Digital pre-distortion method for improving linearity of power amplifier in phased array system

Abstract

The invention relates to the technical field of wireless communication, and discloses a digital pre-distortion method for improving the linearity of a power amplifier in a phased array system, which comprises the following steps of: 1, acquiring a nonlinear parameter of the power amplifier under a dynamic load in real time, the nonlinear parameters comprise an input signal amplitude-phase characteristic, a temperature parameter and a quantum tunneling current parameter; 2, generating a pre-distortion compensation signal based on the nonlinear parameter, wherein the pre-distortion compensation signal comprises a quantum phase compensation item and a dynamic phase change compensation item; and step 3, superposing the pre-distortion compensation signal to an original input signal to generate a pre-distortion modulation signal. The linearity of the power amplifier is improved by introducing the quantum phase compensation item and the dynamic phase change compensation item, nonlinear parameters are accurately obtained in real time in a dynamic load environment, and the precision of pre-distortion compensation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communication, and specifically to a digital predistortion method for improving the linearity of power amplifiers in a phased array system. Background Art

[0002] In modern wireless communication and radar systems, phased array technology is widely used in high-performance signal processing and beamforming. However, as a core device in the RF front-end, the nonlinear distortion of power amplifiers (PAs) seriously affects the overall performance of the system. Especially in the large-signal operating state, the nonlinear effects of power amplifiers will cause gain compression, amplitude-phase distortion (AM-AM and AM-PM distortion), and out-of-band signal leakage, thereby reducing the quality of the transmitted signal. Therefore, improving the linearity of power amplifiers has become one of the key technologies for optimizing the performance of phased array systems, and digital predistortion (DPD) technology is one of the most effective solutions at present.

[0003] Existing digital predistortion technologies mainly rely on static or semi-static parameter estimation methods, such as polynomial modeling based on the static characteristics of power amplifiers, memory effect compensation algorithms, etc. These methods can improve the linearity of power amplifiers to a certain extent, but there are still obvious limitations. On the one hand, traditional nonlinear modeling often ignores the influence of dynamic load effects, temperature changes, and device quantum characteristics on nonlinear distortion, resulting in insufficient compensation accuracy. On the other hand, current optimization algorithms, such as the least mean square error (LMS), gradient descent-based adaptive predistortion methods, have a slow calculation convergence speed in complex scenarios and are difficult to meet the requirements of real-time and high-precision for high-speed phased array systems.

[0004] In view of the above problems, the present invention proposes a method for obtaining nonlinear parameters by combining quantum tunneling current parameters and temperature gradient models. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a digital predistortion method for improving the linearity of power amplifiers in a phased array system, which solves the problems of inaccurate acquisition of nonlinear parameters, single compensation method, slow convergence speed, and insufficient accuracy of existing digital predistortion technologies in a dynamic load environment.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A digital predistortion method for improving the linearity of power amplifiers in a phased array system, comprising the following steps: Step 1: Real-time obtain the nonlinear parameters of the power amplifier under dynamic load, where the nonlinear parameters include the amplitude-phase characteristics of the input signal, temperature parameters, and quantum tunneling current parameters; Step 2: Generate a predistortion compensation signal based on the non-linear parameters, where the predistortion compensation signal includes a quantum phase compensation term and a dynamic phase transition compensation term; Step 3: Superimpose the predistortion compensation signal on the original input signal to generate a predistortion modulation signal; Step 4: Dynamically optimize the predistortion parameters according to the fusion data of the direct connection feedback signal and the air interface feedback signal.

[0007] Preferably, the acquisition of the non-linear parameters includes: Real-time measure the quantum tunneling current parameters through the gate leakage current of the power amplifier; Measure the junction temperature of the power amplifier through a thermoelectric coupling sensitive layer, and calculate the temperature gradient parameters based on the temperature gradient model; The non-linear relationship between the quantum phase compensation term and the quantum tunneling current parameters and temperature gradient parameters is mapped through a pre-trained model.

[0008] Preferably, the dynamic optimization of the predistortion parameters includes: Construct an optimization objective function, which includes a weighted combination of the error vector magnitude and the entropy production rate, where the value range of the weight coefficient λ is from 0.1 to 1.0; Solve the minimum value of the objective function through a quantum annealing algorithm and update the coefficient of the predistortion compensation term.

[0009] Preferably, the generation of the predistortion compensation signal includes: Use an AlGaN / GaN superlattice heterojunction as the quantum tunneling sensitive layer, and the heterojunction includes a boron nitride tunneling barrier layer and a doped zinc oxide thermoelectric conversion layer; Generate a quantum phase compensation term through the tunneling current signal and temperature sensitive signal output by the quantum tunneling sensitive layer.

[0010] Preferably, the preparation of the quantum tunneling sensitive layer includes: Epitaxially grow an AlGaN / GaN superlattice structure on a substrate; Form a boron nitride tunneling barrier layer on the surface of the superlattice through atomic layer deposition, and the thickness of the boron nitride layer is 1.5 nm to 3.0 nm; Deposit a doped zinc oxide thermoelectric conversion layer on the boron nitride layer through magnetron sputtering, and the doping concentration is 3% to 8%.

[0011] Preferably, the calculation architecture for dynamically optimizing the predistortion parameters includes: A hybrid processor of a quantum annealing calculation unit and a classical logic unit; The quantum annealing unit is used to solve the entropy production rate constrained optimization problem, and the classical logic unit is used to generate the predistortion compensation signal in real time.

[0012] Preferably, the transmission rate of the silicon photonic interconnection module is 350 Gbps to 450 Gbps, and the bit error rate is lower than 5×10 -12 , and it is used to transmit carrier density wave data and thermal imaging data.

[0013] Preferably, the acquisition of the real-time feedback signal includes: Collecting the direct connection feedback signal of the output signal of the power amplifier through a directional coupler; Separating the target channel signal in the air interface feedback signal through the receiving end beamforming technology; Performing signal-to-noise ratio weighted fusion on the direct connection feedback signal and the air interface feedback signal, and the weight coefficient is dynamically adjusted according to the signal-to-noise ratio, and the signal-to-noise ratio threshold is 10 dB to 30 dB.

[0014] Preferably, the generation of the predistortion modulation signal includes: Performing 16-bit fixed-point number operation on the quantum phase compensation term and 32-bit floating-point number operation on the dynamic phase transition compensation term; The processing delay of the predistortion modulation signal is 0.5 microseconds to 1.0 microseconds.

[0015] Preferably, a thermoelectric coupling sensitive layer is integrated in the power amplifier package. The temperature sensitivity of the sensitive layer is 0.3% / °C to 0.7% / °C, and the size of the sensitive layer is 1.0 mm×1.0 mm to 2.5 mm×2.5 mm, and it is connected to the power amplifier chip through eutectic soldering.

[0016] The present invention provides a digital predistortion method for improving the linearity of a power amplifier in a phased array system. It has the following beneficial effects: 1. The present invention adopts a non-linear parameter acquisition technology that combines quantum tunneling current parameters and a temperature gradient model, achieving real-time and accurate acquisition of the non-linear characteristics of the power amplifier under dynamic loads. Compared with the traditional static measurement method in the prior art, it solves the problem of insufficient adaptability of the power amplifier performance under dynamic loads.

[0017] 2. The present invention introduces a predistortion compensation signal generation method with a quantum phase compensation term and a dynamic phase transition compensation term, achieving the technical effect of improving the linearity of the power amplifier. Compared with the single compensation method in the prior art, it solves the problem that it cannot fully consider different non-linear influencing factors, and realizes more accurate compensation.

[0018] 3. The present invention dynamically optimizes the predistortion parameters through the quantum annealing algorithm, constructs an objective function to realize the weighted combination of the error vector magnitude and the entropy production rate, achieving the technical effect of efficiently optimizing the predistortion compensation signal. Compared with the traditional optimization method in the prior art, it solves the problems of slow convergence speed and insufficient accuracy, and the optimization effect is more refined.

[0019] 4. The present invention adopts the eutectic welding technology of an integrated thermoelectric coupling sensitive layer and a power amplifier chip, achieving the technical effects of efficient thermal management and temperature monitoring. Compared with the separated temperature sensing method in the prior art, it solves the problem of difficulty in achieving high-precision and real-time temperature monitoring, and improves the stability and reliability of the system. Brief Description of the Drawings

[0020] Figure 1 It is a schematic flow chart of the method of the present invention. Detailed Embodiment

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0022] Please refer to the attached Figure 1 , the embodiments of the present invention provide a digital predistortion method for improving the linearity of a power amplifier in a phased array system, including: In a phased array system, the non-linear characteristics of a power amplifier (PA) are mainly caused by thermal effects and quantum tunneling effects. To achieve precise compensation for PA non-linear distortion, it is first necessary to obtain its non-linear parameters in real time. These parameters include the amplitude-phase characteristics of the input signal, temperature parameters, and quantum tunneling current parameters. Through high-precision sensors and signal processing technologies, the dynamic non-linear behavior of the PA can be accurately captured, providing data support for subsequent predistortion compensation.

[0023] In this embodiment, the acquisition of non-linear parameters is mainly achieved through the following methods: Measurement of quantum tunneling current parameters The gate leakage current of the PA is closely related to the quantum tunneling effect. A high-precision current sensor (such as the Keysight B2900 series) can be used to collect the leakage current signal in real time. Specifically, the current sensor is connected to the PA gate circuit, and the sampling rate is set to 1 MHz to ensure that transient changes can be captured.

[0024] The collected leakage current signal is low-pass filtered, and the cut-off frequency is set to 100 kHz to remove high-frequency noise. The quantum tunneling current J t (t) is calculated by the formula: Among them, e is the electron charge, h is the Planck constant, f L (E) and f R(E) are the Fermi distribution functions on the left and right sides respectively, T(E, x) is the tunneling probability, E is the energy, and x is the spatial coordinate.

[0025] The PA junction temperature can be measured in real time through a thermoelectrically coupled sensitive layer (QTSL). Specifically, the QTSL consists of an AlGaN / GaN superlattice, a boron nitride (h-BN) tunneling barrier layer, and a doped zinc oxide (Sc:ZnO) thermoelectric conversion layer. The output voltage V of the QTSL th and the junction temperature T j are related as follows: V th = S(T)·ΔT; where S(T) is the Seebeck coefficient and ΔT is the local temperature rise.

[0026] Use a high-precision temperature acquisition module (such as ADI ADT7320) to read the output voltage of the QTSL and calculate the local temperature rise ΔT based on the temperature gradient model. The specific form of the temperature gradient model is: where S(T) = S0 + k S ·T 3 / 2 , S0 is a constant, and k S is the temperature coefficient.

[0027] The amplitude-phase characteristics of the input signal can be acquired in real time through a high-speed ADC (such as ADI AD9208). Use quadrature demodulation technology to extract the amplitude A(t) and phase φ(t) of the signal. Decompose the input signal x(t) into an in-phase component I(t) and a quadrature component Q(t): I(t) = A(t)·cos(φ(t)); Q(t) = A(t)·sin(φ(t)); where, A(t) is the amplitude of the signal, φ(t) is the phase of the signal, I(t) is the in-phase component, and Q(t) is the quadrature component.

[0028] Input the quantum tunneling current parameters, temperature parameters, and the amplitude-phase characteristics of the input signal into a pre-trained model (such as an LSTM network) to generate a mapping relationship of the non-linear parameters. The input layer of the pre-trained model receives the above parameters, the hidden layer performs feature extraction, and the output layer generates a non-linear distortion prediction value.

[0029] The acquisition of non-linear parameters can also be enhanced in the following ways: Multi-sensor fusion: Combine an infrared thermal imager and a current sensor to improve the measurement accuracy of the temperature gradient and the quantum tunneling current.

[0030] Adaptive sampling rate: Dynamically adjust the sampling rate according to the PA operating state to increase the data acquisition frequency under high temperature or high power conditions.

[0031] The acquisition of non - linear parameters may also include: Noise suppression technology: Using wavelet transform to denoise the acquired signal to improve the accuracy of parameter extraction.

[0032] Real - time calibration: Calibrating the sensor in real - time through the feedback signal to eliminate measurement errors.

[0033] Through the above - mentioned method, the non - linear parameters of the PA can be obtained in real - time and accurately, providing reliable data support for subsequent predistortion compensation. Experiments show that after adopting this method, the EVM of the PA drops from 8.5% to 1.9%, and the ACPR is stable at - 68 dBc in the frequency band of 94 ± 2 GHz, significantly improving the system performance.

[0034] After obtaining the non - linear parameters of the power amplifier (PA), it is necessary to generate a predistortion compensation signal based on these parameters to cancel the non - linear distortion of the PA. This step constructs a composite compensation signal containing a quantum phase compensation term and a dynamic phase transition compensation term by combining the quantum tunneling effect and the dynamic phase transition characteristics. The generation of the compensation signal needs to respond in real - time to the temperature change of the PA, the input signal power, and the transient characteristics of the quantum tunneling current to ensure that the predistortion effect matches the dynamic characteristics of the PA.

[0035] In this embodiment, the generation of the predistortion compensation signal is realized in the following way: The phase mutation caused by the quantum tunneling current is the main source of high - frequency signal distortion. As an option, the quantum tunneling current parameter J t (t) and the temperature gradient parameter ΔT collected in real - time are input into the pre - trained model to generate the phase compensation amount. Specifically, the pre - trained model adopts a bidirectional long short - term memory (BiLSTM) network architecture.

[0036] The input layer of the BiLSTM network receives J t (t) and ΔT at the current moment and the previous M historical moments, where M is dynamically determined by the thermal relaxation time constant τ(T): In the formula, f s is the signal sampling rate, τ0 is the relaxation time at room temperature, E a is the activation energy, k B is the Boltzmann constant, and T is the current junction temperature.

[0037] The output layer of the network generates the quantum phase compensation term Δφ(t), and its expression is: Among them, α is the tunneling - thermal coupling coefficient, τ Qis the thermal relaxation time, n0 is the equilibrium carrier concentration, μ is the carrier mobility, and P crit is the critical power of dynamic phase transition.

[0038] When the input signal power approaches the phase transition critical point P crit , the dynamic phase transition compensation term Δx(t) needs to be triggered. Specifically, P crit is calculated by the formula:

[0039] In the formula, α(T), β(T), and γ(T) are the temperature-dependent nonlinear coefficients, respectively, which are obtained by S-parameter measurement fitting.

[0040] In a possible implementation, the generation of the dynamic phase transition compensation term Δx(t) adopts a memory polynomial model: where c k,m (T) is the temperature-dependent polynomial coefficient, K is the nonlinear order, and M is the memory depth.

[0041] Generally, the quantum phase compensation term Δφ(t) and the dynamic phase transition compensation term Δx(t) are superimposed on the original input signal x(t) to generate the pre-distorted modulation signal x pd (t). Specifically: x pd (t) = x(t) · e jΔφ(t) + Δx(t); The high-speed digital-to-analog converter (DAC) is used to convert x pd (t) into an analog signal and load it to the input end of the PA through an IQ modulator.

[0042] In some embodiments, the generation of the pre-distortion compensation signal can also be optimized by the following methods: Multi-model fusion: Combine BiLSTM and convolutional neural network (CNN), and use CNN to extract the time-frequency domain features of the input signal to enhance the spatio-temporal correlation of the compensation term.

[0043] Adaptive order adjustment: Dynamically adjust the nonlinear order K according to the peak-to-average power ratio (PAPR) of the input signal, and increase the high-order term compensation in the high PAPR scenario.

[0044] The pre-distortion compensation signal generated by the above method can accurately cancel the nonlinear distortion of the PA. Experiments show that in the 5G NR 100 MHz bandwidth scenario, after adopting this method, the EVM of the PA is reduced from 7.2% to 1.7%, and the ACPR is improved from -45 dBc to -65 dBc.

[0045] After generating the predistortion compensation signal, it is necessary to synthesize it with the original input signal to form a predistortion modulation signal. This signal compensates for the nonlinear characteristics of the PA to ensure the high fidelity of the output signal. This step involves key technologies such as signal superposition, digital-to-analog conversion, and RF modulation, and needs to cooperate closely with the aforementioned compensation signal generation module to achieve low-latency and high-precision signal processing.

[0046] Compound and superimpose the quantum phase compensation term Δφ(t) and the dynamic phase transition compensation term Δx(t) with the original input signal x(t). Specifically, the expression of the predistortion modulation signal x pd (t) is: x pd (t) = x(t)·e jΔφ(t) +Δx(t); In the formula, x(t) is the original baseband signal, Δφ(t) is the quantum phase compensation term, Δx(t) is the dynamic phase transition compensation term, and j is the imaginary unit.

[0047] The superposition operation is completed by a digital signal processor (DSP). A hybrid-precision computing architecture is adopted: Use 16-bit fixed-point arithmetic for the phase compensation term Δφ(t); Use 32-bit floating-point arithmetic for the dynamic phase transition compensation term Δx(t).

[0048] Convert the digital predistortion signal x pd (t) into an analog signal. Specifically, a high-speed digital-to-analog converter (DAC), such as ADI AD9164, is used, whose sampling rate is not less than 4 GS / s and the resolution is ≥14 bits.

[0049] Load the baseband signal onto the RF carrier through an IQ modulator (such as HMC6300). The local oscillator frequency f LO of the IQ modulator is consistent with the operating frequency band of the PA. The specific formula is: where c is the speed of light and λ is the signal wavelength.

[0050] The signal processing delay needs to be strictly controlled within 1 μs. Specifically, a real-time scheduling module is deployed in the FPGA, and the computing path is optimized through hardware pipelining technology.

[0051] Adopt a clock synchronization circuit (such as ADI HMC7044) to ensure the phase alignment of the clocks of the DAC, IQ modulator, and PA driver. The clock jitter needs to be lower than 100 fs to reduce phase noise.

[0052] The generation of the predistortion modulation signal can also be enhanced in the following ways: Multi-channel synchronization: In a phased array multi-PA system, a global clock distribution network is adopted to ensure the phase consistency of the modulation signals of each channel.

[0053] Adaptive gain adjustment: Dynamically adjust the gain coefficient of the DAC according to the PA output power to avoid signal overload.

[0054] The RF modulation process may further include: Harmonic suppression: A band-pass filter (such as Mini-Circuits VBF-1445+) is cascaded after the IQ modulator to suppress high-order harmonic components; Temperature compensation: Based on the junction temperature T measured by QTSL j , the local oscillator frequency f is adjusted in real time LO to compensate for thermal drift.

[0055] The pre-distorted modulation signal generated by the above method can effectively cancel the nonlinear distortion of the PA. Experiments show that in the 5G NR, 100 MHz bandwidth scenario, the EVM of the PA is reduced from 7.2% to 1.7%, and the ACPR is improved from -45 dBc to -65 dBc.

[0056] Dynamic optimization and feedback are the key links to ensure that the pre-distortion system continuously adapts to the changes in the nonlinear characteristics of the power amplifier (PA). Based on the real-time feedback signal, online adjustment of the pre-distortion parameters can effectively cope with dynamic scenarios such as temperature drift and sudden change of load impedance. This step realizes the dynamic update of the pre-distortion model parameters through multi-modal feedback data fusion and quantum optimization algorithm, ensuring the long-term stability of the system.

[0057] In this embodiment, dynamic optimization and feedback are implemented in the following manner: The feedback signal includes two modes: direct connection feedback and air interface feedback. As an option, the direct connection feedback collects the PA output signal through a directional coupler (such as Mini-Circuits ZFDC-20-5). The air interface feedback uses the receiving end beamforming technology to separate the target channel signal.

[0058] Perform signal-to-noise ratio (SNR) weighted fusion on the two feedback signals: w 直连 = 1 - w OTA ; In the formula, SNR OTA is the signal-to-noise ratio of the air interface feedback signal, SNR 直连 is the signal-to-noise ratio of the direct connection feedback signal, w OTA is the ratio of the signal-to-noise ratio of the air interface feedback signal to the sum of the signal-to-noise ratios of the air interface and direct connection signals, w 直连 represents the weight coefficient of the direct connection feedback signal. The fused feedback signal y 融合(t) is: y 融合 (t) = w OTA ·y OTA (t) + w 直连 ·y 直连 (t); Among them, y OTA (t) is the air interface feedback signal, y 直连 is the direct connection feedback signal, y 融合 (t) is the fused feedback signal.

[0059] The optimization objective function takes into account both the distortion index and the thermodynamic stability. Specifically, the objective function F is defined as: In the formula, EVM is the error vector magnitude, is the time derivative of the entropy production rate, and λ is the weight coefficient (the value range is 0.1 to 1.0). The entropy production rate The calculation formula is: Among them, J Q is the heat flux density, is the temperature gradient, J t is the tunneling current density, E is the electric field strength, and T is the local temperature.

[0060] The quantum annealing algorithm is used to solve the minimum value of the objective function. In a possible implementation, the optimization problem is mapped to a quadratic unconstrained binary optimization (QUBO) model: In the formula, is the Pauli Z operator, h i and J ij are the local field strength and the coupling coefficient respectively, which are calculated through the feedback signal error.

[0061] After obtaining the optimal parameters through quantum annealing the updated parameters are sent to the FPGA in real time. The PCIeGen4 bus is used for data transmission to ensure that the update delay is less than 5ms.

[0062] The dynamic optimization process can also be enhanced in the following ways: Multi-modal data caching: Cache recent feedback data (such as a 100ms time window) in the edge computing unit for incremental training; Noise robustness processing: Perform wavelet threshold denoising on the feedback signal to suppress the influence of environmental interference on the optimization process.

[0063] The entropy production rate calculation can also include: Infrared thermal imaging assistance: By integrating an infrared sensor (such as FLIRA655sc), the temperature field distribution on the PA surface is obtained to improve the calculation accuracy; Carrier density inversion: Based on the tunneling current J t (t) and temperature T, the carrier concentration n(t) is inverted as follows: where n0 is the equilibrium carrier concentration and τ(T) is the temperature-dependent relaxation time.

[0064] Through dynamic optimization and feedback mechanism, the predistortion system remains stable in the temperature range of -40°C to 85°C, and the EVM fluctuation is less than 0.5%. Experiments show that in the scenario of sudden change in load VSWR (1.5 → 4.0), the system completes parameter adaptive adjustment within 10 μs.

[0065] Example 1: This example provides a specific usage method based on 5G millimeter-wave base station applications, including the following: Test scenario and hardware configuration: Application scenario: 5G millimeter-wave base station radio frequency front end (28 GHz band, 800 MHz bandwidth).

[0066] PA module: QorvoQPD1025 (GaN process, operating frequency band 26.5 - 29.5 GHz, saturated output power 46 dBm).

[0067] Predistortion system configuration: Quantum tunneling sensitive layer (QTSL): Size 2 mm × 2 mm, integrated inside the PA package; Hybrid processor: XilinxVersalHBM platform, deploying quantum annealing unit and classical logic unit; Feedback link: Rohde&SchwarzFSW85 spectrum analyzer (supporting 40 GHz real-time analysis).

[0068] Experimental steps: Step 1: System deployment and initialization Connect the QTSL and the PA chip by eutectic soldering to ensure that the thermal coupling sensitivity ≥ 0.5% / °C; Load the pre-trained BiLSTM model in the FPGA (input layer: quantum tunneling current J t (t) and temperature gradient ΔT; output layer: quantum phase compensation term Δφ(t)); Configure the silicon optical interconnection module (transmission rate 400 Gbps, bit error rate < 1e-12), and connect the feedback signal processing unit.

[0069] Step 2: Signal Generation and Predistortion Modulation Use a Keysight M8199A arbitrary waveform generator to generate a 5G NR signal (256QAM modulation, subcarrier spacing 120 kHz, PAPR = 12 dB); Collect the amplitude-phase characteristics A(t) and φ(t) of the input signal x(t) in real time, and calculate the predistortion compensation signal through the following formula: In the formula, M = 15, K = 5, c k,m (T) is dynamically optimized through a quantum annealing algorithm; Convert x pd (t) into an analog signal through an ADI AD9164 DAC (sampling rate 4 GS / s), and load it onto a 28 GHz carrier through an HMC6300 IQ modulator.

[0070] Step 3: Dynamic Optimization and Feedback Control Collect the PA output signal through a directional coupler, and directly connect the feedback signal y 直連 (t) is digitized through an ultra-wideband ADC (4 GS / s); The receiving end uses a 64-element phased array antenna to extract the air interface feedback signal y OTA (t); Weight and fuse the feedback signals according to the signal-to-noise ratio: y 融合 (t) = w OTA ·y OTA (t) + w 直连 ·y 直連 (t); Among them, The threshold range is 10 - 30 dB; Solve the optimization objective function through a quantum annealing algorithm: Update the predistortion coefficient c k,m (T), and send it to the FPGA through a PCIe Gen4 bus (update delay < 5 ms).

[0071] Temperature control: The PA junction temperature rises from 25 °C (room temperature) to 85 °C at a rate of 10 °C / min and lasts for 24 hours; Data acquisition: Record the EVM, ACPR, and junction temperature T once every hour j ; Use an infrared thermal imager (FLIR A655sc) to monitor the surface temperature field distribution of the PA; Collect the dynamic change data of the quantum tunneling current J t (t).

[0072] Process according to the above steps and record the data in the following table: Table 1. Comparison data table of linearity indicators Time (h) <![CDATA[Junction temperature T j (°C)]]> EVM Fluctuation 0 25 1.80％ 12 65 1.90％ 24 85 2.00％ Table 2. Temperature stability test data table Dynamic response of key parameters: Quantum tunneling current J t (t): Increases from 1.2 mA to 3.5 mA at high temperature, and the predistortion model adjusts Δφ(t) in real time; Entropy production rate Decreases from 0.05 W / K at 25 °C to 0.03 W / K at 85 °C, and the thermal stability is improved by 40%.

[0073] Comparison experiment: Traditional Generalized Memory Polynomial (GMP) model Index GMP Model The present invention Advantage EVM (85°C) 11.50％ 2.00％ ↓82％ Parameter Convergence Time 50 μs 5 μs ↑ 10 times Power Consumption 30W 22W ↓27％ Table 3. Comparison data table In summary, in this scenario, it supports the 28 GHz millimeter-wave band, the bandwidth ≥ 800 MHz, and has high high-frequency adaptability; at a high temperature of 85 °C, the EVM fluctuation ≤ 0.2% and the signal processing delay ≤ 0.8 μs, meeting the 3GPP delay requirement (< 1 μs).

[0074] Example 2: This example provides a specific usage method based on the application of a low-earth orbit satellite communication terminal, including the following content: Test scenario and hardware configuration: Application scenario: Radio frequency front-end of a low-earth orbit satellite communication terminal (Ka band, 2 GHz bandwidth).

[0075] PA module: Wolfspeed CGHV14800 (GaN process, operating frequency band 27.5 - 31 GHz, saturated output power 40 W).

[0076] Predistortion system configuration: Quantum tunneling sensitive layer (QTSL): Size 1.5 mm × 1.5 mm, integrated inside the PA package; Hybrid processor: Xilinx Versal ACAP platform, integrated with a quantum annealing unit and a classical logic unit; Feedback link: Keysight N9041B spectrum analyzer (supporting 44 GHz real-time analysis).

[0077] Step 1: System deployment and initialization Connect the QTSL and the PA chip by eutectic soldering to ensure that the thermal coupling sensitivity ≥ 0.5% / °C; Load the pre-trained BiLSTM model in the FPGA, input the quantum tunneling current and temperature gradient, and output the quantum phase compensation term; Configure the silicon optical interconnection module (transmission rate 400 Gbps, bit error rate < 1e-12), and connect the feedback signal processing unit.

[0078] Step 2: Signal generation and predistortion modulation Use the Keysight M8199A arbitrary waveform generator to generate satellite communication signals (QPSK modulation, bandwidth 2 GHz, PAPR = 10 dB); Collect the amplitude and phase characteristics of the input signal in real time and generate the predistortion compensation signal; Convert the digital compensation signal into an analog signal through a high-speed DAC, and load it onto the Ka-band carrier through an IQ modulator.

[0079] Step 3: Dynamic optimization and feedback control Collect the PA output signal through a directional coupler, and digitize the direct connection feedback signal through an ultra-wideband ADC; The receiving end uses a 32-element phased array antenna to extract the air interface feedback signal through beamforming technology; Fuse the feedback signals weighted by the signal-to-noise ratio and dynamically adjust the predistortion parameters; Solve the optimization objective function with the quantum annealing algorithm, update the predistortion coefficient, and send it to the FPGA through the PCIe Gen4 bus.

[0080] Test conditions and data acquisition: Temperature control: The PA junction temperature rises from -55 °C (extreme cold) to +125 °C (extreme heat) at a rate of 50 °C / min and lasts for 12 hours; Load mutation: The VSWR mutates from 1.5 to 4.0 to simulate the antenna pointing switch; Data acquisition: Record the EVM, ACPR, and junction temperature every 30 minutes; Use an infrared thermal imager to monitor the PA surface temperature field distribution; Collect the dynamic change data of the quantum tunneling current.

[0081] Test results and analysis: Run according to the above method and record the data in the following table: Table 4. Comparison data table of linearity indicators Time (h) <![CDATA[Junction temperature T j (°C)]]> EVM Fluctuation 0 -55 1.90％ 6 35 2.00％ 12 125 2.10％ Table 5. Comparison data table of linearity indicators Dynamic response of key parameters: Quantum tunneling current: increases from 1.0 mA to 3.2 mA at high temperatures, and the predistortion model adjusts the compensation term in real time; Entropy production rate: decreases from 0.06 W / K at -55 °C to 0.04 W / K at +125 °C, and the thermal stability is improved by 33%.

[0082] Comparative experiment: Traditional Generalized Memory Polynomial (GMP) model Table 6. Comparative data table In summary, within the range of -55 °C to +125 °C, the EVM fluctuation ≤ 0.2%; the recovery time ≤ 8 μs when the VSWR mutates; the signal processing delay ≤ 1 μs, meeting the satellite communication time delay requirements.

[0083] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A digital predistortion method for improving linearity of a power amplifier in a phased array system, characterized in that: The following steps are involved: Step 1: acquiring nonlinear parameters of the power amplifier under dynamic load in real time, wherein the nonlinear parameters include input signal amplitude and phase characteristics, temperature parameters, and quantum tunneling current parameters; Step 2: generating a predistortion compensation signal based on the nonlinear parameter, wherein the predistortion compensation signal includes a quantum phase compensation term and a dynamic phase change compensation term; Step 3: superimposing the predistortion compensation signal onto the original input signal to generate a predistortion modulation signal; Step 4: Dynamically optimize predistortion parameters based on fusion data of the direct connection feedback signal and the air interface feedback signal.

2. The digital predistortion method for improving the linearity of a power amplifier in a phased array system according to claim 1, characterized in that: The acquisition of the nonlinear parameters includes: Real-time measurement of quantum tunneling current parameters through power amplifier gate leakage current; The junction temperature of the power amplifier is measured through a thermoelectrically coupled sensitive layer, and the temperature gradient parameters are calculated based on a temperature gradient model; The nonlinear relationship between the quantum phase compensation term and the quantum tunneling current parameter and the temperature gradient parameter is mapped through a pre-trained model.

3. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 1, characterized in that: The dynamically optimized predistortion parameters include: Constructing an optimization objective function, wherein the objective function comprises a weighted combination of an error vector magnitude and an entropy production rate, wherein a weight coefficient λ ranges from 0.1 to 1.0; The minimum value of the objective function is solved by a quantum annealing algorithm, and the coefficient of the pre-distortion compensation term is updated.

4. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 1, characterized in that: The generating of the predistortion compensation signal comprises: An AlGaN / GaN superlattice heterojunction is used as a quantum tunneling sensitive layer, wherein the heterojunction comprises a boron nitride tunneling barrier layer and a doped zinc oxide thermoelectric conversion layer; A quantum phase compensation term is generated through the tunneling current signal and the temperature sensitive signal output by the quantum tunneling sensitive layer.

5. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 4, characterized in that: The preparation of the quantum tunneling sensitive layer comprises: Epitaxially growing an AlGaN / GaN superlattice structure on a substrate; Forming a boron nitride tunneling barrier layer on the surface of the superlattice by an atomic layer deposition process, wherein the thickness of the boron nitride layer is 1.5 nm to 3.0 nm; The doped zinc oxide thermoelectric conversion layer is deposited on the boron nitride layer by a magnetron sputtering process with a doping concentration of 3% to 8%.

6. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 1, characterized in that: The calculation architecture of the dynamic optimization pre-distortion parameters includes: A hybrid processor of quantum annealing computing units and classical logic units; The quantum annealing unit is used to solve the entropy production rate constraint optimization problem, and the classical logic unit is used to generate a pre-distortion compensation signal in real time.

7. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 6, characterized in that: The transmission rate of the silicon photonic interconnect module is 350Gbps to 450Gbps, and the bit error rate is less than 5×10 -12 , used to transmit carrier density wave data and thermal imaging data.

8. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 1, characterized in that: The acquisition of the real-time feedback signal includes: A direct feedback signal of the power amplifier output signal is collected through a directional coupler; Separate the target channel signal from the air interface feedback signal through the receiving end beamforming technology; The direct connection feedback signal and the air interface feedback signal are subjected to signal-to-noise ratio weighted fusion, the weight coefficient is dynamically adjusted according to the signal-to-noise ratio, and the signal-to-noise ratio threshold is 10 dB to 30 dB.

9. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 1, characterized in that: The generation of the predistortion modulation signal includes: 16-bit fixed-point arithmetic is used for quantum phase compensation terms, and 32-bit floating-point arithmetic is used for dynamic phase change compensation terms; The processing delay of the predistortion modulation signal is 0.5 microseconds to 1.0 microseconds.

10. The digital predistortion method for improving linearity of a power amplifier in a phased array system according to claim 1, characterized in that: A thermoelectric coupling sensitive layer is integrated in the power amplifier package. The temperature sensitivity of the sensitive layer is 0.3% / °C to 0.7% / °C. The size of the sensitive layer is 1.0mm×1.0mm to 2.5mm×2.5mm, and the sensitive layer is connected to the power amplifier chip through eutectic welding.

Images