A radio frequency power calibration method for a wireless module

Abstract

The present application relates to wireless communication transmission and radio frequency power amplifier technical field, disclose a kind of radio frequency power calibration method for wireless module, comprising: the rectified voltage in calibration radio frequency pulse flat top period is collected and synchronous collection collector bias power supply voltage, deduct parasitic electrical voltage drop according to bias voltage difference and power supply push pressure sensitivity and calculate pure thermal induction gain attenuation rate, calculate reference equivalent rectified voltage, deviation exceeds convergence window, update local sensitivity slope according to adjacent iteration data to modify control word, after convergence, the final control word and pure thermal induction gain attenuation rate are associatedly stored, the physical separation of power supply rail dynamic voltage drop and transistor junction temperature attenuation is realized by the present application, eliminate transient thermal droop interference and power supply ripple crosstalk, avoid control word excessive compensation, reach stable convergence in gain compression zone, improve burst transmit power flatness.

Description

Technical Field

[0001] This invention relates to the field of wireless communication transmission and radio frequency power amplifier technology, and in particular to a radio frequency power calibration method for wireless modules. Background Technology

[0002] Transmit power calibration is a fundamental step in radio frequency (RF) testing of wireless modules. It is used to ensure that RF performance meets radiation specifications and maintains the linearity of the transmit link. Current testing typically uses pulsed RF signals to drive the RF power amplifier. A detector voltage reflecting the transmit power is collected through a directional coupler and a detector diode. This voltage is then compared with a target power reference voltage, and the baseband power control word is adjusted iteratively through a closed loop. Due to the inherent thermal resistance of the active region of the semiconductor, the transistor experiences significant transient power consumption during RF pulse transmission. The junction temperature can rise rapidly within tens of microseconds, leading to a decrease in carrier mobility and causing a significant dynamic gain attenuation in the RF power amplifier during the pulse flat-top period.

[0003] To reduce the impact of dynamic gain droop within the pulse on the sampling level, current tests typically extend the pulse dwell time to several milliseconds, waiting for the power amplifier chip to reach local thermal equilibrium before acquiring the detection voltage. This method increases the measurement waiting time for each frequency point and power step, reducing test efficiency. At the same time, continuous pulse transmission causes the RF front-end printed circuit board to continuously accumulate heat, resulting in ambient substrate temperature drift and affecting the stability of the calibration reference. If the pulse width is shortened and single-point detection is performed at a fixed moment during the pulse peak, the sampled value will contain transient thermal attenuation components that have not yet dissipated. In this case, the baseband processing unit is prone to judging the dynamic gain compression caused by the junction temperature rise as insufficient static intrinsic gain of the RF power amplifier, thereby excessively increasing the baseband power control word, causing power overshoot at the leading edge of the wireless module during actual burst communication transmission and worsening adjacent channel leakage.

[0004] Therefore, while maintaining a short pulse and fast test cycle, how to effectively distinguish between the transient self-heating decay of transistors and the static RF intrinsic gain, and avoid excessive compensation of the baseband power control word caused by single-point detection deviation, has become the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems in the background art, the present invention provides the following technical solution: a method for calibrating the radio frequency power of a wireless module, comprising the following steps: Step S1: Obtain the initial substrate temperature of the RF power amplifier; the baseband processing unit outputs the initial feedforward power control word to control the transmit calibration RF pulse signal. Step S2: During the flat-top period of the calibration RF pulse signal, the first detection voltage and the second detection voltage are acquired at the first sampling time and the second sampling time, respectively, and the first supply voltage and the second supply voltage at the collector bias power supply pin are acquired simultaneously. Step S3: Calculate the parasitic electrical voltage drop based on the voltage difference between the first and second supply voltages and the preset power supply push sensitivity coefficient. Subtract the parasitic electrical voltage drop from the difference between the first and second detector voltages and divide by the time difference between the two sampling times to calculate the pure thermal induced gain attenuation rate. Calculate the reference equivalent detector voltage based on the pure thermal induced gain attenuation rate, the first detector voltage, and the time difference between the first sampling time and the transmission reference time when the rising edge of the calibration RF pulse signal reaches 50% of its peak amplitude. Step S4: Calculate the deviation between the reference equivalent detector voltage and the preset target power reference voltage. When the deviation exceeds the preset convergence tolerance window, calculate the control word step correction based on the preset detector sensitivity slope in the first iteration. In subsequent iterations, update the local sensitivity slope based on the difference between the reference equivalent detector voltage and the difference between the feedforward power control word in two adjacent iterations. Use the local sensitivity slope to correct the feedforward power control word and repeat until the deviation falls into the preset convergence tolerance window. When the deviation falls into the preset convergence tolerance window, use the current feedforward power control word as the final control word, and store the final control word, the pure thermal induced gain attenuation rate, and the initial substrate temperature in a non-volatile storage medium.

[0006] Preferably, step S3, calculating the reference equivalent detector voltage, includes sub-step S31, which multiplies the pure thermal induced gain attenuation rate by the time difference to obtain the temperature rise compensation voltage and adds it to the first detector voltage; step S4 includes sub-step S41, which calculates the difference between the preset target power reference voltage and the current iteration reference equivalent detector voltage, divides it by the local sensitivity slope, rounds it to obtain the control word step correction amount, and adds it to the current iteration feedforward power control word to generate the feedforward power control word after the next iteration update.

[0007] Preferably, step S2 includes sub-step S21, configuring the first sampling time to lag behind the calibration RF pulse signal transmission reference time, and the waiting time to be greater than the electrical setup time constant corresponding to the discharge of the decoupling capacitor in the RF power amplifier bias power supply network; configuring the first sampling time to lag behind the flat-top period start edge by more than or equal to 10 μs, and the time difference between the first sampling time and the second sampling time to be in the range of 20 μs to 80 μs.

[0008] Preferably, step S2 further includes sub-step S22, setting a third sampling time at the midpoint between the first and second sampling times during the flat-top period of the calibration RF pulse signal, and acquiring the third detection voltage; step S3 further includes sub-step S32, calculating the second-order curvature deviation between the average voltage value obtained from the first and second detection voltages and the third detection voltage; when the second-order curvature deviation exceeds the linearity judgment threshold, determining the secondary thermal attenuation compensation factor, and superimposing the secondary compensation term determined by multiplying the secondary thermal attenuation compensation factor by the square of the time difference when calculating the reference equivalent detection voltage, so as to eliminate the gain drooping bending error caused by the nonlinearity of transistor thermal resistance and thermal capacitance under saturated operating conditions.

[0009] Preferably, the initial substrate temperature in step S1 is obtained by sensing with a thermistor located next to the RF power amplifier; the calibration method further includes a sub-step S33 before calculating the pure thermal induced gain attenuation rate in step S3, which obtains the temperature difference between the current initial substrate temperature and the factory calibration reference temperature, calculates the static level drift based on the calibrated dark voltage temperature drift slope of the detector diode, and subtracts the static level drift from the measured first detector voltage and second detector voltage respectively to eliminate the DC zero-point level offset caused by the external detector diode with temperature change.

[0010] Preferably, step S1 further includes sub-step S11, which involves calibrating power control parameters at multiple discrete frequency points within the target operating frequency band, obtaining the final control word and pure thermal induced gain attenuation rate corresponding to each discrete frequency point, comparing the pure thermal induced gain attenuation rates between each discrete frequency point, and determining that the discrete frequency point is in an RF impedance mismatch state when the pure thermal induced gain attenuation rate of a specific discrete frequency point exceeds the preset deviation threshold of the benchmark average value under the same power level, outputting a bias adjustment command to lower the programmable static bias current of the RF power amplifier by one level, and recalibrating after the adjustment to update the final control word.

[0011] Preferably, step S4 further includes sub-step S42: when the wireless module is in the burst data frame transmission state, the final control word stored in the non-volatile storage medium is called to configure the transmission power, and the associated thermally induced gain attenuation rate is read; the corresponding modulation peak-to-average ratio (PAPR) conversion factor is obtained according to the modulation coding method of the current transmission service, and the thermally induced gain attenuation rate is multiplied by the PAPR conversion factor to generate the modulation transmission pre-tilt slope; a baseband gain digital pre-tilt compensation sequence that increases linearly with time according to the modulation transmission pre-tilt slope is superimposed in the burst data frame transmission time slot to offset the intra-pulse gain attenuation of the RF power amplifier caused by the junction temperature rise during burst transmission.

[0012] Preferably, step S3 further includes sub-step S34, where when the target output power is in the small-signal linear region of the amplifier, the difference between the measured first detector voltage and the second detector voltage is equal to zero, the calculated pure thermal induced gain attenuation rate is zero, the calculated reference equivalent detector voltage is equal to the first detector voltage, and the cooling interval between two consecutive transmission calibration RF pulse signals is equal to the substrate recovery delay time.

[0013] Preferably, the associated storage in step S4 includes sub-step S43, which combines the final control word, the pure thermal induced gain attenuation rate, the initial substrate temperature, the power supply push sensitivity coefficient, and the corresponding target operating frequency point into an RF power calibration mapping table entry and writes it into a non-volatile storage medium so that the wireless module can call the configuration when transmitting signals in different frequency bands.

[0014] The beneficial effects of this invention are: 1. In the RF power calibration of the wireless module, by processing the change in detector voltage during the flat-top period of the RF pulse, the pure thermal induced gain attenuation rate and the reference equivalent detector voltage can be obtained. This can reduce the influence of transient self-heating of the RF power amplifier on the power calibration results, making the calibration results closer to the static intrinsic gain of the RF power amplifier. This can avoid the baseband processing unit from excessively increasing the feedforward power control word due to transient gain attenuation, reduce the power overshoot at the leading edge of the wireless module's burst transmission, and reduce the resulting adjacent channel leakage deterioration.

[0015] 2. By simultaneously considering the changes in the bias supply voltage of the RF power amplifier and using the power supply push sensitivity coefficient to deduct the influence of the power supply change on the detector voltage, the interference of the transient voltage drop of the bias supply network on the calculation results of the pure thermal induced gain attenuation rate can be reduced. This allows short pulse calibration to more accurately reflect the thermal attenuation of the RF power amplifier itself, thereby improving the accuracy of power calibration and maintaining good consistency in the bias calibration of different transmit links.

[0016] 3. By updating the correction amplitude of the feedforward power control word with the local sensitivity slope during the iteration process, the RF power amplifier can still adjust the calibration step according to the actual response after entering the nonlinear gain compression region, reducing the iteration hysteresis and oscillation caused by fixed correction parameters. At the same time, by determining the cooling interval between adjacent calibration RF pulses based on the pure thermal induced gain attenuation rate, the influence of residual heat from the previous pulse on the subsequent calibration can be reduced, making the closed-loop calibration process converge more stably. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 The flowchart for the iterative calibration of the power control word considering thermal decay and supply voltage drop in this invention is shown below; Figure 2 This is a reverse push curve diagram of the secondary compensation of the detector voltage under the saturation condition of the power amplifier of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0019] A method for calibrating the radio frequency power of a wireless module includes the following steps: Step S1: Obtain the initial substrate temperature of the RF power amplifier; the baseband processing unit outputs the initial feedforward power control word to control the transmit calibration RF pulse signal. Step S2: During the flat-top period of the calibration RF pulse signal, the first detection voltage and the second detection voltage are acquired at the first sampling time and the second sampling time, respectively, and the first supply voltage and the second supply voltage at the collector bias power supply pin are acquired simultaneously. Step S3: Calculate the parasitic electrical voltage drop based on the voltage difference between the first and second supply voltages and the preset power supply push sensitivity coefficient. Subtract the parasitic electrical voltage drop from the difference between the first and second detector voltages and divide by the time difference between the two sampling times to calculate the pure thermal induced gain attenuation rate. Calculate the reference equivalent detector voltage based on the pure thermal induced gain attenuation rate, the first detector voltage, and the time difference between the first sampling time and the transmission reference time when the rising edge of the calibration RF pulse signal reaches 50% of its peak amplitude. Step S4: Calculate the deviation between the reference equivalent detector voltage and the preset target power reference voltage. When the deviation exceeds the preset convergence tolerance window, calculate the control word step correction based on the preset detector sensitivity slope in the first iteration. In subsequent iterations, update the local sensitivity slope based on the difference between the reference equivalent detector voltage and the difference between the feedforward power control word in two adjacent iterations. Use the local sensitivity slope to correct the feedforward power control word and repeat until the deviation falls into the preset convergence tolerance window. When the deviation falls into the preset convergence tolerance window, use the current feedforward power control word as the final control word, and store the final control word, the pure thermal induced gain attenuation rate, and the initial substrate temperature in a non-volatile storage medium.

[0020] Preferably, step S3, calculating the reference equivalent detector voltage, includes sub-step S31, which multiplies the pure thermal induced gain attenuation rate by the time difference to obtain the temperature rise compensation voltage and adds it to the first detector voltage; step S4 includes sub-step S41, which calculates the difference between the preset target power reference voltage and the current iteration reference equivalent detector voltage, divides it by the local sensitivity slope, rounds it to obtain the control word step correction amount, and adds it to the current iteration feedforward power control word to generate the feedforward power control word after the next iteration update.

[0021] Preferably, step S2 includes sub-step S21, configuring the first sampling time to lag behind the calibration RF pulse signal transmission reference time, and the waiting time to be greater than the electrical setup time constant corresponding to the discharge of the decoupling capacitor in the RF power amplifier bias power supply network; configuring the first sampling time to lag behind the flat-top period start edge by more than or equal to 10 μs, and the time difference between the first sampling time and the second sampling time to be in the range of 20 μs to 80 μs.

[0022] Preferably, step S2 further includes sub-step S22, setting a third sampling time at the midpoint between the first and second sampling times during the flat-top period of the calibration RF pulse signal, and acquiring the third detection voltage; step S3 further includes sub-step S32, calculating the second-order curvature deviation between the average voltage value obtained from the first and second detection voltages and the third detection voltage; when the second-order curvature deviation exceeds the linearity judgment threshold, determining the secondary thermal attenuation compensation factor, and superimposing the secondary compensation term determined by multiplying the secondary thermal attenuation compensation factor by the square of the time difference when calculating the reference equivalent detection voltage, so as to eliminate the gain drooping bending error caused by the nonlinearity of transistor thermal resistance and thermal capacitance under saturated operating conditions.

[0023] Preferably, the initial substrate temperature in step S1 is obtained by sensing with a thermistor located next to the RF power amplifier; the calibration method further includes a sub-step S33 before calculating the pure thermal induced gain attenuation rate in step S3, which obtains the temperature difference between the current initial substrate temperature and the factory calibration reference temperature, calculates the static level drift based on the calibrated dark voltage temperature drift slope of the detector diode, and subtracts the static level drift from the measured first detector voltage and second detector voltage respectively to eliminate the DC zero-point level offset caused by the external detector diode with temperature change.

[0024] Preferably, step S1 further includes sub-step S11, which involves calibrating power control parameters at multiple discrete frequency points within the target operating frequency band, obtaining the final control word and pure thermal induced gain attenuation rate corresponding to each discrete frequency point, comparing the pure thermal induced gain attenuation rates between each discrete frequency point, and determining that the discrete frequency point is in an RF impedance mismatch state when the pure thermal induced gain attenuation rate of a specific discrete frequency point exceeds the preset deviation threshold of the benchmark average value under the same power level, outputting a bias adjustment command to lower the programmable static bias current of the RF power amplifier by one level, and recalibrating after the adjustment to update the final control word.

[0025] Preferably, step S4 further includes sub-step S42: when the wireless module is in the burst data frame transmission state, the final control word stored in the non-volatile storage medium is called to configure the transmission power, and the associated thermally induced gain attenuation rate is read; the corresponding modulation peak-to-average ratio (PAPR) conversion factor is obtained according to the modulation coding method of the current transmission service, and the thermally induced gain attenuation rate is multiplied by the PAPR conversion factor to generate the modulation transmission pre-tilt slope; a baseband gain digital pre-tilt compensation sequence that increases linearly with time according to the modulation transmission pre-tilt slope is superimposed in the burst data frame transmission time slot to offset the intra-pulse gain attenuation of the RF power amplifier caused by the junction temperature rise during burst transmission.

[0026] Preferably, step S3 further includes sub-step S34, where when the target output power is in the small-signal linear region of the amplifier, the difference between the measured first detector voltage and the second detector voltage is equal to zero, the calculated pure thermal induced gain attenuation rate is zero, the calculated reference equivalent detector voltage is equal to the first detector voltage, and the cooling interval between two consecutive transmission calibration RF pulse signals is equal to the substrate recovery delay time.

[0027] Preferably, the associated storage in step S4 includes sub-step S43, which combines the final control word, the pure thermal induced gain attenuation rate, the initial substrate temperature, the power supply push sensitivity coefficient, and the corresponding target operating frequency point into an RF power calibration mapping table entry and writes it into a non-volatile storage medium so that the wireless module can call the configuration when transmitting signals in different frequency bands.

[0028] Example 1: In the mass production test line for wireless communication modules, the transmit power calibration operation for the n78 band under the Sub-6GHz frequency band of fifth-generation mobile communication technology, with a target operating frequency of 3500MHz and a target output power of 26dBm, is carried out using a hardware architecture that includes a baseband processing unit, an RF transceiver, a heterojunction bipolar transistor RF power amplifier, a directional coupler, a Schottky detector diode, a dual-channel high-speed analog-to-digital converter, a thermistor, and a non-volatile storage medium. The directional coupler is connected in series between the output port of the RF power amplifier and the antenna interface. Its coupling terminal is connected to a Schottky detector diode, and the detection voltage output by the Schottky detector diode is connected to the first channel of the dual-channel high-speed analog-to-digital converter. A thermistor mounted on the surface of the circuit board next to the RF power amplifier package is used to sense the substrate temperature of the module board. The collector bias power supply pin of the RF power amplifier is connected to the second channel of the dual-channel high-speed analog-to-digital converter through a voltage divider network to acquire the power supply bias level. Under the high-speed cycle requirements of the production line, if the pulse emission dwell time is extended to several milliseconds to wait for the power amplifier chip and package to reach thermal equilibrium, not only will the waiting time for machine testing increase several times, but also... Continuous high-load excitation will cause sustained global heat accumulation on the RF test fixture and RF front-end circuit board, resulting in substrate temperature drift and degrading the physical accuracy of the calibration reference. If the pulse width is simply compressed to the microsecond level and single-point sampling is performed at a fixed time during the flat-top period, the transient high current extraction caused by the high-power RF pulse will cause the power supply filter decoupling capacitor to discharge, resulting in a dynamic voltage drop at the collector power supply pin. The detector voltage drop induced by this electrical drop and the decrease in carrier mobility and dynamic gain droop caused by the rapid rise in junction temperature of the transistor active region are deeply overlapped within the time domain sampling window. If the distortion of the two drops is superimposed, the distortion will be significant. Using sampled values ​​directly to correct the baseband power control word inevitably misjudges the dual distortions of power supply rail dynamic fluctuations and transistor transient self-heating as insufficient static intrinsic gain of the amplifier. This causes the feedforward power control word to be over-adjusted, resulting in significant transient power overshoot and excessive adjacent channel radiation at the leading edge of subsequent actual burst communication transmissions. At the same time, when the RF power amplifier is in the high-power operating region and approaches the gain compression boundary, the amplifier output power increment relative to the step response of the baseband control word exhibits nonlinear compression characteristics. Using a preset detector sensitivity slope to calculate the control word compensation step is prone to causing optimization oscillations or convergence hysteresis in the saturation range.

[0029] To address the aforementioned dynamic constraints involving both electrical and thermal aspects, the test system uses a baseband processing unit to lock the local oscillator of the RF transceiver to 3500MHz, controls the RF antenna switch to the n78 transmit channel, and reads the package substrate level sensed by the thermistor via an analog-to-digital converter interface. Based on a preset resistance-temperature calibration curve, the initial substrate temperature of the RF power amplifier is calculated. The temperature is set to 25°C; subsequently, the baseband processing unit retrieves the preset initial feedforward power control word from the non-volatile storage medium. The baseband generator outputs a rectangular RF pulse with a preset total duration of 100μs. After quadrature modulation and digital-to-analog conversion, the pulse is input to the RF power amplifier. The rise time of the RF pulse is approximately 5μs, and the moment when the pulse rises to 50% of its peak amplitude is defined as the reference moment for calibrating the RF pulse signal transmission. The sampling time is 0 μs. To avoid initial ringing at the pulse leading edge caused by the distributed inductance of the RF matching network and parasitic oscillations of the chip package leads, and to avoid the nonlinear transient settling period caused by the initial large current discharge of the decoupling capacitor in the power supply network, the baseband processing unit selects the first sampling time within the pulse flat-top stabilization period. 20μs and the second sampling time The time interval between two sampling points is 80 μs. The sampling time is 60μs, and the time lag between the first sampling moment and the pulse emission reference moment is greater than the electrical settling time constant of the power supply network decoupling filter; during the first RF pulse emission, the dual-channel high-speed analog-to-digital converter performs synchronous acquisition according to the hardware timing trigger sequence. At a time interval of 20μs, the first channel acquires the first detector voltage. The voltage is 1520mV, and the first supply voltage is synchronously acquired by the second channel. 3385mV; in At a time interval of 80μs, the first channel acquired the second detector voltage. The voltage is 1468mV, and the second supply voltage is synchronously acquired through the second channel. It is 3375mV.

[0030] After receiving the aforementioned multi-dimensional time-domain synchronous sampling data, the baseband processing unit performs power supply fluctuation decoupling and junction temperature dynamic response calculation, and calculates the transient sag of the power supply rail based on the power supply voltage difference between the two sampling times. The voltage is 10mV. The power supply push sensitivity coefficient of the RF power amplifier, previously calibrated in a dark state without RF input, is retrieved. The value is 0.4, representing the detector voltage drop ratio caused by unit supply voltage fluctuation, and is a dimensionless coefficient; the parasitic electrical voltage drop component induced by the power supply rail drop is calculated. The measured total detection voltage drop between the first and second sampling times was 4mV. The baseband processing unit subtracted the parasitic electrical voltage drop component from the total voltage drop and extracted the pure electrothermal attenuation caused by the rise in the junction temperature of the transistor's active region. The value is 48mV, and the thermally induced gain attenuation rate is calculated using the time difference between the two sampling times: ,in, For purely thermally induced gain attenuation rate, This is the first detection voltage. This is the second detection voltage. This is the power supply push sensitivity coefficient. The first power supply voltage collected at the first sampling moment. The second power supply voltage is collected at the second sampling time. This is the first sampling time. At the second sampling time, the thermally induced gain attenuation rate is calculated by substituting the sampled values. The value is 0.8 mV / μs; after obtaining the purely thermally induced gain attenuation rate, the baseband processing unit performs a reverse time-domain deduction to eliminate the self-emission reference time. Up to the first sampling time The dynamic gain compression caused by the rise in semiconductor junction temperature during this period allows us to calculate the reference equivalent detector voltage under zero-temperature-rise reference conditions: ,in, The reference equivalent detector voltage for eliminating transient thermal effects, To calibrate the reference time for RF pulse signal transmission, the reference equivalent detector voltage is calculated by substituting various values. The reference equivalent detector voltage is 1536mV. Physically, this reference voltage is defined as the reference level mapped at the detector end of the virtual isothermal steady-state RF output when the active region of the heterojunction bipolar transistor is at its initial substrate temperature of 25°C and is not modulated by intra-pulse self-heating temperature rise. It does not represent the actual sampling reading during the physical transient process of the pulse rising edge, but rather serves as a static intrinsic gain calibration anchor point after removing junction temperature transient rise interference, eliminating the output power droop caused by lattice thermal relaxation during the pulse flat-top period. The baseband processing unit compares the reference equivalent detector voltage with the preset target power reference voltage of 1650mV. By comparing the two, the absolute deviation between them can be obtained. The value is 114mV. Specifically, during the process of a heterojunction bipolar transistor being triggered from the pulse leading edge to reaching a thermally diffused steady state, the first 20μs period is buffered by the absorption of the micro-scale emitter mesa micro-region thermal capacity of the die. The rise in the active region lattice temperature of the transistor has an electrothermal diffusion delay of about 8μs. During the first 5μs pulse rise period of the emission leading edge, the RF output has not yet reached its peak value, and the heat generation power is in the integral growth stage. From 5μs to 20μs, the slope of the transistor junction temperature rise has smoothly transitioned to a linear evolution stage dominated by the substrate main thermal resistance. As a result, the actual cumulative temperature rise from the pulse emission reference time to the first sampling time drops. Compared with the value derived by inverse linear extrapolation based on the pure thermally induced gain decay rate from 20μs to 80μs during the flat-top period, the total integral deviation converted to the detection level is only 0.8mV, which is far below the preset convergence tolerance window of 10mV. Therefore, using first-order linear back-calculation in the rated operating area can greatly simplify the algorithm calculation overhead while ensuring test accuracy. Since the deviation exceeds the preset convergence tolerance window... The voltage is 10mV, and the preset detector sensitivity slope is used in the first iteration. Calculate the control word step correction amount, given a value of 16.25mV / LSB. : ,in, This is the control word step correction amount for the first iteration. To preset the target power reference voltage, The preset detector sensitivity slope characterizes the change in detector voltage corresponding to a unit step of the feedforward power control word. This indicates the rounding operation; the result is calculated as follows: For 7, update the feedforward power control word. .

[0031] To block interference from residual heat accumulation at the board level caused by continuous pulse iteration, the baseband processing unit, before controlling the transmission of the second calibration pulse, uses the currently calculated thermally induced gain attenuation rate. The inter-pulse cooling interval duration was dynamically calculated to be 0.8 mV / μs. ,in, This refers to the cooling interval duration. The substrate recovery delay time is set to 200 μs. The thermal recovery coefficient is calibrated to 150 μs / (mV / μs); the inter-pulse cooling interval is calculated by substituting the values. The system maintains a quiescent state with no RF output for 320μs until this time, allowing the heat accumulated in the transistor's active region to fully recede to the initial substrate temperature of 25°C. The aforementioned substrate recovery delay time and thermal recovery coefficient settings are derived from the thermal diffusion transient test procedure for heterojunction bipolar transistor dies and package bases. In the initial thermal resistance calibration test, the main thermal resistance relaxation time constant from the transistor junction region to the package lead frame was measured to be 195μs using an infrared transient temperature measurement system. When the pulse is completely turned off, the minimum physical relaxation threshold required for heat dissipation from the chip's active region to the package ground copper is 200μs. If the inter-pulse waiting time... Below 200μs, subsequent pulses will superimpose the undissipated local residual temperature rise from the preceding pulse. The thermal recovery coefficient of 150μs / (mV / μs) is derived from the measured increase in thermally induced gain attenuation rate of approximately 0.18 watts in junction transient power dissipation for every 0.1mV / μs increase in pure thermal induced gain attenuation rate, and the additional extended thermal diffusion time required. The measured attenuation curve shows that an adaptive cooling time of 150μs is required per unit attenuation rate to ensure that the residual junction temperature recovers to within ±0.1℃ of the initial base plate temperature, thus avoiding thermal accumulation hysteresis during continuous pulse optimization. Subsequently, the baseband processing unit loads the updated control word. The value is 127, triggering the transmission of the second calibration pulse, which is synchronously acquired by the high-speed analog-to-digital converter. The first detection voltage at 20 μs 1602mV and supply voltage 3382mV, and The second detector voltage at 80 μs 1544mV and supply voltage The voltage drop is 3372mV, and the calculated voltage drop is... For 10mV and parasitic electrical voltage drop The voltage is 4mV. After deducting the electrical voltage drop, the pure electrothermal attenuation is obtained. The voltage is 54mV. The thermally induced gain attenuation rate at this value is calculated. The value is 0.9 mV / μs, and the reference equivalent detector voltage corresponding to the second transmission is calculated. It is 1620mV; at this point, the absolute deviation is... The value is 30mV, which exceeds the preset convergence tolerance window. A value of 10mV indicates that the output has not yet converged. This is because the power amplifier's transistors enter a gain compression range under high power output, causing the large-signal gain slope to deviate significantly from the nominal small-signal value. Continuing to use the preset detector sensitivity slope will result in insufficient step correction and induce multiple invalid iterations. The baseband processing unit dynamically extracts the local sensitivity slope based on the data evolution between two adjacent iterations. ,in, The slope of local sensitivity. This is the reference equivalent detector voltage for the current iteration period. The reference equivalent detector voltage is the voltage of the immediately preceding iteration cycle. For the feedforward power control word used in the current iteration cycle, The feedforward power control word used in the previous iteration cycle; substituting it into the numerical calculation yields... The extracted local sensitivity slope of 12mV / LSB is lower than the preset detector sensitivity slope of 16.25mV / LSB, indicating that the RF power amplifier transistor has entered the large-signal gain compression nonlinear region. The baseband processing unit uses the local sensitivity slope instead of the preset detector sensitivity slope and calculates the new step correction amount. : ,in, The control word step correction amount for the second iteration is obtained by substituting it into the numerical calculation. This leads to the third update of the feedforward power control word. Before initiating the third calibration transmission, the baseband processing unit uses the updated attenuation rate. Calculate the cooling interval duration as 0.9 mV / μs: It maintains no RF output until the heat in the transistor's active region is fully dissipated before loading the feedforward control word. The value is 130, controlling the transmission of the third calibration pulse; the dual-channel high-speed analog-to-digital converter synchronously samples again according to the preset timing sequence. The first detector voltage was measured at 20 μs. 1639mV and supply voltage The value is 3380mV. The second detector voltage was measured at 80 μs. 1579mV and supply voltage The voltage drop is 3370mV; at this time, the supply voltage drop is... Maintaining a voltage of 10mV, parasitic voltage drop The voltage drop is 4mV. The pure electrothermal attenuation is obtained after removing the electrical voltage drop. The voltage is 56mV. The pure thermally induced gain attenuation rate is calculated as follows: The reference equivalent detector voltage is obtained by reverse-engineering this parameter back to the transmission reference time: The absolute deviation was calculated as follows: Since the deviation satisfies: If the output power falls within the preset convergence tolerance window, the optimization iteration loop is terminated; the baseband processing unit will then process the final control word. Final thermally induced gain attenuation rate Power supply push sensitivity coefficient Initial base temperature The target operating frequency is encapsulated as a structured data entry, with corresponding values ​​of 130, 0.9333mV / μs, 0.4, 25℃ and 3500MHz, respectively, and is written into the non-volatile storage medium of the wireless module.

[0032] During the stage when the module completes mass production calibration and connects to the cellular network to transmit burst service signals, for 5G NR uplink 256QAM high-order modulation burst signals with non-constant envelope and peak-to-average power ratio characteristics, the signal peak-to-average power ratio is 7.5dB and the duration of a single burst data frame is 500μs. The baseband processing unit calls the final control word solidified in the non-volatile storage medium. Configure the RF amplifier transmit power level. The value is 130, and the purely thermally induced gain decay rate is read from the associated storage. Its value is 0.9333 mV / μs; the preset modulation peak-to-average power ratio (PAPR) conversion factor is called according to the 256QAM modulation format. Its value is 0.65, and the modulated transmit pre-tilt slope is calculated as follows: ,in, To modulate the transmit pre-tilt slope, This is the modulation peak-to-average power ratio (PAPR) conversion factor, set to 0.65 for the 256QAM modulation format; The modulated transmit pre-tilt slope is calculated by substituting the parameters into the pure thermally induced gain attenuation rate: The baseband processing unit applies a time-dependent modulation signal to the quadrature modulation signal within a 500μs burst transmit time slot. A linearly increasing baseband gain digital pre-tilt compensation sequence, with a determined reverse slope, in-situ cancels the junction temperature rise and dynamic gain droop caused by RF power dissipation in the transistor active region. This ensures that the RF power radiated from the antenna port maintains a flat envelope throughout the entire burst data frame duration. The power peak fluctuation converges from 0.85dB in the uncompensated state to within 0.05dB, the error vector amplitude EVM improves from -31.5dB to -36.2dB, and the adjacent channel leakage ratio ACLR improves by 5.5dB. This achieves synergistic compensation of transient power supply filtering drop-off isolation, saturation gain compression fast convergence, and transmit pulse thermal droop compensation. This enables the wireless module's RF power amplifier to possess high-precision calibration characteristics under microsecond-level short pulse test cycles, while maintaining the flatness of the actual transmit power and the quality of the modulated signal transmission at the air interface. Specifically, the physical value of the modulation peak-to-average ratio (PAPR) reduction factor is determined based on the statistical PAPR and average transmit heat dissipation ratio of the current communication service modulation format. During the calibration pulse test, the transmit power... The frequency pulse is excited by a continuous constant envelope rectangular wave with a peak-to-average power ratio (PAPR) of 0 dB, and the transistor is in a continuous full-power heat dissipation state. However, in actual burst traffic transmission, the 256QAM modulation signal has non-constant envelope characteristics and a PAPR as high as 7.5 dB. Its root mean square average heat dissipation power of the envelope amplitude is only 65% ​​of that of a constant envelope pulse with the same peak value. Combined with the microsecond-level thermal filtering relaxation time of the semiconductor active region, the junction temperature rise slope caused by it in the burst time slot is attenuated to 0.65 times that in the continuous wave state. Therefore, the modulation... The modulation peak-to-average power ratio (PAPR) conversion factor is configured to 0.65. For other modulation formats, a PAPR lookup table is embedded in the baseband processing unit. The PAPR conversion factor for quadrature phase shift keying modulation with a PAPR of 3.5dB is configured to 0.82, and the PAPR conversion factor for 64QAM modulation with a PAPR of 6.2dB is configured to 0.71. The baseband processing unit dynamically indexes this lookup table to obtain the corresponding PAPR conversion factor based on the modulation coding scheme called in the current frame.

[0033] Example 2: The mass production test of the RF power amplifier in the current wireless communication module uses an n78 transmit channel in the Sub-6GHz band, with a target operating frequency of 3500MHz. The hardware test link includes an RF vector signal generator, a dual-channel 14-bit high-speed analog-to-digital converter, a directional coupler with a coupling of 20dB and a directionality greater than 25dB, a Schottky detector diode operating in the near-linear detection region, a vector signal analyzer with an analysis bandwidth of 100MHz and a noise floor better than -42dB, a programmable DC regulated bias power supply with a transient response time of less than 5μs, and a surface-mounted platinum resistance temperature sensor with a temperature measurement accuracy of 0.1℃. During the test, a dynamic voltage drop ripple with a time-domain step of 10mV is injected into the DC bias power supply rail. The RF detector channel is superimposed with Gaussian white noise background disturbance with a signal-to-noise ratio of 25dB. The temperature of the test fixture substrate is within the natural micro-drift range of 25℃ to 26.5℃. The total duration of the calibration RF pulse signal is set to 100μs. The first sampling time... The second sampling time is 20 μs. The time difference between the two sampling times is 80 μs. The sampling time is 60 μs; the first sampling time lags at least 10 μs from the beginning edge of the flat-top period, and a 20 μs time margin is maintained between the second sampling time and the pulse turn-off, so that the sampling time difference is within the range of 20 μs to 80 μs, and the substrate recovery delay time is... Set to 200 μs, thermal recovery coefficient Set to 150 μs / (mV / μs), both are calibrated based on the measured dissipation curves of the thermal resistance and thermal capacity inside the heterojunction bipolar transistor die, and are used to determine the cooling interval between consecutive iterative emission.

[0034] The experiment included the present invention sample group, four control groups, and three out-of-range control groups. The present invention sample group employed dual-channel synchronous sampling. Before calculating the pure thermally induced gain attenuation rate, the parasitic electrical voltage drop corresponding to the power rail drop was deducted. The reference equivalent detector voltage was calculated backwards from the dual-sampling point data. After the RF power amplifier entered the gain compression region, the feedforward power control word was corrected using a local sensitivity slope. Simultaneously, the inter-pulse cooling interval was determined according to the pure thermally induced gain attenuation rate. Control group 1 used a long-pulse steady-state sampling method with an RF pulse width of 2000 μs and a fixed single-point sampling time set at [time missing]. At the end of the pulse at 1800μs, a preset detection sensitivity slope and a fixed inter-pulse waiting time were used; Control group 2 used a 100μs short pulse, and only collected a single detection voltage at the 50μs peak, without performing thermal attenuation reverse calculation; Control group 3 still collected the first and second detection voltages during the peak, but did not collect the collector bias supply voltage simultaneously, and directly used the total detection voltage drop as the thermal attenuation; Control group 4, after completing the extraction of the pure thermal induced gain attenuation rate and the calculation of the benchmark equivalent detection voltage, always used a preset detection sensitivity slope of 16.25mV / LSB. Calculate the control word correction amount, do not update the local sensitivity slope, and adjust the control group 1 (out-of-range) at the first sampling time. The sampling time was advanced to 8μs; the second sampling time was adjusted for the out-of-range control group 2. The sampling time difference was shortened to 15 μs by setting it to 35 μs; the inter-pulse waiting time was fixed at 50 μs for the out-of-range control group 3. Except for the power gradient test in the small signal linear region and the high compression saturation region, the sample group of this invention and the above control groups were tested at a target output power of 26 dBm and a target operating frequency of 3500 MHz. The initial feedforward power control word Both are 120, preset target power reference voltage The preset convergence tolerance window is 1650mV. It is 10mV.

[0035] Under the rated operating condition of 26dBm, the sample of this invention entered the preset convergence tolerance window after 3 pulse iterations, with a total calibration time of 1.25ms. The final reference equivalent detector voltage was 1657.67mV, and the final control word was 130. After using this control word to transmit burst data frames, the burst flat-top power fluctuation was 0.04dB, the error vector amplitude was -36.2dB, and the adjacent channel leakage ratio was -42.5dBc. This set of data corresponds to a calibration process involving dual-channel power supply decoupling, dual-sampling point time-domain back-inference, local sensitivity slope update, and inter-pulse cooling.

[0036] Control group 1 used a 2000μs long pulse and sampled at 1800μs. The calibration was completed after two iterations, with a total calibration time of 12.8ms. The final reference equivalent detector voltage was 1648.5mV, and the final control word was 129. The subsequent burst transmission had a flat-top power fluctuation of 0.35dB, an error vector amplitude of -33.8dB, and an adjacent channel leakage ratio of -38.2dBc. During the long pulse continuous excitation, the substrate temperature on the fixture surface increased unidirectionally by 3.2℃, causing the calibration reference to drift with the substrate temperature.

[0037] Control group 2 performed single-point sampling at 50μs during the flat-top phase of the 100μs short pulse, with a single-point detection voltage of 1588mV. The control word was corrected based on this sampling result. This group completed the calibration in two iterations, with a total calibration time of 0.45ms, and the final control word was 133. The flat-top power fluctuation during burst emission was 0.88dB, the error vector amplitude was -28.4dB, and the adjacent channel leakage ratio was -32.5dBc. Since the transistor had already experienced thermal droop during single-point sampling, the sampling result also included transient thermal decay, and the control process increased the feedforward power control word accordingly. At the actual burst emission leading edge, the transistor junction temperature had not yet increased by the same magnitude, resulting in power overshoot. The adjacent channel leakage ratio differed by 10dB from the sample group of this invention.

[0038] In control group 3, the collector bias supply voltage was not collected. When a 10mV drop in the supply rail occurred between two sampling times, the total detector voltage drop of 52mV was directly used as the thermal attenuation for calculation. The pure thermal induced gain attenuation rate was 0.8667mV / μs. Based on this result, the final reference equivalent detector voltage was calculated to be 1674mV. The control process set the final control word to 126, and the total calibration time after 3 iterations was 1.2ms. The subsequent burst transmission had a flat-top power fluctuation of 0.42dB, an error vector amplitude of -32.1dB, an adjacent channel leakage ratio of -37dBc, and an actual transmission power deviation of 0.75dB from the target power. In this group, the parasitic electrical voltage drop caused by the dynamic drop in the supply rail was not deducted from the detector voltage difference. Therefore, this part of the voltage change was included in the calculation of both the pure thermal induced gain attenuation rate and the reference equivalent detector voltage.

[0039] Control group 4 retains the power supply decoupling and dual-sampling point reverse derivation, but continues to use the preset detector sensitivity slope of 16.25mV / LSB in the gain compression region. Under this condition, the actual sensitivity of the large signal of the RF power amplifier drops to 12mV / LSB. The control word step calculated using the original preset detector sensitivity slope is insufficient, ultimately requiring 6 iterations, with a total calibration time of 2.65ms. The final reference equivalent detector voltage is 1655.8mV, the final control word is 130, the burst flat-top power fluctuation is 0.05dB, the error vector amplitude is -36dB, and the adjacent channel leakage ratio is -42.3dBc.

[0040] In the sampling time boundary test, the out-of-range control group 1 will use the first sampling time. The sampling point was moved to 8μs, entering the parasitic oscillation region at the leading edge of the pulse. The initial overshoot and ringing harmonics generated by the distributed inductance superimposed on the detector voltage caused high-frequency noise fluctuations in the purely thermally induced gain attenuation rate. After 5 iterations, the total calibration time was 2.2ms. The final reference equivalent detector voltage was 1632.4mV, and the final control word was 128. The sudden flat-top power fluctuation was 0.22dB, the error vector amplitude was -33dB, and the adjacent channel leakage ratio was -36.5dBc.

[0041] Out-of-range control group 2 will have a second sampling time. Set to 35μs to make the sampling time difference The time was shortened to 15μs. At this time, the measured voltage difference between the two sampling points was 12mV. After the difference was affected by the quantization noise of the analog-to-digital converter and the background Gaussian noise, the calculated variance of the pure thermal induced gain attenuation rate increased by 24.5%, and the reference equivalent detector voltage showed random disturbances. After 4 iterations, the total calibration time was 1.75ms. The final reference equivalent detector voltage was 1641.2mV, the final control word was 129, the sudden flat-top power fluctuation was 0.18dB, the error vector amplitude was -34.2dB, and the adjacent channel leakage ratio was -38.8dBc.

[0042] In the out-of-range control group 3, the inter-pulse waiting time was fixed at 50 μs. When the next calibration pulse was triggered, the junction temperature formed by the previous pulse had not yet fully recovered to the initial base temperature. Subsequent pulses continued to add residual temperature rise. After 4 iterations, the total calibration time was 0.7 ms. The final reference equivalent detector voltage was 1668.5 mV, and the final control word was 127. The burst flat-top power fluctuation was 0.28 dB, the error vector amplitude was -33.5 dB, and the adjacent channel leakage ratio was -37.8 dBc.

[0043] In the small-signal linear region of the power gradient test, the target output power was set to 5dBm, and the actual RF radiated power of the RF power amplifier was 3.16mW. At this time, the RF large-signal current swing of the transistor collector was small, and the additional dynamic junction temperature change during the 100μs short pulse excitation was less than 0.04℃. The measured first detector voltage was 400.2mV, and the measured second detector voltage was 400.1mV. The difference between the two was 0.1mV. This difference fell within the quantization noise margin of the analog-to-digital converter chip and was treated as zero in the finite-precision calculation of the system. Therefore, the pure thermally induced gain attenuation rate was taken as zero. The reference equivalent detector voltage was taken as the calibration reference corresponding to the first detector voltage. The calibration was completed in one iteration under this condition, and the total calibration time was 0.35ms. The final reference equivalent detector voltage was recorded as 400.1mV, the final control word was 36, the sudden flat-top power fluctuation was 0.01dB, the error vector amplitude was -39.5dB, and the adjacent channel leakage ratio was -48.2dBc.

[0044] When the target output power is increased to 28dBm, the RF power amplifier enters the high compression saturation region, and the transistor thermal response exhibits nonlinear bending. The third detector voltage obtained at the midpoint sampling time... A second-order curvature deviation of 6mV is formed between the linear mean of the first and second detector voltages, based on which the secondary thermal attenuation compensation factor is activated. The value is 0.00667mV / μs², and a secondary compensation term is superimposed in the calculation of the reference equivalent detector voltage. The reference equivalent detector voltage obtained by the three-point compensation process is 1914.67mV. In the final test record of this operating condition, the reference equivalent detector voltage is 1914.7mV, the final control word is 138, and the calibration is completed after 3 iterations. The total calibration time is 1.4ms. The sudden flat-top power fluctuation is 0.06dB, the error vector amplitude is -34.8dB, and the adjacent channel leakage ratio is -39.6dBc.

[0045] As can be seen from the above tests, during the microsecond-level short pulse calibration process, after synchronously acquiring the collector bias supply voltage and deducting the parasitic electrical voltage drop, the portion of the detector voltage difference used to calculate the pure thermally induced gain attenuation rate no longer includes the dynamic drop of the power supply rail. By reverse-calculating the reference equivalent detector voltage based on the dual-sampling point data and updating the local sensitivity slope with adjacent iteration data in the gain compression region, the control word step deviation caused by the fixed slope in the large signal compression region can be reduced. After the inter-pulse cooling interval is adjusted with the pure thermally induced gain attenuation rate, the next calibration pulse is triggered after the transistor junction region completes the corresponding thermal recovery. When the calibration parameters are used for subsequent burst service transmissions, they can be used to generate compensation amounts that vary with the transmission time slot, so as to reduce the intra-pulse gain attenuation and RF power envelope changes caused by transistor self-heating during burst transmissions.

[0046] Example 3: When calibrating the multi-carrier band transmit power of the wireless communication module under wide temperature conditions, the Schottky detector diode mounted next to the directional coupler has a negative temperature coefficient. After the module substrate temperature deviates from the factory calibration reference temperature, its DC dark level drifts to zero with temperature. In the no-RF dark state with the RF power amplifier excitation signal turned off, the system places the module in a temperature-controlled test chamber and steps the ambient temperature within the range of -20℃ to 85℃. The static dark voltage output by the Schottky detector diode is read through the analog-to-digital converter, and the dark voltage temperature drift slope of the detector diode is calibrated according to the ratio of the dark voltage change to the temperature change. Its value is -1.2mV / ℃. During the production line calibration process, the baseband processing unit reads the initial substrate temperature through a thermistor mounted next to the RF power amplifier package. The factory-calibrated reference temperature is 35℃. The temperature is 25℃, and the temperature difference between the two is 10℃. The static level drift can be calculated from this: ,in, The static level drift is -12mV. Before using the measured first and second detector voltages for attenuation rate calculation, the baseband processing unit subtracts -12mV of static level drift, which is equivalent to adding 12mV compensation to the measured detector voltage to separate the level offset caused by the ambient temperature drift of the external detector device from the transient self-heating change inside the transistor active region.

[0047] Under 28dBm high-power deep saturation emission conditions, the heterojunction bipolar transistor enters the large-signal gain compression region. The heat flow diffusion from the die active region to the packaging substrate is dominated by the exponential temperature rise function of the first relaxation time constant of 45μs and the second relaxation time constant of 195μs within the device. Within the 20μs to 80μs pulse flat-top sampling window, due to the electrothermal integral effect, the temperature rise curve expands according to the Taylor series and exhibits an objectively measurable quadratic parabolic bending component. The detector voltage shows a downward bending during the pulse flat-top period. The baseband processing unit transmits a calibration RF pulse signal with a duration of 100μs, and the first sampling time is selected during the pulse flat-top period. Second sampling time A third sampling time is set at the midpoint between the two sampling times. A dual-channel high-speed analog-to-digital converter synchronously acquires the detector voltages at three sampling times. After deducting the parasitic electrical voltage drop corresponding to the dynamic dropout of the collector bias power supply pin and the static level drift of the detector diode, the three corrected detector voltages are denoted as follows: , and The average value of the first and second detector voltages is 1832mV. Based on this, the baseband processing unit calculates the second-order curvature deviation between this average value and the third detector voltage. ,in, Substituting the second-order curvature deviation into the three correction detector voltages, we get: If the second-order curvature deviation exceeds the linearity threshold of 3mV, the baseband processing unit initiates secondary compensation and calculates the secondary thermal attenuation compensation factor. : Simultaneously, the thermally induced gain attenuation rate is calculated based on the corrected detector voltages at the first and second sampling times: When calculating the reference equivalent detector voltage, a secondary compensation term is superimposed on the first-order thermal attenuation compensation. This term is obtained by multiplying the secondary thermal attenuation compensation factor by the square of the time difference between the transmission reference time and the first sampling time. calibrate the reference time for radio frequency pulse signal transmission. Taking 0 μs, substituting the values, we get: The denominator of the secondary thermal attenuation compensation factor is calculated by multiplying the time difference between the third and first sampling times and the time difference between the second and third sampling times. This is used to extract the second-order change of the detector voltage curve corresponding to the three sampling points. The linear judgment threshold is set to 3mV. The equivalent effective resolution quantization noise level of the dual-channel high-speed analog-to-digital converter is 0.8mV. The extreme value of the thermal noise and random drift of the Schottky detector diode within the microsecond-level sampling window is 1.2mV. The upper limit of the background random noise after the sum of the squares of the two is 1.5mV. Twice of this noise upper limit is used as the 3mV linear judgment threshold to distinguish between the curvature change caused by sampling noise and the nonlinear bending of the detector voltage under deep saturation conditions.

[0048] Within the n78 operating frequency band of 3300MHz to 3800MHz, the baseband processing unit calibrated the 3300MHz low-frequency point, 3500MHz center frequency point, and 3800MHz high-frequency point with the same target output power of 26dBm, and recorded the final control word and pure thermally induced gain attenuation rate for each discrete frequency point. At 3300MHz, the pure thermally induced gain attenuation rate was... It is 0.82 mV / μs; at 3500 MHz, The reference average value formed by the discrete frequency points of the operating frequency band is 0.88mV / μs. For the 3800MHz high-frequency edge frequency point, the RF microstrip impedance mismatch causes the power-added efficiency to decrease from 42.5% at the center frequency point to 34%, and the corresponding pure thermal induced gain attenuation rate increases to 1.25mV / μs. The baseband processing unit uses 0.88mV / μs as the reference average value under the same power level, and forms an attenuation rate judgment boundary of 1.1mV / μs according to the preset deviation threshold. The deviation of 1.25mV / μs at the 3800MHz frequency point from the reference average value exceeds the preset deviation threshold. Therefore, it is determined that the discrete frequency point is in the RF impedance mismatch state.

[0049] After determining the RF impedance mismatch state, the baseband processing unit writes a bias adjustment command to the programmable bias control register of the RF power amplifier via the RF front-end digital control bus. This lowers the base bias voltage by one step, reduces the programmable static bias current by 15mA, and adjusts the DC bias operating point of the RF power amplifier towards deeper Class AB. After completing the bias adjustment, a single-pulse calibration is re-executed. The updated thermally induced gain attenuation rate drops back to 1.02mV / μs, and the feedforward power control word increases from 130 to 131 and converges again. This bias adjustment maintains the impedance matching structure of the external passive microstrip line and antenna end unchanged. Instead, it reduces the collector static DC current and its corresponding heat dissipation under high-frequency reflection mismatch conditions by changing the active DC bias operating point of the RF power amplifier. Subsequently, the feedforward power control word compensates for the RF output change caused by the bias adjustment, allowing the output power to reach the target power again.

[0050] After completing the calibration of each operating frequency, the baseband processing unit combines and encapsulates the corresponding target operating frequency, final control word, pure thermal induced gain attenuation rate, initial substrate temperature, power supply push sensitivity coefficient, and programmable bias register configuration value, and writes them into a non-volatile storage medium to form the corresponding RF power calibration mapping table entry. When the module subsequently transmits broadband burst communication signals, the baseband processing unit reads the corresponding final control word according to the current target operating frequency to configure the static transmit power level. At the same time, it obtains the modulation peak-to-average power ratio (PAPR) conversion factor according to the current modulation format, multiplies it by the pure thermal induced gain attenuation rate to obtain the modulation transmit pre-tilt slope, and generates a baseband gain digital pre-tilt compensation sequence accordingly.

[0051] In the digital baseband hardware link, the baseband gain digital pre-tilt compensation sequence is executed by a digital multiplier located in the baseband processing unit before the digital-to-analog converter. The baseband processing unit uses a 122.88MHz baseband sampling clock as the driving reference and starts a digital ramp generator within a 500μs burst transmit time slot to generate a dimensionless gain scalar sequence that starts from a value of 1 and increases linearly point by point with the sampling clock cycle according to the modulation transmit pre-tilt slope. At each sampling clock cycle, the same gain scalar is fed in parallel into the in-phase baseband digital channel multiplier and the quadrature baseband digital channel multiplier, so that the in-phase and quadrature digital samples are synchronously scaled at the same ratio. At the same time, the baseband processing unit retains a 1.5dB digital dynamic margin in the front-end digital filter so that when the sequence reaches the peak gain at the end of the burst transmit time slot, the peak digital samples of the in-phase and quadrature outputs remain within the full-scale range of the digital-to-analog converter.

[0052] Example 4: This example combines Figures 1 to 2 This paper describes a method for calibrating radio frequency power in a wireless module, such as... Figure 1As shown, in step S1, the initial substrate temperature of the RF power amplifier is obtained, and the baseband processing unit outputs an initial feedforward power control word to control the transmission calibration RF pulse signal; in step S2, during the flat-top period of the calibration RF pulse signal, the first detector voltage and the second detector voltage are acquired at the first sampling time and the second sampling time, respectively, and the first supply voltage and the second supply voltage at the collector bias power supply pin are acquired simultaneously; in step S3, the parasitic electrical voltage drop is calculated based on the voltage difference between the first supply voltage and the second supply voltage and the preset power supply push sensitivity coefficient. The parasitic electrical voltage drop is subtracted from the difference between the first detector voltage and the second detector voltage and then divided by the time difference between the two sampling times to calculate the pure thermal induced gain attenuation rate. Based on the pure thermal induced gain attenuation rate, the first detector voltage, and the first sampling time relative to the calibration RF pulse, the pure thermal induced gain attenuation rate is calculated. The time difference between the signal rising edge reaching 50% of the peak amplitude of the transmission reference time is used to calculate the reference equivalent detector voltage; in step S4, the deviation between the reference equivalent detector voltage and the preset target power reference voltage is calculated. When the deviation exceeds the preset convergence tolerance window, the control word step correction is calculated based on the preset detector sensitivity slope in the first iteration. In subsequent iterations, the local sensitivity slope is updated based on the difference between the reference equivalent detector voltage and the difference between the feedforward power control word in two adjacent iterations. The feedforward power control word is corrected using the local sensitivity slope and the process is repeated until the deviation falls into the preset convergence tolerance window. When the deviation falls into the preset convergence tolerance window, the current feedforward power control word is used as the final control word, and the final control word, the pure thermal induced gain attenuation rate, and the initial substrate temperature are associated and stored in a non-volatile storage medium.

[0053] like Figure 2As shown, the sampling time series during the flat-top period on the horizontal axis is distributed in 10μs steps within the range of 0μs to 80μs, and the corrected detector voltage on the vertical axis is quantized in 30mV steps within the range of 1760mV to 1940mV. The figure shows the correspondence between the saturated measured detector voltage, the first-order linear approximation reference line, and the equivalent trajectory derived by the second-order compensation. Within the sampling time series during the flat-top period of 20μs to 80μs, the saturated measured detector voltage monotonically decays from 1880mV at 20μs to 1784mV at 80μs, and the decay evolution exhibits a downward-bending nonlinear response characteristic. The first-order linear approximation reference line provides a linear reference within the sampling time series during the flat-top period of 20μs to 80μs, with its two ends connecting 1880mV at 20μs and 1784mV at 80μs, respectively. With a 4mV detection level, the measured saturated detection voltage is generally below the first-order linear approximation reference line in the 20μs to 80μs range. This reflects the gain drooping error and second-order curvature deviation caused by the nonlinearity of the transistor's thermal resistance and capacitance under saturated operating conditions. The equivalent trajectory calculated by secondary compensation extends backward from 1880mV at 20μs in the sampling time interval from 0μs to 20μs flat-top period. The corresponding reference equivalent detection voltage of 1914.67mV is obtained by reverse calculation at the 0μs transmission reference time. By introducing secondary thermal attenuation compensation related to the square of the time difference during the reverse calculation process, the gain drooping error caused by the nonlinearity of the transistor's thermal resistance and capacitance under saturated operating conditions is eliminated in situ, and the reference equivalent detection voltage corresponding to the zero-temperature-rise reference state is determined.

[0054] Example 5: During the pre-parameter calibration process before the wireless communication module is connected to the production line testing machine, regarding the electrical coupling relationship between the dynamic drop of the collector bias power supply pin of the RF power amplifier and the detector voltage response, the test system, in a resting bias state without RF power excitation, injects a small step disturbance voltage into the collector bias power supply pin through a programmable DC regulated power supply. The dual-channel high-speed analog-to-digital converter synchronously reads the voltage divider sampling value at the bias power supply pin and the coupled detector voltage at the output of the Schottky detector diode. Using the measured step voltage change at the power supply pin as the independent variable and the in-phase coupling level offset measured at the output of the detector diode as the dependent variable, the voltage ratio of the two is calculated, and multiple measurements are performed to obtain the average value. This calibrates the power supply push sensitivity coefficient under the current operating frequency band. The coefficient is written into the register of the baseband processing unit to remove the parasitic electrical voltage drop component from the total drop of the first and second detector voltages during subsequent pulse flat-top sampling.

[0055] The power supply push sensitivity coefficient, measured under resting bias without RF excitation, characterizes the pure parasitic electrical common-mode coupling ratio between the power supply network of the RF front-end printed circuit board and the cathode reference ground level of the Schottky detector diode and the bias sampling channel. When the RF power amplifier switches to large-signal transmission mode, a 10mV voltage drop occurs at the collector bias power supply pin. This voltage change affects the final detector output through two physical paths: first, the collector voltage change causes parasitic electrical level leakage to the detector output through the board-level ground impedance and the filter network; second, the power supply drop causes the transistor RF output to... The power amplitude changes, which in turn cause changes in the detector voltage. The collector push response curve of the RF power amplifier is linear near the target power of 26dBm, and the 10mV power supply drop is a minor disturbance relative to the 3380mV collector power supply reference. The above-mentioned board-level pure parasitic electrical coupling and the large-signal RF power push modulation are linearly superimposed in phase in the nanosecond to microsecond time domain. During factory calibration, the increment of the large-signal push level is converted and incorporated into the power supply push sensitivity coefficient, so that this coefficient corresponds to the parasitic electrical voltage drop component caused by the voltage drop of the power supply pin during the actual large-signal pulse peak.

[0056] During the step quantization process of loading the feedforward power control word into the baseband processing unit for closed-loop correction, considering the dimensional conversion relationship between the discrete integer characteristics of the feedforward power control word register and the continuous variation of RF power, the slope of the measured and calibrated detection voltage of the Schottky detector diode with output power near the target output power of 26dBm is... Baseband feedforward power control word unit step size The corresponding output power step is 0.25dB / LSB; the baseband processing unit multiplies the two physical parameters to obtain the detector voltage change corresponding to each step of the feedforward power control word, and uses it as the preset detector sensitivity slope. : When the reference equivalent detector voltage after eliminating transient self-heating effects exceeds the preset convergence tolerance window, the baseband processing unit calculates the voltage difference between the preset target power reference voltage and the reference equivalent detector voltage, using a preset detector sensitivity slope. As the conversion slope for the first iteration, the voltage difference is divided by Then, through discrete rounding, the integer step correction amount of the control word is determined, and the feedforward power control word is updated accordingly.

[0057] Example 6: During the calibration of the RF power amplifier for high and low temperature environments and component discreteness in the wireless communication module, before starting RF transmission, the test system reads the voltage divider sampling level of the negative temperature coefficient thermistor mounted next to the RF power amplifier package through the analog-to-digital conversion interface of the baseband processing unit. Combined with a piecewise linearized lookup table of the resistance-temperature characteristics of the temperature sensor pre-embedded in a non-volatile storage medium, the initial substrate temperature of the RF power amplifier is obtained. The initial base temperature is 25°C. When the initial base temperature is detected to exceed the operating boundary of -40°C to 85°C, the baseband processing unit generates a temperature over-limit protection interruption and suspends the subsequent calibration process.

[0058] Determine the reference time for calibrating the radio frequency pulse signal transmission. At that time, the baseband processing unit triggers the high-speed digital-to-analog converter to output the pulse baseband envelope. The first channel of the dual-channel high-speed analog-to-digital converter continuously oversamples the pulse envelope output by the directional coupler coupled to the detector at a sampling rate of 100MHz, forming a sequence of detector envelope samples arranged according to the sampling clock. After obtaining the flat-top period sample of the calibration RF pulse signal, the baseband processing unit determines the actual peak amplitude of the pulse from the sample sequence. Using 50% of the actual peak amplitude as the comparison threshold, the hardware numerical comparator compares the samples along the rising edge of the same pulse sequentially to locate the analog-to-digital converter sampling point where the rising edge first reaches the comparison threshold. This sampling point is then determined as the reference time for the transmission of the calibration RF pulse signal. The baseband processing unit records the timing position corresponding to the sampling point as 0μs, and determines the first sampling time from the same continuous sampling sequence according to the time offset corresponding to the preset delay register. 20μs, second sampling time For 80μs and the third sampling time The preset target power reference voltage is 1650mV, half of which is 825mV. This fixed half value does not replace 50% of the actual peak amplitude of the current calibration RF pulse signal for determining the transmission reference time. The transmission reference time is always determined by the rising edge sampling point corresponding to 50% of the measured peak amplitude of the current pulse.

[0059] During the cyclic optimization of the feedforward power control word and the inter-pulse electrothermal recovery process across iterations, the baseband processing unit employs a preset convergence tolerance window. Reference equivalent detector voltage Compared with the preset target power reference voltage When comparing, when the absolute deviation amount When the voltage is greater than 10mV, the preset detector sensitivity slope is retrieved from the memory during the first iteration. Based on this, the initial control word step correction amount is calculated.

[0060] In subsequent iterations, the baseband processing unit extracts the reference equivalent detector voltage for the current iteration period. and the reference equivalent detector voltage immediately preceding the previous iteration period At the same time, obtain the corresponding feedforward power control word and The local sensitivity slope is calculated based on the difference between the baseline equivalent detector voltage and the difference in the feedforward power control word between two adjacent iterations: When the RF power amplifier enters the large-signal gain compression nonlinear region, the baseband processing unit uses the slope of this local sensitivity as the denominator in the calculation of the current control word correction, resulting in: During the closed-loop cycle, before each time the baseband processing unit controls the RF power amplifier to transmit the next calibration RF pulse signal, it calls a hardware timer to load the cooling interval duration count value: Among them, the preset base recovery delay time Set to 200μs, preset thermal recovery coefficient Set to 150μs / (mV / μs), during the period when the hardware timer monotonically decrements from the loaded value to zero, the system blocks the baseband RF transmission path, allowing the transient heat accumulated in the transistor junction region to recover to the thermal state corresponding to the initial base temperature. After the count ends, the updated feedforward power control word is loaded and the next calibration RF pulse signal is transmitted. The above process is repeated until the absolute deviation is reached. The voltage is reduced to less than or equal to 10mV, the iteration is then terminated, and the current feedforward power control word is written as the final control word to a non-volatile storage medium.

[0061] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A method for calibrating radio frequency power in a wireless module, characterized in that, Includes the following steps: Step S1: Obtain the initial substrate temperature of the RF power amplifier; the baseband processing unit outputs the initial feedforward power control word to control the transmit calibration RF pulse signal. Step S2: During the flat-top period of the calibration RF pulse signal, the first detection voltage and the second detection voltage are acquired at the first sampling time and the second sampling time, respectively, and the first supply voltage and the second supply voltage at the collector bias power supply pin are acquired simultaneously. Step S3: Calculate the parasitic electrical voltage drop based on the voltage difference between the first and second supply voltages and the preset power supply push sensitivity coefficient. Subtract the parasitic electrical voltage drop from the difference between the first and second detector voltages and divide by the time difference between the two sampling times to calculate the pure thermal induced gain attenuation rate. Calculate the reference equivalent detector voltage based on the pure thermal induced gain attenuation rate, the first detector voltage, and the time difference between the first sampling time and the transmission reference time when the rising edge of the calibration RF pulse signal reaches 50% of its peak amplitude. Step S4: Calculate the deviation between the reference equivalent detector voltage and the preset target power reference voltage. When the deviation exceeds the preset convergence tolerance window, calculate the control word step correction based on the preset detector sensitivity slope in the first iteration. In subsequent iterations, update the local sensitivity slope based on the difference between the reference equivalent detector voltage and the difference between the feedforward power control word in two adjacent iterations. Use the local sensitivity slope to correct the feedforward power control word and repeat until the deviation falls into the preset convergence tolerance window. When the deviation falls into the preset convergence tolerance window, use the current feedforward power control word as the final control word, and store the final control word, the pure thermal induced gain attenuation rate, and the initial substrate temperature in a non-volatile storage medium.

2. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, Step S3 includes calculating the reference equivalent detector voltage, which involves sub-step S31, multiplying the pure thermal induced gain attenuation rate by the time difference to obtain the temperature rise compensation voltage and adding it to the first detector voltage; Step S4 includes sub-step S41, calculating the difference between the preset target power reference voltage and the current iteration reference equivalent detector voltage, dividing it by the local sensitivity slope, and rounding it to obtain the control word step correction amount, which is then added to the current iteration feedforward power control word to generate the feedforward power control word updated in the next iteration.

3. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, Step S2 includes sub-step S21, which configures the first sampling time to lag behind the calibration RF pulse signal transmission reference time, and the waiting time is greater than the electrical setup time constant corresponding to the discharge of the decoupling capacitor in the RF power amplifier bias power supply network; and configures the first sampling time to lag behind the flat-top period start edge by more than or equal to 10 μs, and the time difference between the first sampling time and the second sampling time is in the range of 20 μs to 80 μs.

4. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, Step S2 further includes sub-step S22, setting a third sampling time at the midpoint between the first and second sampling times during the flat-top period of the calibration RF pulse signal, and acquiring the third detection voltage; Step S3 further includes sub-step S32, calculating the second-order curvature deviation between the average voltage value obtained from the first and second detection voltages and the third detection voltage; when the second-order curvature deviation exceeds the linearity judgment threshold, determining the secondary thermal attenuation compensation factor, and superimposing the secondary compensation term determined by multiplying the secondary thermal attenuation compensation factor by the square of the time difference when calculating the reference equivalent detection voltage, so as to eliminate the gain drooping bending error caused by the nonlinearity of transistor thermal resistance and thermal capacitance under saturated operating conditions.

5. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, In step S1, the initial substrate temperature is obtained by sensing with a thermistor located next to the RF power amplifier. Before calculating the pure thermal induced gain attenuation rate in step S3, the calibration method also includes a sub-step S33, which obtains the temperature difference between the current initial substrate temperature and the factory calibration reference temperature, calculates the static level drift based on the calibrated dark voltage temperature drift slope of the detector diode, and subtracts the static level drift from the measured first and second detector voltages respectively to eliminate the DC zero-point level offset caused by the external detector diode with temperature changes.

6. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, Step S1 also includes sub-step S11, which involves calibrating power control parameters at multiple discrete frequency points within the target operating frequency band, obtaining the final control word and pure thermal induced gain attenuation rate corresponding to each discrete frequency point, comparing the pure thermal induced gain attenuation rates between each discrete frequency point, and determining that the discrete frequency point is in an RF impedance mismatch state when the pure thermal induced gain attenuation rate of a specific discrete frequency point exceeds the preset deviation threshold of the benchmark average value under the same power level. The output bias adjustment command lowers the programmable static bias current of the RF power amplifier by one level, and recalibrates after the adjustment to update the final control word.

7. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, Step S4 also includes sub-step S42: when the wireless module is in the burst data frame transmission state, the final control word stored in the non-volatile storage medium is called to configure the transmission power, and the pure thermal induced gain attenuation rate is read from the associated storage; the corresponding modulation peak-to-average ratio (PAPR) conversion factor is obtained according to the modulation coding method of the current transmission service, and the pure thermal induced gain attenuation rate is multiplied by the PAPR conversion factor to generate the modulation transmission pre-tilt slope; a baseband gain digital pre-tilt compensation sequence that increases linearly with time according to the modulation transmission pre-tilt slope is superimposed in the burst data frame transmission time slot to offset the intra-pulse gain attenuation of the RF power amplifier caused by the junction temperature rise during burst transmission.

8. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, Step S3 also includes sub-step S34, where when the target output power is in the small-signal linear region of the amplifier, the difference between the measured first detector voltage and the second detector voltage is equal to zero, the calculated pure thermal induced gain attenuation rate is zero, the calculated reference equivalent detector voltage is equal to the first detector voltage, and the cooling interval between two consecutive transmission calibration RF pulse signals is equal to the substrate recovery delay time.

9. The radio frequency power calibration method for a wireless module according to claim 1, characterized in that, The associated storage in step S4 includes sub-step S43, which combines the final control word, the pure thermal induced gain attenuation rate, the initial substrate temperature, the power supply push sensitivity coefficient, and the corresponding target operating frequency point into an RF power calibration mapping table entry and writes it into a non-volatile storage medium so that the wireless module can call the configuration when transmitting signals in different frequency bands.

Images