Search

CN-122026837-A - Energy-saving E-type power amplifier-based energy efficiency optimization method

CN122026837ACN 122026837 ACN122026837 ACN 122026837ACN-122026837-A

Abstract

The invention discloses an energy-saving E-class power amplifier energy efficiency optimization method which comprises the steps of extracting time-frequency characteristics of an audio signal and dynamic impedance characteristics of a load to construct a dynamic working condition state space, identifying a high-priority period with optimized efficiency through an attention network, respectively performing rapid and global optimization on bias voltage, supply voltage and switching time sequence by adopting a hierarchical asynchronous particle swarm algorithm, synthesizing optimization parameters, correcting, updating a predistortion compensation table and outputting a control signal. The invention realizes the accurate tracking and quick response to the double dynamic changes of the audio signal envelope and the load impedance, and obviously improves the average energy efficiency of the power amplifier in the full working condition range.

Inventors

  • YAN WENCHANG

Assignees

  • 广州市科昱音响设备有限公司

Dates

Publication Date
20260512
Application Date
20260202

Claims (8)

  1. 1. The energy efficiency optimization method based on the energy-saving E-class power amplifier is characterized by comprising the following steps of: S1, sampling and frequency band decomposition are carried out on an input audio signal to obtain a plurality of sub-band signals, characteristics of the sub-band signals are extracted through a plurality of transform encoders, and time-frequency characteristic vectors are generated through cross-band attention fusion; S2, injecting a perturbation signal in a low-sensitivity period of the audio signal, collecting the response of the output end of the power amplifier, and analyzing the response through a transducer decoder to obtain a load characteristic vector; s3, constructing and updating a dynamic working condition state space containing power, impedance and junction temperature parameters based on the time-frequency characteristic vector and the load characteristic vector; s4, inputting a dynamic working condition state space sequence into an attention network, calculating an efficiency sensitivity score, marking a high-priority time period according to the score and the efficiency gradient, and generating an optimized priority chart; S5, in a high-priority period, optimizing the bias voltage by using a first particle swarm algorithm to obtain a first parameter set, and simultaneously, optimizing the parameter set containing the power supply voltage and the switching time sequence by using a second particle swarm algorithm in a fixed period to obtain a second parameter set; S6, synthesizing a first parameter set and a second parameter set through a decoupling mapping matrix, and generating a final control parameter set by combining temperature compensation factor correction; and S7, updating a predistortion compensation table based on the final control parameter set, compensating the input signal, and outputting the output signal to the power amplifier together with the control parameter.
  2. 2. The energy-saving class-E power amplifier energy efficiency optimization method according to claim 1, wherein the S1 specifically comprises: s11, carrying out analog-to-digital conversion sampling on an input audio signal to obtain a digital audio sequence; S12, carrying out wavelet transformation decomposition on the digital audio sequence to obtain a plurality of sub-band signals covering the complete audio frequency band; S13, inputting each sub-band signal into an independent transducer encoder for encoding, and outputting a corresponding sub-band characteristic vector by each transducer encoder; S14, inputting all the sub-band feature vectors into a cross-band attention fusion layer, wherein the cross-band attention fusion layer performs the following operations of splicing all the sub-band feature vectors, performing linear transformation and attention weight calculation on the spliced vectors, and performing weighted fusion on all the sub-band feature vectors according to the calculated attention weight; and S15, mapping the weighted and fused vector into a time-frequency characteristic vector through a linear layer.
  3. 3. The energy-saving class-E power amplifier energy efficiency optimization method according to claim 1, wherein the step S2 specifically comprises: s21, monitoring the waveform of an input audio signal in real time, and identifying and positioning the zero crossing point and the peak value point of the signal amplitude of the signal; S22, determining a low sensitivity period between adjacent zero crossing points and peak points, and generating a sinusoidal perturbation signal in the low sensitivity period; S23, superposing the sinusoidal perturbation signal and the input audio signal to form a composite signal containing the perturbation and inputting the composite signal into a power amplifier; s24, synchronously collecting a voltage signal and a current signal at the output end of the power amplifier, and respectively serving as a voltage response and a current response; s25, inputting the voltage response and the current response into a pre-trained transducer decoder, and analyzing the frequency domain impedance characteristic of the load by the transducer decoder based on the time sequence relation of the voltage response and the current response; the transducer decoder outputs a load eigenvector comprising complex impedance magnitude and phase information loaded at a plurality of eigenvalues S26.
  4. 4. The energy-saving class-E power amplifier energy efficiency optimization method according to claim 1, wherein the step S3 specifically comprises: s31, extracting a first characteristic sequence from the time-frequency characteristic vector, extracting a second characteristic sequence from the load characteristic vector, and aligning the first characteristic sequence and the second characteristic sequence in the time dimension; s32, calculating an average power value and a peak power value in a fixed time window in the past based on the aligned first characteristic sequence; S33, calculating the peak-to-average ratio of the signal based on the average power value and the peak power value; S34, calculating the average impedance amplitude and the average impedance phase at a preset characteristic frequency point in a fixed time window in the past based on the aligned second characteristic sequence; s35, estimating the junction temperature of a power device in the power amplifier in real time based on the working voltage, the working current history data and the thermal resistance model of the power amplifier; S36, constructing a multidimensional state vector by taking the average power value, the peak-to-average ratio, the average impedance amplitude, the average impedance phase and the estimated junction temperature as state variables; S37, collecting new time-frequency characteristic vectors and load characteristic vectors according to preset fixed frequency to update the multidimensional state vectors to form a dynamic working condition state space which is dynamically updated.
  5. 5. The energy-saving class-E power amplifier energy efficiency optimization method according to claim 1, wherein the step S4 specifically comprises: s41, extracting a state vector sequence with fixed time length from a dynamic working condition state space; s42, inputting the state vector sequence into a gating attention network for processing; S43, outputting corresponding efficiency sensitivity scores at each moment by the gating attention network; S44, calculating an efficiency change gradient between each adjacent moment based on average power historical data in a dynamic working condition state space; s45, multiplying the efficiency sensitivity score at each moment by the efficiency change gradient at the moment to obtain a comprehensive priority score; s46, comparing the comprehensive priority score with a preset threshold, and marking the moment as a high priority period when the comprehensive priority score is larger than the preset threshold; S47, generating an optimized priority graph with time as horizontal axis and comprehensive priority score as vertical axis, and identifying high priority period in the graph.
  6. 6. The energy-saving class-E power amplifier energy efficiency optimization method according to claim 1, wherein the step S5 specifically comprises: s51, monitoring an optimization priority diagram in real time, and activating a rapid optimization loop when a system enters a high priority period at any time; S52, initializing a particle swarm of a first particle swarm algorithm in a rapid optimization loop, wherein each particle represents a bias voltage candidate value, and the position of the particle is constrained within a bias voltage working range allowed by the power amplifier; S53, calculating the fitness of each particle based on the current parameters in the dynamic working condition state space by the first particle swarm algorithm, wherein the fitness calculation takes the instantaneous energy efficiency at the current moment as a main target; s54, updating the historical optimal position of a particle unit and the global optimal position of a particle group according to the particle fitness by a first particle swarm algorithm, and iteratively updating the speeds and positions of all particles according to the historical optimal position and the global optimal position; s55, terminating iteration when the first particle swarm algorithm meets a first convergence condition, and recording a bias voltage value corresponding to the global optimal particle position as a first optimization result to form a first parameter set; S56, activating a global optimization loop in a preset fixed time period, and initializing a particle group of a second particle swarm algorithm in the global optimization loop, wherein each particle represents a multidimensional vector, and the vector dimension comprises a power supply voltage value, an on time value and an off time value; S57, calculating the fitness of each particle based on historical data of a past period in a dynamic working condition state space by a second particle swarm algorithm, wherein the fitness calculation takes the average energy efficiency in the period as a main target, and constrains the total harmonic distortion and the temperature change rate; s58, updating the historical optimal position of the particle unit and the global optimal position of the particle group according to the particle fitness by a second particle swarm algorithm, and iteratively updating the speeds and positions of all particles according to the historical optimal position and the global optimal position; And S59, stopping iteration when the second particle swarm algorithm meets a second convergence condition, and recording a power supply voltage value, an on-time value and an off-time value corresponding to the global optimal particle position as a second optimization result to form a second parameter set.
  7. 7. The energy-saving class-E power amplifier energy efficiency optimization method according to claim 1, wherein the step S6 specifically comprises: S61, acquiring a predefined decoupling mapping matrix; s62, extracting a bias voltage optimized value from a first parameter set, and extracting a power supply voltage optimized value, an on-time optimized value and an off-time optimized value from a second parameter set; s63, combining the extracted bias voltage optimized value, the power supply voltage optimized value and the start time optimized value into an input vector; S64, performing linear transformation on the input vector by using a decoupling mapping matrix to obtain an intermediate vector; S65, reading the current estimated junction temperature of the power device from a dynamic working condition state space; S66, calculating a temperature compensation factor based on the current estimated junction temperature of the power device and a preset reference junction temperature; And S67, scalar correction is carried out on the intermediate vector by using a temperature compensation factor, elements in the corrected vector are respectively determined to be a final bias voltage, a final power supply voltage and a final turn-on time, and turn-off time optimization values are determined to be final turn-off times, so that a final control parameter set is formed together.
  8. 8. The energy-saving class-E power amplifier energy efficiency optimization method according to claim 1, wherein the step S7 specifically comprises: S71, calculating a corresponding index address in an address space of a predistortion compensation table according to a final bias voltage, a final power supply voltage, a final turn-on time and a final turn-off time in a final control parameter set; S72, based on the historical input and output data of the power amplifier, establishing a query mechanism taking an index address as input, and acquiring an amplitude compensation coefficient and a phase compensation coefficient corresponding to the index address; S73, writing the amplitude compensation coefficient and the phase compensation coefficient into a storage position designated by an index address in the predistortion compensation table to finish updating the predistortion compensation table; s74, before the input audio signal enters the main path of the power amplifier, reading an amplitude compensation coefficient and a phase compensation coefficient corresponding to the current final control parameter set in a predistortion compensation table; s75, multiplying and correcting the amplitude of the input audio signal by using the read amplitude compensation coefficient, and adding and correcting the phase of the input audio signal by using the read phase compensation coefficient to generate a pre-distortion compensated audio signal; s76, outputting the audio signal after predistortion compensation to a signal input end of a power amplifier; And S77, converting the final bias voltage value in the final control parameter set into an analog voltage value and outputting the analog voltage value to a bias control end of the power amplifier, converting the final supply voltage value into the analog voltage value and outputting the analog voltage value to a power supply modulation end of the power amplifier, and converting the final on time and the final off time into pulse width modulation signals and outputting the pulse width modulation signals to a switch driving end of the power amplifier.

Description

Energy-saving E-type power amplifier-based energy efficiency optimization method Technical Field The invention relates to the technical field of audio electronics and high-efficiency power amplifier control, in particular to an energy-saving E-class power amplifier-based energy efficiency optimization method. Background In the field of audio electronics, switch-mode power amplifiers have become a key component in high-fidelity sound systems, portable audio devices, and professional sound amplifying equipment by virtue of their theoretical high efficiency advantages. The class E amplifier is used as a specific topological structure, and the state switching of the power switching device is completed under the condition of zero voltage or zero current by carefully designing an external resonant network, so that the switching loss is obviously reduced, and higher energy conversion efficiency is realized. Around the technical route for improving the efficiency of the class-E power amplifier, the prior proposal mainly focuses on circuit topology optimization, static adjustment of an operating point, analog or digital feedback control and other modes. One representative scheme is to approach the theoretical efficiency extremum by accurately designing the inductance and capacitance parameters of the output matching network so that the amplifier satisfies the soft switching state at a specific frequency and under fixed load conditions. Another common scheme is to introduce an envelope tracking technique, and dynamically adjust the supply voltage of the power tube according to the amplitude variation of the input signal, so that the working point of the power tube is always close to a high-efficiency area. In addition, part of digital control schemes use a table look-up or simple algorithm to fine tune the bias points by collecting output voltage and current to cope with slow changes of the load. The prior art solutions have difficulty in simultaneously co-responding to fast transients in the signal envelope and real-time fluctuations in the load impedance. The static parameter optimization scheme is free from the design working condition, the efficiency is rapidly reduced, the envelope tracking technology can track the signal change, but is insensitive to the load impedance change, the adjustment bandwidth is limited by a power supply modulation loop, and the response speed and the control precision of the digital control method based on simple feedback cannot be considered, so that the average energy efficiency is often far lower than the peak value when the complex music signal and the dynamic load are combined. These deficiencies together cause that the existing class-E audio power amplifier has poor comprehensive energy efficiency in the whole working condition range, and can not continuously maintain high-efficiency operation under the condition of double dynamic changes of audio signal amplitude and load impedance. Therefore, how to provide an energy-saving class E power amplifier based energy efficiency optimization method is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention provides an energy-saving E-class power amplifier energy efficiency optimization method, and provides a technical scheme for integrating depth feature perception and intelligent dynamic optimization, aiming at the problem that energy conversion efficiency can not be kept close to an optimal value in a full working condition range when the amplitude of an audio frequency band signal and load impedance of an existing E-class power amplifier are dynamically changed. The scheme comprises the steps of firstly extracting multi-resolution time-frequency characteristics of an audio signal through wavelet transformation and a transducer encoder, utilizing perturbation injection and decoder technology to sense load impedance in real time, further constructing a dynamic working condition state space, identifying a key period of efficiency optimization through a gating attention network, finally adopting a hierarchical asynchronous self-adaptive particle swarm optimization algorithm to respectively conduct collaborative optimization on bias voltage, power supply and switching parameters in a fast loop and a slow loop, and generating final control parameters through decoupling mapping and temperature compensation. The invention can realize accurate tracking and quick response to the dynamic and load changes of the audio signal, effectively improve the average energy efficiency of the power amplifier under the full working condition, and simultaneously consider the constraint on harmonic distortion and thermal state. According to the embodiment of the invention, the energy efficiency optimization method based on the energy-saving E-class power amplifier comprises the following steps of: S1, sampling and frequency band decomposition are carried out on an input audio signal to obtain a plurality of sub-b