravity, which makes curved THz propagation path. This paper discusses the calculation method of THz wave atmospheric absorption loss. Based on atmospheric observation re-analysis data, using ITU-R standard as bridge, a constructive relationship is built between meteorological data and atmospheric absorption loss. The proposed method can be used in the ground detection engineering of THz wave for aircraft, to make THz communication link’s atmospheric absorption loss estimation.

Aiming at the application of remote sensing satellite for multi-band and multi-polarization detection, a quasi-optical feed system which is the important device of millimeter-wave radiometer antenna on satellite is designed. This feed system can respectively receive the electromagnetic signals at 54?GHz and 183?GHz to complete the effective detection of earth atmospheric and humidity. Each frequency band has H/V polarization. The quasi-optical feed system is constructed and simulated. The simulated results can meet the requirements on antenna system. The far-field radiation parameters of prototype, like taper angle, cross-polarization, beam efficiency, are measured. The measured results demonstrate that the proposed feed-system has a good performance and its feasibility and rationality are verified.

The communication blackout when the space vehicle at ultra-high speed is reentering the atmosphere has always been a challenging task. Based on the Epstein distribution of electron density, the models of oblique incidence terahertz wave propagation are established for magnetized plasma inhomogeneous plasma sheath. The transmittance and attenuation of terahertz wave are numerically calculated. The effects of collision frequency, magnetic field intensity, incident angle, etc. on THz wave attenuation are analyzed. The results shows that with the increase of magnetic field intensity, the minimum transmittance and attenuation peak move to the direction of higher THz frequency. When the electromagnetic wave incident angle becomes larger, the transmittance decreases, and the attenuation increases. When the incident angle is less than or equal to 60o, the maximum attenuation at 0.22 THz of atmospheric window is 5.32 dB. This study provides a theoretical basis for solving the blackout problem.

Based on the Terahertz Time-Domain Spectrum(THz-TDS) technology, a multi-feature joint medicine inspection method is proposed to study three different medicines. First, the measurement data is acquired with a THz-TDS system. For consistency, medicines are dried, crushed into powder and then made into capsules. Then, different features like refractive index and material factor are obtained by using feature extraction method. Finally, medicines are classified and identified by a multi-feature joint detection method, in which three different machine learning methods, Back Propagation(BP) neural network, Support Vector Machine(SVM) and Learning Vectorization Quantization(LVQ), are adopted to improve the efficiency and accuracy. In the training process, all parameters are combined as the training sample in order to improve the characteristic ability. Experiment results show that the accuracy with machine learning are above 95%, and for SVM, the accuracy reaches 99%. The results confirm the application of the terahertz multi-feature joint method in medicine quality inspection and identification.

In view of the fact that the current full duplex systems cannot effectively deal with external interference, a joint algorithm based on self-interference suppression technology and adaptive beamforming technology, is proposed for eliminating self-interference and external interference simultaneously. Firstly, the system model of the full duplex arrays under this scenario is established. Then, according to time-domain cancellation method and Minimum Variance Distortionless Response(MVDR) method, the mathematical model as well as the performance analysis of the algorithm is derived. At last, the numerical simulation results indicate that the proposed algorithm can restrain both self-interference and external interference to the noise level.

10 gigabit communication technology has become another research hotspot in the field of satellite communications, and it is especially important to provide high-speed and reliable 10 gigabit data sources for on-board devices. The system adopts the newly launched K7 series Field Programmable Gate Array(FPGA) from Xilinx, and realizes the 10 gigabit protocol conversion function by using its GTX interface module and 10 gigabit MAC module. The state machine design method realizes the mutual conversion between the Ethernet protocol frame and the dedicated link protocol frame. The test results show that the system achieves a conversion bandwidth of 10 Gbit/s per channel, and the total bandwidth reaches 40 Gbit/s, which can meet the testing requirements of satellite equipment.

In recent years, anti-interference technology of Global Navigation Satellite System(GNSS) has gradually become a research hotspot. In anti-interference research, satellite navigation signal simulation on array antennas in the context of broadband interference plays a vital role. In order to improve the efficiency of anti-interference research, the simulation problem of digital Intermediate Frequency(IF) signal at the entrance of satellite navigation software receiver is studied by Matlab. Firstly, power control is performed according to the signal to interference and noise ratio. Then satellite navigation signals, interference signals and noise signals are modeled separately. And then the delay caused by the array antenna is calculated. Finally, the satellite signals of the array antenna in the background of broadband interference are simulated and the Graphical User Interface(GUI) is designed. After the simulation, the signal is input into the research system of the anti-interference algorithm. It implements the complex interference signal source simulation in anti-interference research, improves its efficiency and verifies its practicability and correctness.

In order to further improve the target detection and monitoring capability of airborne early warning radar, this paper first analyzes the factors affecting the detection range of single platform airborne early warning radar, introduces the characteristics and necessity of networked collaborative detection mode, and gives the definition of stable detection coverage and Doppler blind zone. Secondly, the cooperative detection modes among airborne early warning radar and other airborne early warning radar, shipborne radar, Unmanned Aerial Vehicle(UAV) radar and other different platform radars are analyzed and the simulations and evaluations are carried out. The performance parameters are compared with that of various cooperative detection modes. Finally, the development trend of networked cooperative detection of airborne early warning radar is prospected.

The Licklider Transmission Protocol(LTP) is an important protocol in deep space communication, but its Automatic Repeat-reQuest(ARQ) mechanism cannot meet the transmission delay requirements. In order to take advantage of the fountain code without feedback and meet the different reliability requirements of red data and green data in the LTP, first of all, the fountain code with characteristics of Unequal Error Protection(UEP) is introduced into the LTP to obtain the LTP with characteristics of UEP(UEP-LTP). Then an exponential window selection probability function is constructed by assigning different selection probabilities of windows to different degrees of the fountain code, and the function is optimized to obtain an Advanced UEP-LTP(AUEP-LTP) with better performance. The simulation results show that the bit error rate of red data and green data of AUEP-LTP can reach 10-10.473 2 and 10-3.687 8 respectively, which can meet the different reliability requirements of red data and green data, reduce the retransmission of LTP and improve the performance, and is more suitable for deep space communication.

In view of the existing problems in the current Orthogonal Frequency Division Multiplexing (OFDM) radar signal recognition method, this paper proposes an interpretable method for identification of OFDM radar signals. The method which is based on Tree-based Pipeline Optimization Tool(TPOT) and Local Interpretable Model-agnostic Explanations(LIME) is to identify OFDM radar signals. Firstly, according to the characteristics of OFDM radar signals, the complexity features and singular value entropy of time-frequency image matrix are extracted to form the feature vectors. Then through the TPOT，the best performing machine learning process is obtained. Finally，the interpretation result is interpreted by the interpreter, and the result of the recognition is given as a risk assessment; meanwhile，according to the interpretability of OFDM radar signals, those signals difficult to distinguish are determined. The experimental results show that the recognition rate of the OFDM radar signal with RSN=0 dB is 91%. The interpretability given by LIME can be utilized to determine the type of radar signal that is difficult to distinguish in the data set.

The transcoding between Distributed Video Coding(DVC) and traditional video provides an effective way to realize low-power video communication between mobile terminal devices. In this paper, DVC and HEVC(High Efficiency Video Coding) transcoding is taken as the research object, and the decoding end information of DVC is utilized to carry out complexity optimization research on the division process of coding unit with high complexity in HEVC. Firstly, texture complexity, motion vector and prediction residuals related to Coding Unit(CU) partition are extracted from DVC decoder. Then, based on naive Bayes principle, a CU fast partition model is established at the HEVC coding end. After the model is generated, current CU division can be quickly determined by inputting feature information, avoiding a cost calculation process of large number of Rate Distortion(RD). Experimental results show that the proposed scheme significantly reduces the HEVC coding time by 58.26% with a slight increase in the encoding bit rate, and hardly affects the quality of video.

A novel flexible implantable antenna is proposed. The antenna is a half cutting of a Coplanar Waveguide(CPW)-fed antenna with symmetric geometry. The half-cut antenna is optimized and analyzed in a single-layered body phantom by using HFSS. Simulation shows that there are frequency bandwidth shifts when the body phantom is filled with different tissues of muscle, small intestine, and stomach, respectively. The antenna bandwidth covers Industrial, Scientific, and Medical(ISM) band in the case of skin, muscle and small intestine. Simulation also shows that the antenna bandwidth has little change with the varying of substrate thickness. Measurements are done by arranging the antenna in meat mince and skin-mimicking gel. For the two cases, the measured |S11|<-10 dB bandwidth of the proposed antenna is 1.91-2.65 GHz and 2.19-2.79 GHz covering ISM band, respectively. The proposed antenna has a small size of 6 mm×13.5 mm×24 ?m. The proposed antenna is also investigated with a conformal structure adhering to a cylinder. The measured bandwidth of the conformal antenna is 2.26-3.11 GHz. Compared with published literatures, the antenna has advantages of low profile, light weight and wide bandwidth. The proposed antenna can be applied in wireless implantable medical devices, such as capsule endoscope, wireless cardiac pacemaker etc.

The Method of Moments(MoM) combined with Multi-Layer Fast Multipole Algorithm (MLFMA) and Adaptive Cross Approximation(ACA) is proposed to calculate the electromagnetic characteristics of the target. The computation of the scattering of electrically large composite targets is speeded up and the memory requirement is reduced. For the near-field region of self-acting target, MoM and MLFMA are combined to speed up the matrix vector multiplication, reduce the storage and decrease the complexity of calculation. The far-field impedance matrix has a low rank characteristic and can be compressed by ACA to speed-up the filling rate of the matrix. Matrix filling is stored sequentially according to the interaction among the blocks divided by tree structure. ACA algorithm is utilized to solve the problem of the compression of matrix between each block and block. The proposed hybrid algorithm based on ACA can compress the impedance matrix of the coupling between two targets, reduce the filling time and the required memory of the matrix. At the same time, the computation time of matrix vectors can be shortened in the iterative process, thus the total time of electromagnetic scattering calculation can be reduced greatly. Numerical simulation results show that the algorithm is more efficient than the traditional method, and the accuracy of electromagnetic scattering is consistent.

A gas-solid coupling model is designed, which uses microwave heating porous medium(silicon carbide) to heat gas indirectly, aiming for the application of gas-solid model coupling and gas-solid reaction under microwave assisted heating. Finite element simulation software is adopted to build the model. The accuracy and feasibility of the gas-solid coupling model are verified by coupling simulation of three physical fields: electromagnetic field, fluid heat transfer field, free and porous media. The results show that the microwave conversion efficiency is the highest when the porosity of silicon carbide is 0.5 and the radius of quartz tube is 20 mm, and the microwave energy efficiency is above 90%.

Compared to conventional heat conduction heating, microwave heating has the effect of internal and external heating and has been widely used in various fields such as industry and agriculture. Nevertheless, the problems of low heating efficiency and low heating uniformity in industrial applications are still widespread. A heating method based on dual-port non- synchronous rotating structure is proposed to reduce the mutual coupling between ports, and a microwave heating model of dual-port rotating structure is established based on COMSOL multiphysics. This paper analyzes the effect of the rotation speed between the two microwave source ports on the mutual coupling and heating uniformity between the ports, realizes the low mutual coupling and high uniformity heating between the ports, and greatly improves the efficiency of microwave heating.

In recent years, microwave heating has been widely used in various fields due to its high efficiency and internal heating. However, the uniformity of microwave heating limits the development of its large-scale industrial application. High uniformity and high efficiency are the focus of microwave heating research. Low efficiency will cause the waste of energy and non-uniform heating will lead to poor heating quality and even thermal runaway. In this paper, a method of frequency-selected heating with dual-port is proposed and the power is fed by two orthogonal polarized antennas in the cavity. Then, the heating efficiency of each frequency point and the influence of different frequencies on the heating effect are analyzed by using sweep frequency. Finally, frequency-selected heating is carried out on the frequency point with high efficiency, so as to improve the heating uniformity under the premise of maintaining high heating efficiency.

To solve the problem of parameter diagnosis of microwave excited plasma, the equivalent relative complex permittivity of low-temperature plasma excited by microwave is measured by the two-port transmission reflection method. Firstly, the sample-test structure is designed based on the standard BJ22 waveguide structure with the finite element method. The structure can ensure that the permittivity and port parameters of |S11|,φS11,|S21| keep good monotonicity in a large range of permittivity, avoiding multi-value problems. Then, artificial neural network is utilized to train the S parameters which are corresponding to each permittivity obtained by simulation, and the training results achieve sufficient accuracy. Finally, through experiments, the measured S parameters are inverted by neural network to obtain the relative complex permittivity of the substance to be measured. This method can be applied to measure the permittivity of materials like plasma which has negative real part of complex permittivity.

Inverse Synthetic Aperture Radar(ISAR) images are usually affected by various noises which degrade the quality of ISAR images and seriously affect the subsequent feature extraction and target recognition applications. Improving the image quality and reducing noise interference have become important steps in the application of target recognition. A method based on Constant False-Alarm Rate (CFAR) detection and density clustering is proposed to suppress speckle interference and transverse fringe interference in ISAR images. Compared with traditional methods, this method can effectively preserve the details of the target while ensuring the effect of interference suppression. The image area, length and Doppler spread are extracted as ISAR recognition feature vectors. Field experiments show that the proposed method effectively suppresses the interference components in the image, retains more image details, and effectively improves the stability of ISAR recognition features.

In order to take into account the transparency and anti-geometric transformation ability of watermarking algorithm, by introducing the optical encryption method, the robust watermarking algorithm based on phase truncated Fresnel transform and total variational decomposition is proposed. Logistic mapping is introduced to obtain a set of chaotic sequences by iteration for constructing a random mask. Based on Fresnel transform and random mask, the optical encryption of watermarking information is completed. And then two private keys are obtained by using phase truncation mechanism. The host image is processed by total variational decomposition to obtain the cartoon and texture parts. The Discrete Wavelet Transform(DWT) is introduced to decompose the encrypted watermarking and texture parts, and the corresponding sub-band information are obtained. These sub-band information are fused and texture information embedded in watermarking is obtained by reversible DWT method. The watermarking image is formed by combining the watermarking texture part with the original cartoon part. The watermarking extraction method is designed to detect the encrypted watermarking information from the watermarking image, and two private keys are adopted to decrypt the encrypted watermarking information. The test data show that this algorithm takes into account the ideal transparency and robustness of watermarking, and under various geometric transformations, the reconstructed watermarking distortion is small with corresponding correlation coefficient above 0.9.

In recent years, deep learning of convolutional networks has achieved great success in the fields of image processing and target detection. It has become a new research hotspot to apply Convolutional Neural Networks(CNN) to traditional video compression standards. An improved High Efficiency Video Coding(HEVC) compression algorithm integrated with convolutional neural network is proposed, which integrates down-sampling, HEVC codec, up-sampling and quality enhancement process. In order to extract the structural features of video frames efficiently, two convolutional neural networks are integrated in the proposed compression algorithm. Down Sampling CNN(DwSCNN) replaces bicubic down-sampling, which preserves the detailed information while reducing the resolution, obtaining a more compact low resolution video sequence. The low-resolution video sequence is further compressed by HEVC intra coding, and a quality-enhanced Post Processing CNN(PPCNN) is proposed to improve the degraded video sequence that is restored to the original resolution after decoding. The experimental results show that the proposed compression improvement algorithm can achieve better quality reconstruction than the standard HEVC in the low code rate segment, and can save time by 39.46% and bit rate by 11.04% when the PSNR value is close to the same. The video compression performance of the algorithm is superior to the HEVC standard algorithm and other related literature methods.

An image compression coding framework based on adaptive sub-sampling and super-resolution reconstruction is designed for the Joint Photographic Experts Group(JPEG) standard. At the encoder side, a variety of Different Sampling Modes(DSM) and Quantization Modes(QM) are designed for the original image to be encoded. Then, the rate distortion optimization algorithm selects the optimal downsampling and quantization modes from various modes. Finally, the image to be encoded will be sampled and compressed by the standard JPEG compression under the selected optimal mode. In the decoder side, the super-resolution reconstruction algorithm based on convolutional neural network is utilized to reconstruct the decoded sub-sampled image. In addition, the proposed framework is also effective and feasible under the JPEG2000 compression standard. The experimental results show that compared with the mainstream coding and decoding standards and advanced encoding and decoding methods, the framework can effectively improve the rate distortion performance and obtain better visual effects.

In order to reduce the negative impact of the horror image on social development and adolescent growth, a violent image annotation algorithm based on ensemble learning is proposed, assisting in screening out the horror information in the webpage. The annotation of violent image is considered as a multi-label classification problem in this method. Multiple sub-networks are trained through transfer learning, and then the ensemble learning is introduced to fuse the outputs of sub-networks. In the process of fusion, weights are allocated according to the precision of each label on different networks, thus the annotation result is obtained through a series of matrix operations. The experimental results show that the proposed method achieves a great improvement in precision and recall than traditional machine learning algorithm, and also improves the problem that the precision of model annotation on different labels varies greatly due to the label category imbalance.

The new-energy vehicles are mainly powered by electricity. Its power cables transmit the transient signals of high voltage and power. In addition, the antenna effect of the cable makes the cable become the dominant emission source of electromagnetic interference in electric vehicles. The power cable also exerts the crosstalk on the surrounding signal cables. This research analyzes the source of electromagnetic interference in power cables. Then the simplified model of power cable and adjacent cable is established using FEKO tool. The effect of these factors on restraining crosstalk is compared and analyzed by changing the settings of the cable shield layer and the metal groove in the simulation model. Finally, based on the simulation results, it is concluded that reducing the impedance of the cable shielding layer and increasing the use of the metal groove are helpful to minimize the cable’s crosstalk effects.

Two-step dry etch method in the Inductive Coupled Plasma(ICP) chamber is proposed and applied to the SiN gate foot definition during the self-aligned 0.5 micron T-gate fabrication for GaN High Electron Mobility Transistor(GaN HEMT). The main etching part forms a tilted silicon nitride side wall, which reduces the electric field intensity in the channel under the gate and improves the gate metal filling in the silicon nitride recess. The soft landing part performs the over-etching process with a very low bias power to ensure the complete removal of silicon nitride and reduce the channel damage. Compared with the control device without any optimization, the off-state breakdown voltage of the optimized device shows an obvious increase from 140 V to more than 200 V. Moreover, the output power density and the Power Added Efficiency(PAE) at 3.5 GHz are promoted from 5.8 W/mm to 8.7 W/mm and 55.5% to 66.7%, respectively. After un-biased highly accelerated stress test for 96 hours, no obvious change in the appearance of the optimized device can be observed, and the change of the maximum drain current is less than 5%, indicating that the device reliability is pretty good.etch；performance improvement；reliability

This paper introduces the research and design of the tunable filter based on the GaAs varactor technology, including the preparation of the variable capacitance material, the establishment of the diode model and the circuit design. Several tunable filters are designed, and the frequency range covers 1-19 GHz. The test results show that the tunable filter has more than one frequency doubling range, the breakdown voltage of the device is greater than 30 V. In this paper, a tunable filter with central frequency 2-5 GHz adjustable is introduced in detail. The control voltage range is 0-15 V, the insertion loss is about 10 dBm, and the input and output return VSWR(Voltage Standing Wave Ratio) are better than 1.8. The series of tunable filters designed bear the advantages of high consistency, miniaturization, low cost and debugging-free, and have good application prospects.

Medical cyclotron is widely used in the fields of national defense basic science research, isotope production, diagnosis and treatment. High stability and security of medical cyclotron system are required in these situations. Based on these application requirements, this paper outlines the composition of the control system, briefly introduces the role of each subsystem, and finally the system integration with RX3i programmable controller as the front-end controller, the control strategy of low-level control of high frequency system and the security interlock of each system under different working modes are elaborated in detail. The results of continuous operation show that the control system ensures the stable operation of the whole system and meets the requirements of accelerator operation and debugging.

In view of the different cleaning frequency ranges required by different cleaning materials in the industry, the ultrasonic transducers that need to be connected are also different. In order to improve the adaptability of the ultrasonic generator to transducers of different frequency ranges, this paper designs a closed-loop control system for frequency reference and frequency-locked loop circuit tracking by the upper monitor. The whole system generates Pulse Width Modulation(PWM) pulse signals by STM32 main controller. It controls the EXB841 optimized driving circuit, driving high frequency full bridge invert circuit, and ensuring the maximum output power of the transducer through impedance matching and output current detection. At the same time, for different frequency ranges of ultrasonic transducers, it is necessary to adjust the given input and keep the generator in a certain range of base-value for frequency tracking. The working frequency of the ultrasonic transducer test prototype is 28.8?kHz and the maximum power is 1?500?W. The resonant frequency of the system is kept around 28.8?kHz after the system is connected, and the output power is approximately the maximum. The test results show that the system has good adaptability to the ultrasonic transducer whose working frequency is 20-40?kHz.

A practical bandgap reference voltage source is designed with quick start and high stability. It selects 110?nm CMOS(Complementary Metal Oxide Semiconductor) device technology and simulation software of Cadence. The simulation results indicate that under room temperature, the output voltage is 1.2 V under 3.3 V power supply; the temperature drift coefficient is 33 ppm/℃ between -40?℃-85 ℃; it takes 0.5?μs to start the circuit; and the power consumption is 0.967?mW; Power Supply Rejection Ratio(PSRR) reaches -61?dB in the range of low frequency; the layout covers an area of 50?μm×180?μm. Because its simple structure and being easy to be integrated, the circuit can be applied to DAC(Digital to Analog Converter) with high speed and high precision.