Ai-based power amplifier parameter predition and radio frequency calibration optimization
Patent Information
- Authority / Receiving Office
- TW · TW
- Patent Type
- Applications
- Current Assignee / Owner
- ARCADYAN
- Filing Date
- 2025-01-06
- Publication Date
- 2026-07-16
AI Technical Summary
Manual calibration of power amplifiers in wireless network communication products is time-consuming, slowing down production speed and impacting yield.
An artificial intelligence-based method for power amplifier parameter calibration, where initial manual calibration of a subset of parameters is followed by using an AI background calculation program to predict and calculate the remaining parameters, thereby reducing manual intervention.
This approach significantly increases production speed and yield by accurately calibrating power amplifiers within preset specifications with minimal human effort.
Smart Images
Figure TWG2TA001067784_001 
Figure TWG2TA001067784_002 
Figure TWG2TA001067784_003
Abstract
Description
[Technical Field]
[0001] This invention relates to a power amplifier parameter prediction and RF calibration optimization technology for network communication products, and in particular, a power amplifier parameter prediction and RF calibration optimization technology for network communication products based on artificial intelligence. [Previous Technology]
[0002] The wireless network communication products on the market mainly include wireless communication components for transmitting and receiving wireless signals, especially radio frequency communication components for transmitting and receiving radio frequency wireless signals. The transmitter part includes a power amplifier, which is used to amplify the power of the radio frequency wireless signal to be transmitted.
[0003] Upon completion of manufacturing, the power amplifier within the aforementioned radio frequency communication component requires parameter calibration to meet product specifications. Currently, the industry primarily employs manual methods for power amplifier parameter calibration, using a trial-and-error approach. However, this manual method consumes considerable time, thus slowing down production speed and impacting yield.
[0004] Therefore, an efficient method for calibrating power amplifier parameters is needed to overcome the above problems, thereby increasing production speed and yield.
[0005] According to the required specifications of wireless network communication products, each parameter of the power amplifier has a corresponding preset standard power output value range. The purpose of power amplifier parameter calibration is to adjust the parameter value of each parameter so that the power value of the signal output by the radio frequency communication components containing the power amplifier conforms to the preset standard power output value range corresponding to the parameter being calibrated.
[0006] Figure 1 is a flowchart of a method for manually calibrating power amplifier parameters according to the prior art.
[0007] Please refer to Figure 1. First, in step A, manually calibrate the parameter values of N parameters (items) of the power amplifier to obtain the calibration parameter values of N parameters, where N is the total number of parameters. Next, in step B, load the calibration parameter values of the N parameters into the power amplifier to complete the calibration of the power amplifier parameters.
[0008] In the prior art, manual calibration of power amplifiers can be performed using a test computer and a standalone tester (OBT), with the standalone tester connected to the test computer. The RF communication component containing the power amplifier (as a device under test (DUT)) can be first installed on the standalone tester, and then the operator (OP) controls the RF communication component to emit RF signals through the operating interface of the test computer. The standalone tester can receive the RF signals emitted by the RF communication component, measure the power output value of the RF signal, and then transmit the power output value information back to the test computer.
[0009] According to the trial and error method, the operator can first calibrate the first parameter among the N parameters of the power amplifier, which has a corresponding first preset standard power output value range. The operator can manually or automatically select and apply the parameter value of the first parameter through the test computer, and then confirm whether the power output value of the radio frequency signal emitted by the radio frequency communication element falls within the first preset standard power output value range. If so, the parameter value of the first parameter selected and applied is the calibration parameter value; if not, other parameter values of the first parameter are selected and applied again until the calibration parameter value of the first parameter corresponding to the power output value of the first preset standard power output value range is found.
[0010] By repeating the above trial-and-error steps, the calibration of all parameters of the power amplifier can be completed, and all calibration parameter values can be obtained. Finally, all calibration parameter values are loaded into the power amplifier to complete the power amplifier parameter calibration. For example, the calibration parameter values can be loaded into the power amplifier via a test computer, or they can be loaded into the power amplifier via a standalone tester.
[0011] However, the manual calibration of power amplifiers in the prior art requires a lot of time, which slows down production speed and affects yield. Therefore, the present invention proposes an artificial intelligence-based power amplifier parameter prediction and RF calibration optimization technology for network communication products to shorten the calibration time of power amplifiers, speed up production, and improve yield. [Summary of the Invention]
[0012] To effectively solve the above problems, according to one aspect of the present invention, a calibration method for power amplifier parameters is proposed, wherein the power amplifier has N parameters, each of the N parameters having a preset standard power output value range, the method comprising: step 1, manually calibrating the parameter values of N1 of the N parameters to obtain calibration parameter values of the N1 parameters; step 2, inputting the calibration parameter values of the N1 parameters into an artificial intelligence background calculation program to calculate the calibration parameter values of the remaining N2 parameters by the artificial intelligence background calculation program, wherein N2 = N - N1; and step 3, loading the calibration parameter values of the N parameters into the power amplifier to complete the calibration of the power amplifier parameters.
[0013] According to an embodiment of the present invention, preferably, step 1 includes: step 1-1, manually inputting the parameter value of a first parameter among the N1 parameters into the power amplifier; step 1-2, reading the power output value of the power amplifier; step 1-3, confirming whether the power output value of the power amplifier is within the range of a preset standard power output value corresponding to the first parameter. If yes, the input parameter value is the calibration parameter value of the first parameter, and the calibration of another parameter continues. If no, steps 1-1 to 1-3 are repeated; and step 1-4, obtaining the calibration parameter values of the N1 parameters by repeating steps 1-1 to 1-3.
[0014] According to an embodiment of the present invention, preferably, step 2 includes: step 2-1, providing a first power amplifier, the first power amplifier having the N parameters, the first power amplifier having been manually calibrated, inputting the calibration parameter values of the N parameters of the first power amplifier and the corresponding power output values to an artificial intelligence server to establish a first parameter prediction model, and loading the first parameter prediction model into the artificial intelligence background calculation program; and step 2-2, inputting the calibration parameter values of the N1 parameters of the power amplifier into the first parameter prediction model of the artificial intelligence background calculation program, so as to calculate the calibration parameter values of the remaining N2 parameters of the power amplifier by means of the first parameter prediction model.
[0015] According to an embodiment of the present invention, preferably, the method further includes: step 4, inputting the calibration parameter values and corresponding power output values of the N parameters of the power amplifier to the artificial intelligence server to establish a second parameter prediction model, and loading the second parameter prediction model into the artificial intelligence background computing program to update the parameter prediction model of the artificial intelligence background computing program.
[0016] According to an embodiment of the present invention, preferably, the artificial intelligence background computing program is loaded in an artificial intelligence edge computing box, which is connected to a test computer.
[0017] According to an embodiment of the present invention, preferably, the method for establishing a parameter prediction model using the artificial intelligence server includes: linear regression algorithm, random forest regression algorithm, gradient boosting regression algorithm, support vector regression algorithm, neural network-like algorithm, or a combination thereof.
[0018] According to an embodiment of the present invention, preferably, step 1-1 includes: using a stand-alone tester connected to a test computer; connecting a radio frequency communication element including the power amplifier to the stand-alone tester; and manually inputting the parameter value of the first parameter among the N1 parameters to the power amplifier via the test computer and the stand-alone tester to control the radio frequency communication element to transmit a radio frequency signal.
[0019] According to an embodiment of the present invention, preferably, step 1-2 includes: using a stand-alone tester to receive the radio frequency signal and read the power value of the radio frequency signal, wherein the power value is the power output value of the power amplifier corresponding to the parameter value of the first parameter.
[0020] According to an embodiment of the present invention, preferably, steps 1-3 include: using the stand-alone tester to transmit the power value of the radio frequency signal back to the test computer; and using the test computer to confirm whether the power value of the radio frequency signal is within the range of a preset standard power output value corresponding to the first parameter.
[0021] According to another aspect of the present invention, a radio frequency communication element is provided, comprising: a power amplifier calibrated by the calibration method described above.
[0022] The method for calibrating power amplifier parameters based on artificial intelligence of the present invention can achieve the effects of increasing production speed and yield compared with the prior art.
[0023] In order to enable those skilled in the art to understand the purpose, features and effects of the present invention, the present invention will be described in detail below with reference to the following specific embodiments and accompanying drawings.
Implementation Method
[0024] Figure 2 is a flowchart of a method for calibrating power amplifier parameters based on artificial intelligence according to an embodiment of the present invention; Figure 3 is a flowchart of step 1 in the method of Figure 2; and Figure 4 is a flowchart of step 2 in the method of Figure 2.
[0025] Referring to Figures 2 and 3, firstly, in step 1, as with the methods and procedures of the prior art, only the parameter values of N1 out of the N parameters of the power amplifier are manually calibrated, where N is the total number of parameters, and N1 < N. According to embodiments of the present invention, preferably, N1 ≤ N / 5, N1 ≤ N / 10, N1 ≤ N / 15, or N1 ≤ N / 20, but the present invention is not limited thereto.
[0026] Step 1 in Figure 2 may further include steps 1-1 to 1-4 as shown in Figure 3.
[0027] Referring to Figure 3, firstly, in step 1-1, manually input the parameter value of the first parameter out of N1 parameters into the power amplifier. Secondly, in step 1-2, read the power output value of the power amplifier. Thirdly, in step 1-3, confirm whether the power output value of the power amplifier is within the preset standard power output value range corresponding to the first parameter. If yes, the input parameter value is the calibration parameter value of the first parameter; otherwise, repeat steps 1-1 to 1-3. Finally, in step 1-4, obtain the standard parameter values of the N1 parameters by repeating steps 1-1 to 1-3.
[0028] According to an embodiment of the present invention, step 1-1 may include: using a stand-alone tester, which may be wired and / or wirelessly connected to a test computer; connecting a radio frequency communication element including a power amplifier to the stand-alone tester; and manually inputting the parameter value of the first parameter among N1 parameters to the power amplifier via the test computer and the stand-alone tester to control the radio frequency communication element to transmit radio frequency signals.
[0029] The stand-alone tester used in this invention can be a commercially available industry standard product, such as: the IQxel-M™ multi-DUT and multi-communication wireless connection test system provided by LITEPOINT, the WT-428 wireless connection tester provided by LinkWen, the E6640A wireless tester provided by Keysight Technologies, etc., but this invention is not limited to these.
[0030] According to an embodiment of the present invention, steps 1-2 may include: using a stand-alone tester to receive a radio frequency signal transmitted by a radio frequency communication element and reading the power value of the radio frequency signal. At this time, the power value of the radio frequency signal is the power output value of the power amplifier corresponding to the parameter value of the first parameter.
[0031] According to an embodiment of the present invention, steps 1-3 may include: transmitting the power value of the radio frequency signal back to the test computer using a stand-alone tester; and using the test computer to confirm whether the power value of the radio frequency signal is within the range of the preset standard power output value corresponding to the first parameter.
[0032] Next, please refer to Figures 2 and 4. After completing the manual calibration of the amplifier's N1 parameters in step 1, in step 2, the calibration parameter values of the N1 parameters are input into the AI background calculation program (AI Daemon) so that the AI background calculation program can calculate the calibration parameter values of the remaining N2 parameters, where N2 = N - N1.
[0033] The AI background computing program used in this invention can be installed in commercially available industry standard products such as: the RISC-based Box computer provided by Advantech, the AIX-600 and AIX-800 advanced interactive execution systems provided by System Electronics, the BOXER-8220AI embedded computer provided by AAEON, etc., but this invention is not limited to these.
[0034] Step 2 in Figure 2 may further include steps 2-1 to 2-2 as shown in Figure 4.
[0035] Referring to Figure 4, firstly, in step 2-1, a first power amplifier is provided. The first power amplifier has N parameters and has been manually calibrated. The calibration parameter values of the N parameters of the manually calibrated first power amplifier and the corresponding power output values are input to an artificial intelligence server to establish a first parameter prediction model, and the first parameter prediction model is loaded into an artificial intelligence background calculation program. In this invention, the relevant information data (calibration parameter values and / or corresponding power output values) of the manually calibrated first power amplifier are used to provide the basis for training and establishing a first parameter prediction model by an artificial intelligence server, so as to calculate the calibration parameter values of the remaining N2 parameters based on the calibration parameter values of the N1 parameters of the (to be calibrated) power amplifier. In step 2-2, the calibration parameter values of the N1 parameters of the power amplifier are input to the first parameter prediction model of the artificial intelligence background calculation program, so as to calculate the calibration parameter values of the remaining N2 parameters of the power amplifier by means of the first parameter prediction model. Here, the first parameter prediction model of the AI background computing program can not only calculate the calibration parameter values of the remaining N2 parameters, but also calculate the power output values corresponding to the calibration parameter values of the remaining N2 parameters. In this invention, the AI server can be connected wired and / or wirelessly to a computer and an edge AI computing box, wherein the edge AI computing box is equipped with the AI background computing program.
[0036] According to another embodiment of the present invention, in step 2-2, the calibration parameter values of the N1 parameters of the power amplifier and the corresponding power output values can be input to the first parameter prediction model of the artificial intelligence background computing program to improve the computing (prediction) accuracy of the first parameter prediction model of the artificial intelligence background computing program.
[0037] According to an embodiment of the present invention, the method for establishing a parameter prediction model for an artificial intelligence server may include: linear regression algorithm, random forest regression algorithm, gradient boosting regression algorithm, support vector regression algorithm, neural network algorithm, or a combination thereof.
[0038] Finally, please refer to Figure 2. In step 3, the calibration parameter values of N parameters are loaded into the power amplifier to complete the calibration of the power amplifier parameters.
[0039] For example, Table 1 below shows a comparison between the power output value corresponding to the calibration parameter values of 20 parameter items of the power amplifier calculated by the artificial intelligence background calculation program of the present invention and the power output value actually measured by a single-machine tester after the aforementioned calculated (predicted) calibration parameter values of 20 parameter items are loaded into the power amplifier.
[0040] [Table 1] parameter project Preset standard power output range Calculated power output value Actual measured power output value error 01 18~22 18.55 18.46 0.5% 02 18~22 18.01 18.50 -2.6% 03 18~22 19.81 18.43 7.5% 04 18~22 18.47 18.44 0.2% 05 15~19 16.42 17.22 -4.6% 06 15~19 16.98 17.11 -0.8% 07 15~19 16.35 17.09 -4.3% 08 15~19 16.75 17.25 -2.9% 09 15~19 16.77 17.45 -3.9% 10 15~19 16.92 17.42 -2.9% 11 15~19 16.41 17.29 -5.1% 12 15~19 16.89 17.30 -2.4% 13 15~19 16.98 17.36 -2.2% 14 15~19 17.42 17.50 -0.5% 15 15~19 16.80 17.39 -3.4% 16 15~19 16.43 17.18 -4.4% 17 15~19 16.70 17.81 -6.2% 18 15~19 17.02 17.87 -4.8% 19 15~19 17.20 17.96 -4.2% 20 15~19 17.17 17.77 -3.4%
[0041] Table 1 lists the parameter item codes in the first column "Parameter Items"; the second column "Preset Standard Power Output Value Range" lists the preset standard power output value ranges corresponding to each parameter item, in any unit; the third column "Calculated Power Output Value" lists the power output values corresponding to the calibration parameter values of each parameter item of the power amplifier calculated by the artificial intelligence background calculation program of the present invention, in any unit; the fourth column "Actually Measured Power Output Value" lists the power output values actually measured by a single-machine tester after the aforementioned calculated calibration parameter values of each parameter item are loaded into the power amplifier, in any unit; the fifth column "Error" lists the error value between the calculated power output value and the actual measured power output value.
[0042] Clearly, for the 20 parameters of the power amplifier, both the "calculated power output value" and the "actually measured power output value" fall within the "preset standard power output value range," with a maximum error of no more than 8%. In other words, the AI background calculation program of this invention has high accuracy, small error, and is reliable in calculating (predicting) the calibration parameter values and corresponding power output values of the power amplifier. The actual power output value of the power amplifier loaded with the calculated calibration parameter values also falls within the preset standard power output value range corresponding to each parameter. That is to say, the power amplifier assisted by the AI background calculation program of this invention for parameter calibration meets the preset specifications after calibration.
[0043] According to an embodiment of the present invention, the method for calibrating power amplifier parameters based on artificial intelligence may further include step 4: inputting the calibration parameter values of N parameters of the calibrated power amplifier and the corresponding power output values to an artificial intelligence server to establish a second parameter prediction model, and loading the second parameter prediction model into an artificial intelligence background computing program to update the parameter prediction model of the artificial intelligence background computing program. Repeating this process can improve the prediction accuracy of the parameter prediction model.
[0044] The following is an embodiment of the complete calibration process of the method for calibrating power amplifier parameters based on artificial intelligence according to the present invention.
[0045] First, the operator obtains the calibration parameter values and corresponding power output values of the eight parameters (e.g., parameter numbers 01 to 08) of the power amplifier to be calibrated and the corresponding power output values through a test computer using a trial-and-error method. The power amplifier has been installed in the radio frequency communication element, which is connected to a stand-alone tester, which is connected to the test computer. The power amplifier in this embodiment has 60 parameters that need to be calibrated, but the present invention is not limited thereto.
[0046] Next, the operator inputs the calibration parameter values of the 8 parameters (parameter numbers 01 to 08) obtained by manual testing into the edge AI computing box via the test computer, so as to obtain the calibration parameter values of the remaining 52 parameters (e.g., parameter numbers 09 to 60) of the power amplifier to be calibrated and their corresponding power output values through the calculation of the AI computing box. The parameter prediction model has been established in the edge AI computing box, and the edge AI computing box is connected to the test computer.
[0047] Next, the operator loads the calibration parameter values of 60 parameters (parameter numbers 01~60) (including 8 calibration parameter values obtained by manual testing and 52 calibration parameter values obtained by edge AI computing box) into the power amplifier via the test computer, thus completing the parameter calibration of the power amplifier in the radio frequency communication element.
[0048] Next, the operator verifies the RF power output value of the calibrated RF communication component through the test computer, and creates a power output value table, which is stored in the test computer. The power output value table includes at least the calibration parameter values and their corresponding power output values.
[0049] Finally, the power output value table is uploaded to and stored in the artificial intelligence server via the test computer. Based on the data in the power output value table and preset conditions, new parameter prediction models are established periodically or irregularly to update the parameter prediction model in the edge artificial intelligence computing box operation, thereby improving and enhancing the operation (prediction) accuracy of the artificial intelligence computing box. The artificial intelligence server is connected to the edge artificial intelligence computing box, or is connected to the edge artificial intelligence computing box via the test computer.
[0050] The above description illustrates the implementation of the present invention through specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification.
[0051] The above description is only a preferred embodiment of the present invention and is not intended to limit the scope of the present invention; any equivalent changes or modifications made without departing from the spirit disclosed in the present invention should be included within the scope of the following patent. [Simplified Explanation of the Diagram]
[0052] Figure 1 is a flowchart of a method for manually calibrating power amplifier parameters according to the prior art; Figure 2 is a flowchart of a method for calibrating power amplifier parameters based on artificial intelligence according to an embodiment of the present invention; Figure 3 is a flowchart of step 1 in the method of Figure 2; and Figure 4 is a flowchart of step 2 in the method of Figure 2.
Claims
1. A method for calibrating power amplifier parameters, wherein, The power amplifier has N parameters, each of which has a preset standard power output value range. The method includes: Step 1, manually calibrating the parameter values of N1 parameters out of the N parameters to obtain calibration parameter values for the N1 parameters; Step 2, inputting the calibration parameter values of the N1 parameters into an artificial intelligence background calculation program to calculate the calibration parameter values of the remaining N2 parameters, where N2 = N - N1; and Step 3, loading the calibration parameter values of the N parameters into the power amplifier to complete the calibration of the power amplifier parameters, wherein Step 1 includes: Step 1-1, manually inputting the parameter value of a first parameter out of the N1 parameters into the power amplifier; Step 1-2, reading the power output value of the power amplifier; Steps 1-3: Confirm whether the power output value of the power amplifier is within the preset standard power output value range corresponding to the first parameter. If yes, the input parameter value is the calibration parameter value of the first parameter, and continue to calibrate another parameter. If no, repeat steps 1-1 to 1-3; and Step 1-4: Obtain the calibration parameter values of the N1 parameters by repeating steps 1-1 to 1-3.
2. The method as described in request item 1, wherein, Step 1-1 includes: using a standalone tester connected to a test computer; connecting an RF communication element containing the power amplifier to the standalone tester; and manually inputting the parameter value of the first parameter among the N1 parameters to the power amplifier via the test computer and the standalone tester to control the RF communication element to transmit an RF signal.
3. The method as described in claim 2, wherein, Step 1-2 includes: using a standalone tester, receiving the radio frequency signal and reading the power value of the radio frequency signal, which is the power output value of the power amplifier corresponding to the parameter value of the first parameter.
4. The method as described in request item 3, wherein, Steps 1-3 include: using the standalone tester to transmit the power value of the radio frequency signal back to the test computer; and using the test computer to confirm whether the power value of the radio frequency signal is within the range of the preset standard power output value corresponding to the first parameter.
5. A method for calibrating power amplifier parameters, wherein, The power amplifier has N parameters, each of which has a preset standard power output value range. The method includes: Step 1, manually calibrating the parameter values of N1 of the N parameters to obtain calibration parameter values for the N1 parameters; Step 2, inputting the calibration parameter values of the N1 parameters into an artificial intelligence background calculation program to calculate the calibration parameter values of the remaining N2 parameters, where N2 = N - N1; and Step 3, loading the calibration parameter values of the N parameters into the power amplifier to complete the calibration of the power amplifier parameters, wherein Step 2 includes: Step 2-1: Provide a first power amplifier having the N parameters. The first power amplifier has been manually calibrated. Input the calibration parameter values of the N parameters of the first power amplifier and the corresponding power output values into an artificial intelligence server to establish a first parameter prediction model. Load the first parameter prediction model into the artificial intelligence background calculation program. Step 2-2: Input the calibration parameter values of the N1 parameters of the power amplifier into the first parameter prediction model of the artificial intelligence background calculation program to calculate the calibration parameter values of the remaining N2 parameters of the power amplifier using the first parameter prediction model.
6. The method as described in claim 5, further comprising: Step 4: Input the calibration parameter values of the N parameters of the power amplifier and the corresponding power output values into the artificial intelligence server to establish a second parameter prediction model, and load the second parameter prediction model into the artificial intelligence background computing program to update the parameter prediction model of the artificial intelligence background computing program.
7. The method as described in request item 5, wherein, The AI background computing program is loaded into an AI edge computing box, which is connected to a test computer.
8. The method as described in request item 5, wherein, Methods for establishing parameter prediction models using this artificial intelligence server include: linear regression algorithm, random forest regression algorithm, gradient boosting regression algorithm, support vector regression algorithm, neural network-like algorithm, or combinations thereof.
9. A radio frequency communication element comprising: a power amplifier calibrated by any one of claims 1 to 8.