Design method and system of high-integration-level low-power-consumption radio frequency front-end chip
By integrating key modules of RF front-end chips on a single silicon substrate, and adopting adaptive bias control system and dynamic power consumption management strategies, the balance problem of RF front-end chips between high integration, low power consumption and high performance is solved. , achieving the unity of low power consumption and high performance.
Patent Information
- Application Number
- CN202411479306.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-23
- Publication Date
- 2025-02-11
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing RF front-end chips are difficult to balance between high integration, low power consumption and high performance, and are difficult to achieve effective power consumption management in multimode and multiband support and dynamic environments.
High integration and low power consumption of the chip is achieved by integrating RF transceivers, power amplifiers, low noise amplifiers and switching modules on a single silicon substrate, and adopting an adaptive bias control system and dynamic power consumption management strategy.
It realizes the lowest power consumption operation of the RF front-end chip in different working states, improves the chip's performance and integration, and reduces power consumption and adapts to the needs of multimode and multi-bands.
Smart Images

Figure CN119420373A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to chip technology, and in particular to a design method and system for a high-integration, low-power radio frequency front-end chip. Background Art
[0002] With the rapid development of wireless communication technology and the widespread popularity of IoT applications, the demand for high-performance, low-power, and miniaturized RF front-end chips is growing. As the core component of wireless communication systems, the performance of RF front-end chips directly affects the communication quality and energy efficiency of the entire system. However, existing RF front-end chips still face many challenges in balancing integration, power consumption, and performance.
[0003] Traditional RF front-end chips are usually designed with discrete components, and each functional module such as low noise amplifier (LNA), power amplifier (PA), mixer and filter is manufactured separately and then integrated together. Although this method guarantees the performance of each module to a certain extent, it also leads to problems such as large chip size, high power consumption and high cost. With the advancement of integrated circuit technology, monolithic integrated RF front-end chips have gradually become mainstream, but how to ensure the performance of each module and reduce mutual interference while being highly integrated is still an important issue.
[0004] In terms of power consumption, traditional RF front-end chips often use fixed bias and working modes, which makes it difficult to dynamically adjust performance according to the actual working environment and signal characteristics, resulting in high power consumption even in low signal strength or standby mode. In addition, the need for multi-mode and multi-band support has also made the design of RF front-end chips more complex. How to maintain low power consumption while supporting multiple communication standards has become a major challenge.
[0005] In the existing technology, some studies have tried to improve the performance and reduce power consumption of RF front-end chips by improving circuit topology, using advanced processes and new materials, etc. For example, using SOI (silicon on insulator) process to reduce substrate loss and parasitic effects, using GaN (gallium nitride) materials to improve the efficiency of power amplifiers, etc. However, these methods often increase manufacturing costs and are still limited in improving certain performance indicators.
[0006] On the other hand, some studies have proposed adaptive bias technology and dynamic power management strategies, trying to optimize power consumption by adjusting the chip working state in real time. However, these methods usually require additional control circuits and complex algorithms, and there is a trade-off between achieving high integration and low power consumption.
[0007] In addition, with the development of 5G and future 6G communications, RF front-end chips are facing challenges of higher frequency bands, wider bandwidths, and more complex modulation methods. How to maintain high performance and low power consumption in these new application scenarios while achieving high integration has become the focus and difficulty of current research.
[0008] In summary, although RF front-end chip technology has made significant progress, there is still a lot of technical space to achieve the unity of high integration, low power consumption and high performance. How to develop a high-integration, low-power RF front-end chip that can meet the needs of future wireless communications through innovative circuit design and system-level optimization based on advanced technology is an urgent problem that technicians in this field need to solve. Summary of the invention
[0009] The embodiments of the present invention provide a design method and system for a highly integrated and low-power radio frequency front-end chip, which can solve the problems in the prior art.
[0010] According to a first aspect of the embodiments of the present invention,
[0011] Provides a design method for highly integrated, low-power RF front-end chips, including:
[0012] Integrating a radio frequency transceiver, a power amplifier, a low noise amplifier and a switch module on a single silicon substrate, wherein the radio frequency transceiver adopts a zero intermediate frequency architecture, the power amplifier adopts an envelope tracking technology, the low noise amplifier adopts a noise elimination technology, and the switch module adopts a silicon-based radio frequency switch technology; connecting the radio frequency transceiver, the power amplifier, the low noise amplifier and the switch module through an on-chip interconnection technology to form a monolithic integrated radio frequency front-end chip;
[0013] Implementing an adaptive bias control system in the monolithic integrated RF front-end chip, including setting a plurality of temperature sensors, voltage sensors and current sensors, the sensors being respectively located at key positions of the RF transceiver, power amplifier, low noise amplifier and switch module; monitoring the working status of each module of the chip in real time through the sensors, and transmitting the monitoring data to the on-chip digital signal processor; the digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through a digital-to-analog converter;
[0014] Based on the adaptive bias control system, dynamic power consumption management of the RF front-end chip is realized, specifically including: when the RF signal input is detected, the digital signal processor analyzes the signal characteristics and determines the best working mode; according to the determined working mode, the local oscillator frequency and gain of the RF transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module are dynamically adjusted; at the same time, the digital signal processor continuously optimizes the bias parameters of each module according to the real-time monitored temperature, voltage and current data to achieve the lowest power consumption operation of the chip under different working conditions; when the end of the RF signal transmission is detected, the digital signal processor controls each module to enter a low-power standby state to further reduce the overall power consumption of the chip.
[0015] In an optional embodiment,
[0016] The digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter, including:
[0017] Constructing a hybrid optimization algorithm combining a particle swarm optimization algorithm and an adaptive neural fuzzy inference system, encoding bias parameters of the RF transceiver, power amplifier, low noise amplifier and switch module of the RF front-end chip into a multi-dimensional vector, each dimension corresponding to an adjustable bias parameter;
[0018] Design a multi-objective optimization function, including minimizing total power consumption, maximizing signal quality, and maximizing thermal reliability, and use the weighted sum method to convert the multi-objective optimization function into a single-objective optimization problem; use the particle swarm optimization algorithm to perform a global search, initialize the particle swarm, each particle represents a set of possible bias parameter configurations, and update the optimal position of each particle and the global optimal position through iterative optimization; build an adaptive neural fuzzy inference system model based on chip working data, use a sub-clustering algorithm to determine the number of fuzzy rules and the initial membership function, and establish a mapping relationship between bias parameters and power consumption and performance;
[0019] Arrange temperature sensors, voltage sensors and current sensors at key positions of the RF front-end chip, collect sensor data through a low-noise and high-speed analog-to-digital conversion circuit, and use a digital filtering algorithm to remove measurement noise and interference;
[0020] A real-time operating system is implemented on a digital signal processor to process data acquisition and preprocessing tasks, power consumption optimization algorithm execution tasks, bias parameter update tasks, and communication interface tasks in parallel; the multi-core architecture of the digital signal processor is used to divide the particle swarm into multiple sub-groups, with each processor core responsible for the calculation of a sub-group, and data exchange between cores is achieved through a shared memory mechanism; an optimal bias parameter lookup table under common working conditions is established, a fast indexing algorithm and an online learning mechanism are implemented, and the lookup table content is dynamically updated;
[0021] The optimized bias parameters are converted into analog control signals through a high-precision digital-to-analog converter, an active low-pass filter is added to the output end of the digital-to-analog converter to filter out high-frequency noise, and digital pre-distortion technology is implemented to compensate for the nonlinear error of the digital-to-analog converter; a feedback sampling circuit is designed to monitor the actual bias voltage and bias current in real time, and the digital-to-analog converter output is dynamically adjusted through the PID control algorithm to ensure bias accuracy, and a self-calibration mechanism is designed to regularly correct the digital-to-analog converter output error.
[0022] In an optional embodiment,
[0023] Based on the chip working data, an adaptive neural fuzzy inference system model is built, and the sub-clustering algorithm is used to determine the number of fuzzy rules and the initial membership function, and the mapping relationship between bias parameters and power consumption and performance is established, including:
[0024] S1. Collect the working data of the RF front-end chip under different working conditions, the working data includes the bias voltage, bias current, dynamic power consumption, static power consumption, gain, noise figure and linearity parameters of each module;
[0025] S2. Preprocessing the working data, including:
[0026] S2.1 Use principal component analysis method to reduce the dimension of the working data and extract representative features;
[0027] S2.2 Use the Z-score standardization method to normalize the reduced-dimensional data to obtain a standardized data set;
[0028] S3. Constructing an adaptive neural fuzzy inference system model based on the standardized data set, comprising:
[0029] S3.1 uses an adaptive sub-clustering algorithm based on density peaks to process the standardized data set and determine the number of fuzzy rules and initial membership functions, wherein:
[0030] S3.1.1 Calculate the local density of each data point and the distance to high-density points;
[0031] S3.1.2 Automatically identify cluster centers through decision diagrams;
[0032] S3.1.3 introduces an adaptive threshold mechanism to dynamically adjust the clustering threshold according to data distribution characteristics;
[0033] S3.2 constructs an initial fuzzy rule base based on the identified cluster centers, each cluster center corresponds to a fuzzy rule;
[0034] S3.3 selects Gaussian membership function as the membership function of the input variable and initializes the model structure;
[0035] S4. The adaptive neuro-fuzzy inference system model constructed by training includes:
[0036] S4.1 adopts a hybrid learning algorithm, using the least squares method to optimize the conclusion parameters in the forward propagation stage and the gradient descent method to adjust the antecedent parameters in the backward propagation stage;
[0037] S4.2 introduces early stopping method and L2 regularization technology to prevent overfitting;
[0038] S4.3 implements an adaptive learning rate adjustment mechanism to dynamically adjust the learning rate according to changes in training error;
[0039] S5. Evaluate the performance of the trained adaptive neuro-fuzzy inference system model by using the k-fold cross-validation method to calculate the average relative error of the model on the test set;
[0040] S6. Based on the trained model, implement the feature importance analysis module, including:
[0041] S6.1 uses the SHAP value method to quantify the impact of different bias parameters on power consumption and performance indicators;
[0042] S6.2 Determine key bias parameters based on the quantification results;
[0043] S7. Integrate the trained adaptive neural fuzzy inference system model and feature importance analysis results into the digital signal processor of the RF front-end chip, including:
[0044] S7.1 Based on the results of feature importance analysis, optimize the model structure and retain the model parts related to key bias parameters;
[0045] S7.2 quantizes the optimized model parameters to adapt to the computing power of the digital signal processor;
[0046] S8. Using the integrated adaptive neural fuzzy inference system model, according to the real-time collected chip working data, perform the following steps:
[0047] S8.1 preprocesses the real-time collected working data using the same method as step S2;
[0048] S8.2 inputs the preprocessed data into the optimized model to calculate the optimal bias parameters;
[0049] S9. Apply the calculated optimal bias parameters to each module of the RF front-end chip to reduce chip power consumption and optimize performance.
[0050] In an optional embodiment,
[0051] When a radio frequency signal input is detected, the digital signal processor analyzes the signal characteristics and determines the best working mode; according to the determined working mode, dynamically adjusting the local oscillator frequency and gain of the radio frequency transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module includes:
[0052] A method combining fast Fourier transform and wavelet transform is used to detect radio frequency signals, and an adaptive threshold technology is used to dynamically adjust the detection threshold; after the radio frequency signal is detected, the characteristics of the radio frequency signal are analyzed by a multidimensional feature extraction algorithm, and the multidimensional features include center frequency, bandwidth, signal strength, signal-to-noise ratio, modulation mode, signal duration, duty cycle and spectrum shape; the multidimensional features are used as input, and a pre-trained dual Q network deep reinforcement learning model is used to determine the optimal working mode of the radio frequency front-end chip, wherein the deep reinforcement learning model improves generalization ability through pre-training and transfer learning technology;
[0053] Based on the determined optimal working mode, a multi-input multi-output neural network model is used to calculate the specific parameter settings of each module of the RF front-end chip, and the neural network model also takes into account the current environmental parameters; according to the calculated specific parameter settings, each module of the RF front-end chip is adjusted separately: the local oscillator frequency of the RF transceiver is adjusted using an adaptive frequency control system based on a fractional frequency synthesizer, and the frequency control system uses digital pre-distortion technology and a Σ-Δ modulator to optimize frequency accuracy and phase noise; the output power of the power amplifier is adjusted using dynamic bias technology combined with an improved Hilbert transform envelope detection algorithm; the noise coefficient and gain of the low-noise amplifier are adjusted through a variable feedback network and an adaptive impedance matching algorithm; and the isolation of the switch module is adjusted by precise control using RF MEMS technology and electrostatic force;
[0054] The performance indicators of each module are monitored in real time, and a self-calibration algorithm based on Kalman filtering is used to compensate for temperature drift and aging effects. The monitored performance indicators are used as input to continuously optimize parameter settings using a model predictive control algorithm, where the model predictive control algorithm considers the long-term impact of parameter adjustments.
[0055] According to a second aspect of the embodiments of the present invention,
[0056] Provides a design system for highly integrated, low-power RF front-end chips, including:
[0057] The first unit is used to integrate a radio frequency transceiver, a power amplifier, a low noise amplifier and a switch module on a single silicon substrate, wherein the radio frequency transceiver adopts a zero intermediate frequency architecture, the power amplifier adopts an envelope tracking technology, the low noise amplifier adopts a noise elimination technology, and the switch module adopts a silicon-based radio frequency switch technology; the radio frequency transceiver, the power amplifier, the low noise amplifier and the switch module are connected through an on-chip interconnection technology to form a monolithic integrated radio frequency front-end chip;
[0058] The second unit is used to implement an adaptive bias control system in the monolithic integrated RF front-end chip, including setting a plurality of temperature sensors, voltage sensors and current sensors, which are respectively located at key positions of the RF transceiver, power amplifier, low noise amplifier and switch module; the working status of each module of the chip is monitored in real time by the sensors, and the monitoring data is transmitted to the on-chip digital signal processor; the digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter;
[0059] The third unit is used to realize dynamic power consumption management of the RF front-end chip based on the adaptive bias control system, specifically including: when the RF signal input is detected, the digital signal processor analyzes the signal characteristics and determines the optimal working mode; according to the determined working mode, dynamically adjusts the local oscillator frequency and gain of the RF transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module; at the same time, the digital signal processor continuously optimizes the bias parameters of each module according to the real-time monitored temperature, voltage and current data to realize the lowest power consumption operation of the chip under different working conditions; when the end of the RF signal transmission is detected, the digital signal processor controls each module to enter a low-power standby state to further reduce the overall power consumption of the chip.
[0060] According to a third aspect of the embodiments of the present invention,
[0061] An electronic device is provided, comprising:
[0062] processor;
[0063] a memory for storing processor-executable instructions;
[0064] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.
[0065] A fourth aspect of the embodiments of the present invention is:
[0066] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented. BRIEF DESCRIPTION OF THE DRAWINGS
[0067] Figure 1 A schematic diagram of a flow chart of a design method for a highly integrated, low-power radio frequency front-end chip according to an embodiment of the present invention;
[0068] Figure 2 A schematic diagram of the structure of a design system for a highly integrated, low-power radio frequency front-end chip according to an embodiment of the present invention. DETAILED DESCRIPTION
[0069] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0070] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.
[0071] Figure 1 FIG. 1 is a flow chart of a method for designing a high-integration, low-power RF front-end chip according to an embodiment of the present invention. Figure 1 As shown, the method includes:
[0072] S101. Integrate a radio frequency transceiver, a power amplifier, a low noise amplifier and a switch module on a single silicon substrate, wherein the radio frequency transceiver adopts a zero intermediate frequency architecture, the power amplifier adopts an envelope tracking technology, the low noise amplifier adopts a noise elimination technology, and the switch module adopts a silicon-based radio frequency switch technology; connect the radio frequency transceiver, the power amplifier, the low noise amplifier and the switch module through an on-chip interconnection technology to form a monolithic integrated radio frequency front-end chip;
[0073] S102. Implementing an adaptive bias control system in the monolithic integrated RF front-end chip, including setting a plurality of temperature sensors, voltage sensors and current sensors, the sensors being respectively located at key positions of the RF transceiver, power amplifier, low noise amplifier and switch module; monitoring the working status of each module of the chip in real time through the sensors, and transmitting the monitoring data to the on-chip digital signal processor; the digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter;
[0074] S103. Based on the adaptive bias control system, dynamic power consumption management of the RF front-end chip is implemented, specifically including: when the RF signal input is detected, the digital signal processor analyzes the signal characteristics and determines the optimal working mode; according to the determined working mode, the local oscillator frequency and gain of the RF transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module are dynamically adjusted; at the same time, the digital signal processor continuously optimizes the bias parameters of each module according to the real-time monitored temperature, voltage and current data to achieve the lowest power consumption operation of the chip under different working conditions; when the end of the RF signal transmission is detected, the digital signal processor controls each module to enter a low-power standby state to further reduce the overall power consumption of the chip.
[0075] In an optional embodiment,
[0076] The digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter, including:
[0077] Constructing a hybrid optimization algorithm combining a particle swarm optimization algorithm and an adaptive neural fuzzy inference system, encoding bias parameters of the RF transceiver, power amplifier, low noise amplifier and switch module of the RF front-end chip into a multi-dimensional vector, each dimension corresponding to an adjustable bias parameter;
[0078] Design a multi-objective optimization function, including minimizing total power consumption, maximizing signal quality, and maximizing thermal reliability, and use the weighted sum method to convert the multi-objective optimization function into a single-objective optimization problem; use the particle swarm optimization algorithm to perform a global search, initialize the particle swarm, each particle represents a set of possible bias parameter configurations, and update the optimal position of each particle and the global optimal position through iterative optimization; build an adaptive neural fuzzy inference system model based on chip working data, use a sub-clustering algorithm to determine the number of fuzzy rules and the initial membership function, and establish a mapping relationship between bias parameters and power consumption and performance;
[0079] Arrange temperature sensors, voltage sensors and current sensors at key positions of the RF front-end chip, collect sensor data through a low-noise and high-speed analog-to-digital conversion circuit, and use a digital filtering algorithm to remove measurement noise and interference;
[0080] A real-time operating system is implemented on a digital signal processor to process data acquisition and preprocessing tasks, power consumption optimization algorithm execution tasks, bias parameter update tasks, and communication interface tasks in parallel; the multi-core architecture of the digital signal processor is used to divide the particle swarm into multiple sub-groups, with each processor core responsible for the calculation of a sub-group, and data exchange between cores is achieved through a shared memory mechanism; an optimal bias parameter lookup table under common working conditions is established, a fast indexing algorithm and an online learning mechanism are implemented, and the lookup table content is dynamically updated;
[0081] The optimized bias parameters are converted into analog control signals through a high-precision digital-to-analog converter, an active low-pass filter is added to the output of the digital-to-analog converter to filter out high-frequency noise, and digital pre-distortion technology is implemented to compensate for the nonlinear error of the digital-to-analog converter; a feedback sampling circuit is designed to monitor the actual bias voltage and bias current in real time, and the digital-to-analog converter output is dynamically adjusted through a PID control algorithm to ensure bias accuracy, and a self-calibration mechanism is designed to regularly correct the digital-to-analog converter output error.
[0082] In an optional embodiment,
[0083] Based on the chip working data, an adaptive neural fuzzy inference system model is built, and the sub-clustering algorithm is used to determine the number of fuzzy rules and the initial membership function, and the mapping relationship between bias parameters and power consumption and performance is established, including:
[0084] S1. Collect the working data of the RF front-end chip under different working conditions, the working data includes the bias voltage, bias current, dynamic power consumption, static power consumption, gain, noise figure and linearity parameters of each module;
[0085] S2. Preprocessing the working data, including:
[0086] S2.1 Use principal component analysis method to reduce the dimension of the working data and extract representative features;
[0087] S2.2 Use the Z-score standardization method to normalize the reduced-dimensional data to obtain a standardized data set;
[0088] S3. Constructing an adaptive neural fuzzy inference system model based on the standardized data set, comprising:
[0089] S3.1 uses an adaptive sub-clustering algorithm based on density peaks to process the standardized data set and determine the number of fuzzy rules and initial membership functions, wherein:
[0090] S3.1.1 Calculate the local density of each data point and the distance to high-density points;
[0091] S3.1.2 Automatically identify cluster centers through decision diagrams;
[0092] S3.1.3 introduces an adaptive threshold mechanism to dynamically adjust the clustering threshold according to data distribution characteristics;
[0093] S3.2 constructs an initial fuzzy rule base based on the identified cluster centers, each cluster center corresponds to a fuzzy rule;
[0094] S3.3 selects Gaussian membership function as the membership function of the input variable and initializes the model structure;
[0095] S4. The adaptive neuro-fuzzy inference system model constructed by training includes:
[0096] S4.1 adopts a hybrid learning algorithm, using the least squares method to optimize the conclusion parameters in the forward propagation stage and the gradient descent method to adjust the antecedent parameters in the backward propagation stage;
[0097] S4.2 introduces early stopping method and L2 regularization technology to prevent overfitting;
[0098] S4.3 implements an adaptive learning rate adjustment mechanism to dynamically adjust the learning rate according to changes in training error;
[0099] S5. Evaluate the performance of the trained adaptive neuro-fuzzy inference system model by using the k-fold cross-validation method to calculate the average relative error of the model on the test set;
[0100] S6. Based on the trained model, implement the feature importance analysis module, including:
[0101] S6.1 uses the SHAP value method to quantify the impact of different bias parameters on power consumption and performance indicators;
[0102] S6.2 Determine key bias parameters based on the quantification results;
[0103] S7. Integrate the trained adaptive neural fuzzy inference system model and feature importance analysis results into the digital signal processor of the RF front-end chip, including:
[0104] S7.1 Based on the results of feature importance analysis, optimize the model structure and retain the model parts related to key bias parameters;
[0105] S7.2 quantizes the optimized model parameters to adapt to the computing power of the digital signal processor;
[0106] S8. Using the integrated adaptive neural fuzzy inference system model, according to the real-time collected chip working data, perform the following steps:
[0107] S8.1 preprocesses the real-time collected working data using the same method as step S2;
[0108] S8.2 inputs the preprocessed data into the optimized model to calculate the optimal bias parameters;
[0109] S9. Apply the calculated optimal bias parameters to each module of the RF front-end chip to reduce chip power consumption and optimize performance.
[0110] In an optional embodiment,
[0111] When a radio frequency signal input is detected, the digital signal processor analyzes the signal characteristics and determines the best working mode; according to the determined working mode, dynamically adjusting the local oscillator frequency and gain of the radio frequency transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module includes:
[0112] A method combining fast Fourier transform and wavelet transform is used to detect radio frequency signals, and an adaptive threshold technology is used to dynamically adjust the detection threshold; after the radio frequency signal is detected, the characteristics of the radio frequency signal are analyzed by a multidimensional feature extraction algorithm, and the multidimensional features include center frequency, bandwidth, signal strength, signal-to-noise ratio, modulation mode, signal duration, duty cycle and spectrum shape; the multidimensional features are used as input, and a pre-trained dual Q network deep reinforcement learning model is used to determine the optimal working mode of the radio frequency front-end chip, wherein the deep reinforcement learning model improves generalization ability through pre-training and transfer learning technology;
[0113] Based on the determined optimal working mode, a multi-input multi-output neural network model is used to calculate the specific parameter settings of each module of the RF front-end chip, and the neural network model also takes into account the current environmental parameters; according to the calculated specific parameter settings, each module of the RF front-end chip is adjusted separately: the local oscillator frequency of the RF transceiver is adjusted using an adaptive frequency control system based on a fractional frequency synthesizer, and the frequency control system uses digital pre-distortion technology and a Σ-Δ modulator to optimize frequency accuracy and phase noise; the output power of the power amplifier is adjusted using dynamic bias technology combined with an improved Hilbert transform envelope detection algorithm; the noise coefficient and gain of the low-noise amplifier are adjusted through a variable feedback network and an adaptive impedance matching algorithm; and the isolation of the switch module is adjusted by precise control using RF MEMS technology and electrostatic force;
[0114] The performance indicators of each module are monitored in real time, and a self-calibration algorithm based on Kalman filtering is used to compensate for temperature drift and aging effects. The monitored performance indicators are used as input to continuously optimize parameter settings using a model predictive control algorithm, where the model predictive control algorithm considers the long-term impact of parameter adjustments.
[0115] Figure 2 FIG. 1 is a schematic diagram of a design system for a highly integrated, low-power radio frequency front-end chip according to an embodiment of the present invention. Figure 2 As shown, the system comprises:
[0116] The first unit is used to integrate a radio frequency transceiver, a power amplifier, a low noise amplifier and a switch module on a single silicon substrate, wherein the radio frequency transceiver adopts a zero intermediate frequency architecture, the power amplifier adopts an envelope tracking technology, the low noise amplifier adopts a noise elimination technology, and the switch module adopts a silicon-based radio frequency switch technology; the radio frequency transceiver, the power amplifier, the low noise amplifier and the switch module are connected through an on-chip interconnection technology to form a monolithic integrated radio frequency front-end chip;
[0117] The second unit is used to implement an adaptive bias control system in the monolithic integrated RF front-end chip, including setting a plurality of temperature sensors, voltage sensors and current sensors, which are respectively located at key positions of the RF transceiver, power amplifier, low noise amplifier and switch module; the working status of each module of the chip is monitored in real time by the sensors, and the monitoring data is transmitted to the on-chip digital signal processor; the digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter;
[0118] The third unit is used to realize dynamic power consumption management of the RF front-end chip based on the adaptive bias control system, specifically including: when the RF signal input is detected, the digital signal processor analyzes the signal characteristics and determines the optimal working mode; according to the determined working mode, dynamically adjusts the local oscillator frequency and gain of the RF transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module; at the same time, the digital signal processor continuously optimizes the bias parameters of each module according to the real-time monitored temperature, voltage and current data to realize the lowest power consumption operation of the chip under different working conditions; when the end of the RF signal transmission is detected, the digital signal processor controls each module to enter a low-power standby state to further reduce the overall power consumption of the chip.
[0119] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.
[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A design method for a highly integrated, low-power radio frequency front-end chip, characterized in that: include: Integrating a radio frequency transceiver, a power amplifier, a low noise amplifier and a switch module on a single silicon substrate, wherein the radio frequency transceiver adopts a zero intermediate frequency architecture, the power amplifier adopts an envelope tracking technology, the low noise amplifier adopts a noise elimination technology, and the switch module adopts a silicon-based radio frequency switch technology; connecting the radio frequency transceiver, the power amplifier, the low noise amplifier and the switch module through an on-chip interconnection technology to form a monolithic integrated radio frequency front-end chip; Implementing an adaptive bias control system in the monolithic integrated RF front-end chip, including setting a plurality of temperature sensors, voltage sensors and current sensors, the sensors being respectively located at key positions of the RF transceiver, power amplifier, low noise amplifier and switch module; monitoring the working status of each module of the chip in real time through the sensors, and transmitting the monitoring data to the on-chip digital signal processor; The digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter; Based on the adaptive bias control system, dynamic power consumption management of the RF front-end chip is realized, specifically including: when the RF signal input is detected, the digital signal processor analyzes the signal characteristics and determines the best working mode; according to the determined working mode, the local oscillator frequency and gain of the RF transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module are dynamically adjusted; at the same time, the digital signal processor continuously optimizes the bias parameters of each module according to the real-time monitored temperature, voltage and current data to achieve the lowest power consumption operation of the chip under different working conditions; when the end of the RF signal transmission is detected, the digital signal processor controls each module to enter a low-power standby state to further reduce the overall power consumption of the chip.
2. The method according to claim 1, characterized in that: The digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter, including: Constructing a hybrid optimization algorithm combining a particle swarm optimization algorithm and an adaptive neural fuzzy inference system, encoding bias parameters of the RF transceiver, power amplifier, low noise amplifier and switch module of the RF front-end chip into a multi-dimensional vector, each dimension corresponding to an adjustable bias parameter; Design a multi-objective optimization function, including minimizing total power consumption, maximizing signal quality, and maximizing thermal reliability, and use the weighted sum method to convert the multi-objective optimization function into a single-objective optimization problem; use the particle swarm optimization algorithm to perform a global search, initialize the particle swarm, each particle represents a set of possible bias parameter configurations, and update the optimal position of each particle and the global optimal position through iterative optimization; build an adaptive neural fuzzy inference system model based on chip working data, use a sub-clustering algorithm to determine the number of fuzzy rules and the initial membership function, and establish a mapping relationship between bias parameters and power consumption and performance; Arrange temperature sensors, voltage sensors and current sensors at key positions of the RF front-end chip, collect sensor data through a low-noise and high-speed analog-to-digital conversion circuit, and use a digital filtering algorithm to remove measurement noise and interference; A real-time operating system is implemented on a digital signal processor to process data acquisition and preprocessing tasks, power consumption optimization algorithm execution tasks, bias parameter update tasks, and communication interface tasks in parallel; the multi-core architecture of the digital signal processor is used to divide the particle swarm into multiple sub-groups, with each processor core responsible for the calculation of a sub-group, and data exchange between cores is achieved through a shared memory mechanism; an optimal bias parameter lookup table under common working conditions is established, a fast indexing algorithm and an online learning mechanism are implemented, and the lookup table content is dynamically updated; The optimized bias parameters are converted into analog control signals through a high-precision digital-to-analog converter, an active low-pass filter is added to the output of the digital-to-analog converter to filter out high-frequency noise, and digital pre-distortion technology is implemented to compensate for the nonlinear error of the digital-to-analog converter; a feedback sampling circuit is designed to monitor the actual bias voltage and bias current in real time, and the digital-to-analog converter output is dynamically adjusted through a PID control algorithm to ensure bias accuracy, and a self-calibration mechanism is designed to regularly correct the digital-to-analog converter output error.
3. The method according to claim 1, characterized in that Based on the chip working data, an adaptive neural fuzzy inference system model is built, and the number of fuzzy rules and the initial membership function are determined using a sub-clustering algorithm. The mapping relationship between bias parameters and power consumption and performance is established, including: S1. Collect the working data of the RF front-end chip under different working conditions, the working data includes the bias voltage, bias current, dynamic power consumption, static power consumption, gain, noise figure and linearity parameters of each module; S2. Preprocessing the working data, including: S2.1 Use principal component analysis method to reduce the dimension of the working data and extract representative features; S2.2 Use the Z-score standardization method to normalize the reduced-dimensional data to obtain a standardized data set; S3. Constructing an adaptive neural fuzzy inference system model based on the standardized data set, comprising: S3.1 uses an adaptive sub-clustering algorithm based on density peaks to process the standardized data set and determine the number of fuzzy rules and initial membership functions, wherein: S3.1.1 Calculate the local density of each data point and the distance to high-density points; S3.1.2 Automatically identify cluster centers through decision diagrams; S3.1.3 introduces an adaptive threshold mechanism to dynamically adjust the clustering threshold according to data distribution characteristics; S3.2 constructs an initial fuzzy rule base based on the identified cluster centers, where each cluster center corresponds to a fuzzy rule; S3.3 selects Gaussian membership function as the membership function of the input variable and initializes the model structure; S4. The adaptive neuro-fuzzy inference system model constructed by training includes: S4.1 adopts a hybrid learning algorithm, using the least squares method to optimize the conclusion parameters in the forward propagation stage and the gradient descent method to adjust the antecedent parameters in the backward propagation stage; S4.2 introduces early stopping method and L2 regularization technology to prevent overfitting; S4.3 implements an adaptive learning rate adjustment mechanism to dynamically adjust the learning rate according to changes in training error; S5. Evaluate the performance of the trained adaptive neuro-fuzzy inference system model by using the k-fold cross-validation method to calculate the average relative error of the model on the test set; S6. Based on the trained model, implement the feature importance analysis module, including: S6.1 uses the SHAP value method to quantify the impact of different bias parameters on power consumption and performance indicators; S6.2 Determine key bias parameters based on the quantification results; S7. Integrate the trained adaptive neural fuzzy inference system model and feature importance analysis results into the digital signal processor of the RF front-end chip, including: S7.1 Based on the results of feature importance analysis, optimize the model structure and retain the model parts related to key bias parameters; S7.2 quantizes the optimized model parameters to adapt to the computing power of the digital signal processor; S8. Using the integrated adaptive neural fuzzy inference system model, according to the real-time collected chip working data, perform the following steps: S8.1 preprocesses the real-time collected working data using the same method as step S2; S8.2 inputs the preprocessed data into the optimized model to calculate the optimal bias parameters; S9. Apply the calculated optimal bias parameters to each module of the RF front-end chip to reduce chip power consumption and optimize performance.
4. The method according to claim 1, characterized in that: When a radio frequency signal input is detected, the digital signal processor analyzes the signal characteristics and determines the best working mode; according to the determined working mode, dynamically adjusting the local oscillator frequency and gain of the radio frequency transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module includes: A method combining fast Fourier transform and wavelet transform is used to detect radio frequency signals, and an adaptive threshold technology is used to dynamically adjust the detection threshold; after the radio frequency signal is detected, the characteristics of the radio frequency signal are analyzed by a multidimensional feature extraction algorithm, and the multidimensional features include center frequency, bandwidth, signal strength, signal-to-noise ratio, modulation mode, signal duration, duty cycle and spectrum shape; the multidimensional features are used as input, and a pre-trained dual Q network deep reinforcement learning model is used to determine the optimal working mode of the radio frequency front-end chip, wherein the deep reinforcement learning model improves generalization ability through pre-training and transfer learning technology; Based on the determined optimal working mode, a multi-input multi-output neural network model is used to calculate the specific parameter settings of each module of the RF front-end chip, and the neural network model also takes into account the current environmental parameters; according to the calculated specific parameter settings, each module of the RF front-end chip is adjusted separately: the local oscillator frequency of the RF transceiver is adjusted using an adaptive frequency control system based on a fractional frequency synthesizer, and the frequency control system uses digital pre-distortion technology and a Σ-Δ modulator to optimize frequency accuracy and phase noise; the output power of the power amplifier is adjusted using dynamic bias technology combined with an improved Hilbert transform envelope detection algorithm; the noise coefficient and gain of the low-noise amplifier are adjusted through a variable feedback network and an adaptive impedance matching algorithm; and the isolation of the switch module is adjusted by precise control using RF MEMS technology and electrostatic force; The performance indicators of each module are monitored in real time, and a self-calibration algorithm based on Kalman filtering is used to compensate for temperature drift and aging effects. The monitored performance indicators are used as input to continuously optimize parameter settings using a model predictive control algorithm, where the model predictive control algorithm considers the long-term impact of parameter adjustments.
5. A design system for a highly integrated, low-power radio frequency front-end chip, used to implement the method described in any one of claims 1 to 4, characterized in that: include: The first unit is used to integrate a radio frequency transceiver, a power amplifier, a low noise amplifier and a switch module on a single silicon substrate, wherein the radio frequency transceiver adopts a zero intermediate frequency architecture, the power amplifier adopts an envelope tracking technology, the low noise amplifier adopts a noise elimination technology, and the switch module adopts a silicon-based radio frequency switch technology; the radio frequency transceiver, the power amplifier, the low noise amplifier and the switch module are connected through an on-chip interconnection technology to form a monolithic integrated radio frequency front-end chip; The second unit is used to implement an adaptive bias control system in the monolithic integrated RF front-end chip, including setting a plurality of temperature sensors, voltage sensors and current sensors, which are respectively located at key positions of the RF transceiver, power amplifier, low noise amplifier and switch module; the working status of each module of the chip is monitored in real time through the sensors, and the monitoring data is transmitted to the on-chip digital signal processor; The digital signal processor calculates the optimal bias parameters of each module according to a preset power consumption optimization algorithm, and sends the bias control signal to the corresponding module through the digital-to-analog converter; The third unit is used to realize dynamic power consumption management of the RF front-end chip based on the adaptive bias control system, specifically including: when the RF signal input is detected, the digital signal processor analyzes the signal characteristics and determines the optimal working mode; according to the determined working mode, dynamically adjusts the local oscillator frequency and gain of the RF transceiver, the output power of the power amplifier, the noise figure and gain of the low noise amplifier, and the isolation of the switch module; at the same time, the digital signal processor continuously optimizes the bias parameters of each module according to the real-time monitored temperature, voltage and current data to realize the lowest power consumption operation of the chip under different working conditions; when the end of the RF signal transmission is detected, the digital signal processor controls each module to enter a low-power standby state to further reduce the overall power consumption of the chip.
6. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 4.
7. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 4 is implemented.
Citation Information
Cited By
Power amplifier system with low-power-consumption dynamic optimization matching network
CN120342343A
Multi-band adaptive control system for radio frequency front-end module
CN120896598A
A multi-band adaptive control system for radio frequency front-end modules
CN120896598B
Power EPDT wireless private network radio frequency optimization method and system based on deep learning
CN121194227A
Deep learning-based electric power departmental trunking radio frequency optimization method and system
CN121194227B
