Light Emitting Diode (LED) is an important nonlinear optical signal generating device in Visible Light Communication (VLC) systems. The static nonlinearity of LED and the dynamic nonlinear memory effect caused by frequency response can lead to signal distortion and degrade the system performance.

The nonlinearity of LED can usually be compensated by estimating a pre-distortion model, and the wireless VLC is usually intensity modulated/directly detected.

Therefore, the LED nonlinearity modeling is investigated, and an adaptive pre-distortion device based on Amplitude Factorization Polynomial (AFP) is proposed to compensate LED nonlinearity in this paper. Simulation results show that the proposed scheme exhibits good Bit Error Rate (BER), Amplitude Modulation/Amplitude Modulation (AM/AM) correction and constellation diagram performance with higher modeling accuracy.

The normalized mean square error performance improvement of more than 15 dB compared with the traditional memory polynomial is achieved.

Aiming at the limited transmission distance of ultra-long distance optical transmission system using remote pump technology, this paper proposes a single span repeaterless transmission system with mixed transmission of 100 and 200 Gbit/s service. The experiment tests the maximum span loss of the system by adjusting the input power of the system and the position of the remote pump gain unit in the system without considering the influence of other factors such as dispersion.

The key amplification technologies in the experiment mainly include remote pump amplification technology, Erbium-doped optical fiber amplification technology and Raman amplification technology. In the transmission, Erbium-doped optical fiber amplifier, Raman amplifier and remote pump gain unit are used to carry out non-relay amplification of optical signals in the system, and the relay transmission distance is further improved by adjusting the position of remote pump gain unit. The remote pump bypass access method which can improve the amplification gain of remote pump is used in the experiment.

Three different modules were used in the experiment, with a total of 9 service waves. The modulation formats of 200 Gbit/s Polarization Multiplexed - Quadrature Phase Shift Keying (PM-QPSK), 100 Gbit/s PM-QPSK and 200 Gbit/s Polarization Multiplexed -16 Quadrature Amplitude Modulation (PM-16QAM) were used respectively. The transmission experiments were constructed by G. 654b ultra-low loss optical fiber with a loss of 0.17 dB/km. Finally, the single-span ultra-long distance transmission with a total capacity of 1.5 Tbit/s and a cross-segment loss of 75 dB is realized.

The results show that the transmission distance of the system can be improved by setting the optimal fiber input power and adjusting the position of the gain unit of the remote pump. The experimental results provide reference value for the construction of the ultra-long distance optical transmission system based on the remote pump technology, and provide reference data for the development of the repeaterless transmission based on the remote pump technology.

With the continuous emergence of information-based bandwidth-consuming services such as 5th Generation Mobile Communication Technology (5G), the Internet of Things, and cloud computing, people’s demand for high-speed information transmission has increased dramatically. However, due to the inherent nonlinear effects, the transmission capacity of single-mode optical fiber has approached the Shannon limit, and it will no longer be able to meet people’s needs for ultra-high-speed and large-capacity transmission. Solving the transmission capacity problem has become a top priority. In order to solve the needs of large-capacity communication systems and long-distance high-speed transmission problems, we have built a two-mode single-channel C-band Few Mode Fiber (FMF) transmission system.

At the transmitting end, an arbitrary waveform generator is used to convert the digital signal into an electrical signal and drive the In-phase and Quadrature (IQ) modulator to modulate the optical carrier. The modulated signal is transmitted simultaneously using two multiplexing technologies: Mode Division Multiplexing (MDM) and Polarization Division Multiplexing (PDM). In order to achieve long-distance transmission of dual-mode signals of 1 000 km, we construct a dual-circulation loop system. Each time the signal passes through the loop, it will pass through a 50 km FMF. After being transmitted to the target distance, the coupler outputs the signal to the demultiplexing module, and the coherent optical receiver performs homodyne detection on the demultiplexed modulated signal. Finally, the transmitted signal is stored in an oscilloscope for offline Digital Signal Processing (DSP). The signal is sequentially subjected to frequency domain dispersion compensation, downsampling, clock recovery, and least mean square algorithm to restore the original signal.

It was found that within the range of Optical Signal-to-Noise Ratio (OSNR) of each channel in the experiment, the Bit Error Rate (BER) under low Signal to Noise Ratio (SNR) is close to the theoretical channel result. Under high SNR condition, the BER is 1×10^-2, which is 2.5 dB away from the theoretical value. We test the BER of LP_11a and LP_11b modes at Back-To-Back (BTB) and 250, 500, 750 and 1 000 km transmission cases respectively. The BER at all distances are lower than the Low-Density Parity-Check (LDPC) soft decision threshold with 28% redundancy (5.2×10^-2 Soft Decision - Forward Error Correction (SD-FEC)). The BER after 1 000 km transmission in the two modes are 1.7×10^-2 and 1.8×10^-2 respectively, and the total net transmission rate is 400 Gbit/s.

This article demonstrates the transmission of a 32 Gbaud MDM-PDM-16 Quadrature Amplitude Modulation (QAM) C-band signal in a 1 000 km two-mode single-channel FMF system. At the receiving end, the advanced Multiple Input Multiple Output (MIMO)-DSP algorithm is used for channel equalization, and the obtained two-mode BER of 1.7×10^-2 and 1.8×10^-2 are both lower than the LDPC SD-FEC threshold with 28% redundancy. The result reachs a domestic record of 400 Gbit/s net transmission rate based on FMF transmission, and highlight the potential of FMF in large-capacity long-distance transmission.

There is a problem that traditional unrepeatered transmission system need to use disturbed Raman amplifier and remote optical pump amplifier, while both of these two amplifiers need to use high power pump laser which will increase the complexity of fiber link in unrepeatered transmission system. In this paper, we propose a new structure of unrepeatered transmission system using Erbium Doped Optical Fiber Amplifier (EDFA) to replace the forward disturbed Raman amplifier.

We analyze and compare the performance of the proposed unrepeatered transmission system, and realize a real-time unrepeatered transmission experiment with the proposed structure.

The results show that the gain effect of forward Raman pump in traditional unrepeatered communication system can be achieved by using high power amplifier to increase the signal launch power, and the new structure of unrepeatered transmission system is more suitable to achieve higher bitrate with the help of Wavelength Division Multiplexing (WDM) technology. We confirm the unrepeatered transmission system can achieve 500 km span distance at 10 Gbit/s bit rate in single channel communication transmission, and 500 km span distance at 4×10 Gbit/s bit rate in the WDM technology.

The proposed unrepeatered transmission system can simplify the traditional unrepeatered transmission system, and increase the bit rate by using multi-channel WDM system with EDFA. It has important practical significance to simplify the structure and improve the bit rate in future unrepeatered transmission system.

The use of random mapping and traditional constellations in Optical-Orthogonal Frequency Division Multiplexing (O-OFDM) - Index Modulation (IM) system cannot reach the optimal Bit Error Rate (BER) performance. Therefore, a step-by-step optimal signal mapping method is proposed based on the optimal selection of active sub-carriers and the optimal rotation of constellation to improve the BER performance of O-OFDM-IM system.

At the transmitter side, the method firstly selects the optimal combination of active subcarriers based on the channel state information and the channel norm maximization criterion. Then considering ?=0.5° as the step size, an exhaustive search algorithm is used to obtain the optimal constellation by using the channel state information. Next, the optimal constellation graph obtained by rotation is loaded on the active sub-carrier for signal transmission. At last, the Maximum Likelihood (ML) detection algorithm is used to recover the original signal at the receiver to minimize the system error probability.

The results show that at the condition of BER=10^-4 and strong turbulence, the Signal to Noise Ratio (SNR) is improved by about 1.45 dB when the modulation order is 16 in (4, 2) system. In addition, with the increase of turbulence intensity, the improvement of BER performance is more obvious. For example, when the BER=10^-4 the SNR of the (4, 2) system at 8 Quadrature Amplitude Modulation (QAM) is improved by about 1.35 dB in strong turbulence and 1.2 dB in weak turbulence.

Based on the known channel state information, a step-by-step optimal signal mapping method for O-OFDM-IM is proposed in this paper, which can effectively improve the BER performance of the system in strong turbulence channels.

In order to solve the problem that it is difficult to make the temperature sensor of the Photonic Crystal Fiber (PCF) with Surface Plasma Resonance (SPR), a D-structure PCF temperature sensor is proposed. The micro-structure and arrangement of the air hole are also designed. The method of coating the polishing surface of the optical fiber with gold layer is used to stimulate the SPR of the optical fiber.

Ethanol is used as a temperature-sensitive material, and the temperature-sensitive effect is used as the temperature sensing mechanism to achieve the effect of temperature sensing. The finite element analysis method is used to simulate the thickness of the gold layer and the size of the elliptical air hole of the fiber respectively. The coupling characteristics between the core mode and the surface plasma mode are analyzed under different conditions, and the structure of the sensor is optimized.

The temperature characteristics of the sensor are simulated. In the range of -80~80 ℃, the sensitivity of the sensor is 3.4 nm /℃, and the linearity is better than 0.969 63.

The temperature sensor designed in this paper has a wide temperature detection range, satisfactory sensitivity, and linearity, which can be used for real-time temperature monitoring in industrial, agricultural, and medical fields.

In order to solve the problem of cross sensitivity of temperature and strain in Fiber Bragg Grating (FBG) sensing, the response characteristics of temperature and strain in each mode of Few Mode Fiber (FMF)-FBG were studied in this paper. A four-mode FBG temperature and strain dual parameter sensor was also proposed.

The basic mode and a few higher modes can be transmitted simultaneously in the low-mode fiber, taking into account the advantages of low mode dispersion of Single Mode Fiber (SMF) and low nonlinearity of multi-mode fiber, which can be used for sensing multiple physical quantities at the same time. By analyzing the different optical power sensitivity of the LP₀₁ mode in the FMF-FBG at different temperatures, and considering the good linearity of wavelength, temperature and strain under this mode. Thus, the temperature and strain dual parameter sensing of FMF-FBG is realized.

The results show that the sensor can better solve the problem of cross sensitivity of temperature and strain.

Compared with SMF-FBG, the FMF-FBG has several reflection peaks of different modes at the same time, which can not only solve the cross-sensitivity problem of temperature and strain, but also greatly improve the accuracy and stability of sensing, which may have a good application prospect in the field of new type of sensing.

In the production test process of 50 Gbit/s Small Form Pluggable (SFP) 56 optical module, the number of instruments and equipment to be matched for the test is large. Therefore, the test instruments are used in isolation, and the test process is complex. These issues make the optical module test longer, and the test efficiency is relatively low. It brings a lot of inconvenience to the test personnel, and can not guarantee the consistency of the test products. The purpose of this paper is to enhance debugging and testing efficiency while alleviating testers' workload.

Python language and its robust third-party libraries, including the VISA and pyserial libraries, are utilized to develop an automatic testing system for SFP56 optical modules compatible with the Windows operating system.

The system can debug and test the optical module automatically. The test instruments can be used in parallel, reducing the waste of time and resources. According to the test requirements, the optical module can also be tested step by step, which greatly improves the efficiency of the debugging and testing of the optical module in the production line. It can also reduce the requirements of the tester. The verification shows that compared with the traditional optical module test, the test time of each optical module is saved by about 70 s, which greatly improves the test efficiency of the optical module. In the optical module test, the result deviation caused by human operation is avoided, and the product consistency can be guaranteed.

In comparison to Labview-based automatic optical module testing systems, which are currently prevalent in the market, the system designed in this study is more scalable and convenient for upgrades and maintenance. Moreover, it can be expanded to support optical module testing at speeds of 100, 200, 400, and 800 Gbit/s, thus laying a technical foundation for future high-speed optical module testing.

The symbol decomposition technique aims to decompose Optical Orthogonal Frequency Division Multiplexing (O-OFDM) symbols into multiple symbols with a small Peak to Average Power Ratio (PAPR) to mitigate the non-linear effects of Light Emitting Diodes (LEDs). However, O-OFDM generates more decomposed symbols with a higher PAPR, thereby reducing the information rate and degrading the Bit Error Rate (BER) performance.

The proposed design of Precoding O-OFDM Adaptive Symbol Decomposition with Serial Transmission (PCO-OFDM-ASDST) system uses precoding scheme to reduce the PAPR of O-OFDM symbols. The proposed method can reduce the average number of symbol decompositions for adaptive symbol decomposition. The theoretical Signal-to-Noise Ratio (SNR) expressions are derived for multipath channels, and the performance of PAPR, BER, and information rate is analyzed using Monte Carlo BER simulations.

The results demonstrate that under 4 Quadrature Amplitude Modulation (QAM) modulation in Asymmetrically Clipped Optical Orthogonal Frequency Division Multiplexing (ACO-OFDM) systems, the symbol power required for PCO-OFDM-ASDST is 7 dB less than that of Adaptive Symbol Decomposition with Serial Transmission (ASDST) when the BER is 10^-4. Moreover, under 64QAM, the information rate can be increased by 10 Mbit/s when the symbol power is 20 dBm.

The results indicate that the PCO-OFDM-ASDST outperforms ASDST systems in terms of BER and information transmission rate.

The light source of fiber grating sensing system provides energy for the system. The stable and continuous light source is crucial for the application of the whole demodulation system. When the Modulated Grating Y-branch (MG-Y) laser is used as the light source of the fiber grating demodulation system, it is necessary to solve the problem that it is difficult for the MG-Y laser to quickly obtain the stable and continuous tuning within the required specific wavelength range.

In order to obtain the stable and continuous wavelengths quickly for MG-Y lasers, the K-Means (KMeans) clustering model is proposed and applied to obtain high-quality current-wavelength Look Up Table (LUT) for MG-Y lasers.

By combining the tuning characteristics of the MG-Y laser with the KMeans clustering model method, the centroid of the optimal tuning parameter curve in the phase tuning region of the LUT can be quickly obtained with time of about 18.26 s. Then, according to the tuning characteristics of the left, right and phase of the MG-Y laser itself, the fine tuning of the MG-Y laser wavelength can be achieved by uniform interpolation method. The LUT with the full tuning range is obtained from the KMeans clustering model, and the target LUT is selected at a certain interval. According to the output wavelength issued by the laser according to the target LUT, the mean value of the measured wavelength and the absolute value of the error of the target wavelength is 0.18 pm. The standard deviation of the absolute value of the error is 0.52 pm. Therefore, the MG-Y laser can be used as a stable light source for the fiber grating sensing system.

It is verified by experiments that the traditional manual search for each phase tuning region needs at least 30 min to find the centroid point. The same operation can be completed by KMeans clustering model method, which can greatly shorten the time to find the centroid point. Based on this method, the stable wavelength of quasi-continuous tuning can be achieved by controlling the MG-Y laser.

The baseline drift of the Fiber Bragg Grating (FBG) spectral signal is usually one of the main problems, caused by the complex external environment. A spectral baseline correction method based on the improved Long Short Term Memory (LSTM) model is proposed in this paper.

Compared with LSTM model, the improved LSTM model extracts feature information of FBG spectral signal by the Convolutional Neural Network (CNN). The improved LSTM model is composed of CNN, full connection, and LSTM network. In this paper, the improved LSTM model is trained by artificial datasets and measured datasets. The artificial datasets are made up of feature noise, baseline, and FBG spectroscopy. Five methods including wavelet soft threshold method, penalty least square method, Recurrent Neural Network (RNN), LSTM, and the improved LSTM model are used as baseline correction. Identification signal probability and Root Mean Square Error (RMSE) are used to evaluate correction results by the five methods.

The artificial datasets of FBG signal are corrected by the improved LSTM model, and the identification signal probability is increased by 60.8%. The improved LSTM model with training by artificial datasets and measured datasets shows better correction results, compared with training by measured datasets. The mean of the RMSE for FBG spectrum decreases by 10.95%. The standard deviation of RMSE decreases by 4%. The measured datasets of FBG signal are corrected by the improved LSTM model, and the identification signal probability is increased by 50.5%. Compared with wavelet soft threshold method, penalty least square method, RNN and LSTM, the improved LSTM model shows best correction results. The mean values of RMSE and the standard deviation of RMSE are 0.012 2 and 0.002 4, respectively. The RMSE value of the demodulated central wavelength is 0.036 pm. And the baseline correction process takes only 9.68 ms.

The improved LSTM model is an effective method to achieve baseline correction, and has wide range of application prospects in complex external environment.

In this paper, machine learning method is applied to 30 Tbit/s (60 × 500 Gbit/s) Nyquist Dual Polarization-16 Quadrature Amplitude Modulation (DP-16QAM) system after 6 300 km transmission in G. 654E optical fiber. Nonlinear channel equalization is used to reduce the transmission Bit Error Rate (BER).

Referring to the "receptive field" mechanism of convolution neural network, the size of "convolution core" is designed, and the data set is constructed according to the divided sampling data. The artificial neural network is constructed by optimizing the parameters. The one-to-one data corresponding to the transmission and reception of different wavelengths, different optical signal-to-noise ratios, and different fiber input powers in the C-band are collected. Refer to the classic full-connection neural network structure, the neural network is constructed according to the data structure of the data set. The network fitting is carried out for the real part and the imaginary part respectively. After training stage, the test data is sent into the network, and the performances are compared with the traditional methods.

Two kinds of neural networks are used to fit the transmission BER under 60 different wavelength transmission conditions of C band frequency from 191.562 5 to 195.987 5 THz. Compared with Maximum Likelihood Sequence Estimation (MLSE), Network 1 has an average reduction of 23% in BER, and Network 2 has an average reduction of 41% in BER. A frequency of 193.812 5 THz is then selected for the calculation of the fiber input power ranging from 14 to 19 dBm. The average improvement in network 1 and network 2 are 32% and 52%, respectively. Under different optical signal-to-noise ratios, Network 1 has an average improvement of 30%, and Network 2 has an average improvement of 57%.

The two neural networks have excellent performance in nonlinear equalization of coherent transmission systems. At the same time, the number of network layers and nodes will jointly affect the fitting results. Increasing the number of layers and nodes can obtain better fitting results, but the corresponding parameters, training time and the required space will also increase. Therefore, in the application, the actual situation should be considered to choose between the fitting performances and the model attributes.

The optical communication system working at the traditional telecommunication band is faced with "capacity crunch" with the rapidly development of optical communication technology. The acceptable loss of 0.2 dB/km (@2 μm) from the hollow-core photonic bandgap fibers as well as the high optical amplification gains (30 dB@2 μm) of thulium-doped fiber amplifiers (CTFA type) provides the potential for the 2 μm wavelength band to alleviate the communication "capacity crunch". As a result, the needs of the 2 μm silicon-based photodetectors is also raised due to its mature and convenient manufacture process. The main solutions of the silicon-based photodetectors include the usage of Ⅲ-Ⅴ compounds with tunable band gap, the introduction of low band gap width materials as absorption regions, the use of new absorption mechanism in optical absorbed material at the 2 μm wavelength band and so on. The Ge material has a high Two Photon Absorption (TPA) coefficient (1 225 GW/cm@2 μm). And the Ge-on-Silicon-On-Insulator (SOI) photodetector is an excellent method to realize 2 μm optical signal detection, which is compatible with the standard silicon photonics device manufacturing process that has the advantages of low production difficulty and cost. The purpose of this paper is to verify the feasibility of the Ge-on-SOI photodetector to achieve 2 μm wavelength band photoelectric detection by using the special optical absorption mechanism of TPA.

In this paper, the photoelectric detection at the 2 μm wavelength band based on the physical absorption mechanism of current generated from high TPA of Ge material is realized. The quantification of the photocurrent generated by Ge material through TPA effect is analyzed and discussed in this paper. In the experimental test, we first apply a high power 2 μm wavelength band input light source magnified by thulium-doped fiber amplifiers to the input port (grating coupler) of the commercial waveguide-type Ge-on-SOI photodetector in the on-chip active silicon photonics device test system. Then we adjust the alignment of the optical fiber with the grating coupler and the input light polarization to reduce the optical transmission loss from the input optical fiber to the Ge absorption region. Finally, the TPA photocurrent of the on-chip photodetector is obtained by a probe.

A net photogenerated current up to 651 nA and an estimated value of responsivity greater than 10 mA/W is obtained experimentally under a 2 μm wavelength optical input power of 21.9 dBm.

The work of this paper verifies the scientific feasibility of 2 μm waveband photodetector through the TPA effect of Ge material, and provides experimentally support for the design of the 2 μm wavelength band Ge-on-SOI photodetector based on TPA.

In long-distance optical communication systems, compensating fiber nonlinear impairment through traditional Digital Signal Processing (DSP) is difficult due to intractable interactions between Kerr nonlinearity, chromatic dispersion and amplified spontaneous emission noise. Machine learning algorithm can be used to further process signals on the basis of traditional DSP to mitigate fiber nonlinear impairment and improve long-distance transmission performance.

In this paper, the traditional Digital Back Propagation (DBP) algorithm is combined with Deep Neural Network (DNN), where the linear step size and nonlinear step size in DBP are taken as one neuron. It means that the linear step size is taken as the weight matrix of DNN, the nonlinear step size is taken as the activation function, and DSP is taken as the static layer of DNN. A DNN-based Learned Digital Back Propagation (LDBP) algorithm is proposed.

In order to verify the feasibility of the proposed LDBP algorithm, the simulation was carried out in a single-channel polarization division multiplexing 16-ary Quadrature Amplitude Modulation (QAM) optical transmission system. The numerical simulation results demonstrate that the 1-step-per-span LDBP algorithm improves the optimal launched power from -2 dBm to 1 dBm in compared to linear equalization. Meanwhile, compared with DBP with the same computational complexity, the proposed algorithm improves Signal-to-Noise Ratio (SNR) by 0.82 dB at the transmission distance of 1 200 km. In addition, compared with DBP with the same computational complexity and linear equalization, the SNR of the transmission system corresponding to LDBP method decreases more slowly with the increase of transmission distance, and the algorithm can work without knowing the link parameters, showing the characteristic of universality and robustness.

The proposed LDBP algorithm is more suitable for practical long-distance coherent optical communication system than the traditional DBP algorithm.

Single-photon detection and single-photon ranging systems have widespread applications in three-dimensional imaging and long-distance spatial remote sensing. However, device limitations and background noise from the sun light limit the application to only nighttime conditions.

We have improved traditional laser radar systems by implementing all-fiber optics systems with ultra-narrowband fiber optic filtering, leading to significantly improved system stability. By implementing mechanical motion control, we are able to attain single-photon imaging systems with ultra-wide scanning and ultra-high resolution. The systems can operate in optical environments that are typically challenging, such as daytime or foggy weather. Moreover, the laser radar system is designed based on the 1 550 nm wavelength which has good atmospheric penetration and low transmission loss, enhancing its ability to work in foggy and rainy conditions.

In a foggy condition with a visibility of about 500 m, the system achieves three-dimensional imaging of objects 1.6 km away, which is a distance 3 times longer than the visibility at the time. The distance resolution can be significantly enhanced by optimizing the intensity and distance information of the histogram using the minimized negative logarithmic likelihood function, with the assistance of a high-resolution time/digital converter and the model constructed by the system response function of the laser radar. Compared to the maximum value method with a distance resolution of 0.05 m, this system achieves a distance resolution of 0.006 m via the minimized negative logarithmic likelihood function method. Additionally, the image normalization processing is performed to suppress fluctuations in photon technology and eliminate noise in the image. The 1 550 nm all-fiber laser radar system has strong mechanical stability and can work in adverse weather conditions.

It has significant application value in remote sensing and mapping, ground and vehicle-mounted laser radar, and other fields.

In order to solve the problem of the limited bandwidth of polarization insensitive optical power splitter, an ultra-broadband polarization insensitive optical power splitter based on Silicon Nitride (Si₃N₄) waveguide is proposed.

The ultra-broadband polarization insensitive optical power splitter is designed based on adiabatic coupling Si₃N₄ waveguide.Optical bandwidth of 300 nm (from 1 400 ~ 1 700 nm) is demonstrated for both Transverse Electric (TE) and Transverse Magnetic (TM) modes.

Over the entire optical bandwidth, the losses for TE and TM mode are less than 0.10 and 0.15 dB, respectively.The footprint of the proposed ultra-broadband polarization insensitive optical power splitter is only 60.00 μm×2.75 μm, which means that the optical power splitter can be used for large-scale integration.Furthermore, the relationship between the performance and the structure parameters of the proposed optical power splitter is studied.

The results show that the performance of the optical power splitter maintains almost unchanged by varying the structure parameters with ±50 nm.In other words, the optical power splitter possesses large fabrication tolerance which can be easily fabricated.

To improve the linear fit between the reflected spectral center wavelength and external environmental variables in Fiber Bragg Grating (FBG) sensing system, this paper proposes to use particle swarm optimization of Gaussian process regression model to the field of FBG stress sensing.

For the reflectance spectral characteristics of the FBG, the paper studies the impact of the linear fit in the spectral fitting of FBG sensing system. The particle swarm algorithm is used to search for the optimal hyperparameters in the Gaussian process regression model in order to enhance the predictive performance of the reflectance spectral wavelength of the center. A FBG stress sensing experimental platform was built, and the FBG was laid on the strength beam. Different weights were applied to one end of the equal strength beam to produce axial strain on the FBG, and the reflectance spectral data were collected by the spectrometer and analyzed by linear fitting with the studied model. The results obtained by the unoptimized Gaussian process regression model, the maximum value method, the Gaussian fitting method, and the center of mass method were used as the control group.

The results show that under the conditions of erbium-doped fiber amplifier output power of 10 dBm, transmission fiber distance of 50 m, and the number of sampling points of the spectrometer of 501, the linear fit between the reflected spectral center wavelength and the mass of the weights is better than that of the control group. The linear fit of the studied model can reach up to 0.951 9, which is improved compared with that of the control group. Under the conditions of 501, 251, 167 and 126 spectral sampling points, the studied model can improve the linear fit of the system to 0.990 0, which is a maximum improvement of 0.258 7 compared with the maximum value method.

The analysis results show that the Gaussian process regression model optimized by the particle swarm is able to effectively improve the linear fit of the FBG stress sensing system.

E/W band millimeter-wave frequency covers 60~110 GHz, with abundant frequency band resources and potential for high-capacity transmission. Compared with lasers, E/W band millimeter-wave beam is wider, with advantages of directionality, anti-interference and ease of alignment, making it a promising frequency band for long-distance wireless communication at sea. However, the characteristics of near-sea surface channels such as atmospheric absorption and multipath at the sea surface limit the transmission distance of E/W-band millimeter waves, resulting in bottlenecks in E/W-band millimeter wave transmission near-sea surface. This paper focuses on the direction of long-distance millimeter-wave wireless transmission at sea, using microwave photonics technology to propose an ultra-wideband millimeter-wave modulation and demodulation system architecture to improve the millimeter-wave frequency conversion and modulation performance. We then study the optimal high-order modulation format in near-sea surface long-distance transmission systems to achieve high-capacity long-distance transmission.

The power amplifier in long-distance communication systems can introduce nonlinear distortion to the signal. It analyzes the factors affecting the performance of high-order modulation formats such as Amplitude Phase Shift Keying (APSK) and Quadrature Amplitude Modulation (QAM) due to the nonlinear characteristics of solid-state power amplifiers. The optimal distribution, Euclidean distance and inner and outer radius ratio of 16APSK signals have been studied. By the equivalent power experiments, the nonlinear resistance of 16QAM and 16APSK signals are compared and analyzed.

The millimeter-wave long-distance transmission experiment is carried out for performance investigation. It proves that the APSK modulation format is more suitable for long-distance transmission system than QAM format.

Based on the 16APSK modulation format, the performance of 16 Gbit/s, equivalent 34 km long-distance transmission are achieved in the experiment, providing high-value technical means for long-distance high-capacity transmission at sea.