WO2017092399A1 - 数字预失真表生成方法、装置及数字预失真系统 - Google Patents

数字预失真表生成方法、装置及数字预失真系统 Download PDF

Info

Publication number
WO2017092399A1
WO2017092399A1 PCT/CN2016/094674 CN2016094674W WO2017092399A1 WO 2017092399 A1 WO2017092399 A1 WO 2017092399A1 CN 2016094674 W CN2016094674 W CN 2016094674W WO 2017092399 A1 WO2017092399 A1 WO 2017092399A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
optimization
dpd
digital predistortion
module
Prior art date
Application number
PCT/CN2016/094674
Other languages
English (en)
French (fr)
Inventor
梁忠杰
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2017092399A1 publication Critical patent/WO2017092399A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/38—Synchronous or start-stop systems, e.g. for Baudot code
    • H04L25/40—Transmitting circuits; Receiving circuits
    • H04L25/49—Transmitting circuits; Receiving circuits using code conversion at the transmitter; using predistortion; using insertion of idle bits for obtaining a desired frequency spectrum; using three or more amplitude levels ; Baseband coding techniques specific to data transmission systems
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
    • H04B1/02—Transmitters
    • H04B1/04—Circuits
    • H04B1/0475—Circuits with means for limiting noise, interference or distortion

Definitions

  • the present application relates to, but is not limited to, the field of mobile communications, and in particular, to a digital predistortion table generating method and apparatus, and a digital predistortion system.
  • DPD Digital Pre-Distortion
  • the data in the table is only collected. It is possible to select several discrete conditions among these continuously changing factors. How to select such discrete conditions, the DPD data used for pre-distortion under each operating condition may not be available in the DPD table when the device is running. How the data not available in the DPD table is calculated by the limited data in the table. How to ensure the accuracy, validity and pre-distortion performance of the collected and calculated DPD data is a variety of DPD modules of related technologies and The focus of the algorithm.
  • the related DPD table generation method only considers one influencing factor, such as predistortion training and table generation based only on amplifier characteristics or feedback signals, and does not detect and evaluate the predistortion effect of data in the DPD table;
  • the related DPD table generation method considers the influencing factors to be single, and does not detect and evaluate the data of the DPD table, and the effect is poor.
  • the present application provides a digital predistortion table generation method and device, and a digital predistortion system, to solve the related factors of the DPD table of the related technology, and does not detect the data of the DPD table. Evaluate the problem of poor digital predistortion.
  • the application provides a DPD table generation method, including:
  • the digital predistortion effect detection and optimization are performed on the optimization data, and the optimization result is used as the model coefficient of the nonlinear mathematical model;
  • a DPD table is generated from the model coefficients using a full range of data generation methods.
  • generating the data collection use case according to the evaluation data and the full range data generation method comprises: analyzing the evaluation data by using a clustering and fitting method, and obtaining effective data collection by combining the limited values of all the working parameters.
  • Use cases and full range of data generation methods comprises: analyzing the evaluation data by using a clustering and fitting method, and obtaining effective data collection by combining the limited values of all the working parameters.
  • obtaining the optimization data includes: performing digital predistortion testing on the radio frequency signal of the device to be processed according to the data collection use case; displaying the digital predistorted radio frequency signal in a spectrum form, and sequentially scanning the frequency to obtain a set of power values, Generate optimization data.
  • the digital pre-distortion effect detection and optimization of the optimization data is performed, and the optimization result is used as a model coefficient of the nonlinear mathematical model, and generating the DPD table includes: comparing the pre-distorted signal with the standard signal, Detect the DPD effect until the optimal value of the power difference is found.
  • the imaginary part of the DPD is fixed first, the real part is traversed, the optimal real part data is determined, and the real part of the DPD is fixed to the optimal real part.
  • the detection and optimization of the digital predistortion effect of the optimized data includes: large-step optimization and small-step optimization, and looping to the optimal value.
  • the application further provides a computer readable storage medium storing computer executable instructions that are implemented when the computer executable instructions are executed.
  • the application provides a DPD table generating device, which includes an evaluation module, an acquisition module, an optimization module, and a generation module, where
  • the evaluation module is configured to perform, according to the nonlinear mathematical model of the device to be processed, all the working parameters affecting the DPD table of the device to be processed, using the collection module to perform full traversal data collection for all working parameters of the product sample of the device to be processed, and obtain evaluation data according to Evaluation data generation data collection use case and full range data generation method;
  • the acquisition module is configured to use the data collection use case to collect data from the processing device to obtain optimization data;
  • the optimization module is set to detect and optimize the digital predistortion effect of the optimization data, and the optimization result is used as the model coefficient of the nonlinear mathematical model;
  • the generation module is configured to generate the DPD table based on the model coefficients using the full range data generation method.
  • the evaluation module is configured to analyze the evaluation data by using a clustering and fitting method, and obtain an effective data collection use case and a full range data generation method by combining the limited values of all the working parameters.
  • the acquisition module is configured to perform digital predistortion testing on the radio frequency signal of the device to be processed according to the data collection use case; display the digital predistortion radio frequency signal in a spectrum form, and perform a stepwise scan to obtain a set of power values to generate a homing Excellent data.
  • the optimization module is configured to compare the pre-distorted signal with the standard signal, and detect the DPD effect until the power difference optimal value is found.
  • the DPD imaginary part is fixed first. The traversal, determine the optimal real data, fix the real part of the DPD to the optimal real data, perform imaginary traversal, determine the optimal imaginary data, and cycle to the optimal value.
  • the optimization module is further configured to alternately perform large step optimization and small step optimization, and cycle to an optimal value.
  • the present application also provides a DPD system including a DPD table generated by the DPD table generating apparatus provided by the present application.
  • the present application provides a method for generating a DPD table, considering all working parameters of a DPD table that affects a device to be processed, using statistical methods to evaluate and analyze all working parameters, and obtaining a combination of influencing factors of the number of acquisitions, using these use cases.
  • the number of acquisitions solves the problem of discrete condition selection.
  • the use of automation also ensures the efficiency of data collection and generation.
  • the algorithm of the related technology is improved in the acquisition algorithm and the complete DPD data generation algorithm.
  • the DPD table generation method of the related technology whether it is offline writing DPD The table, or the DPD table obtained in real-time training, will not perform pre-distortion effect detection and optimization on the data in the table.
  • the DPD table generation method given in this application has the data in the premise of detecting and optimizing the pre-distortion effect. After the acquisition, the data is used to enable the DPD, and a receiving and transmitting link is configured and connected to detect the signal of the receiving end device, and the received DPD data is further traversed, verified and optimized by the receiving signal, thereby ensuring the DPD table.
  • the data is the data after the pre-distortion effect is optimized, so that the pre-distortion effect achieved by the data in the DPD table in the present application is better than the pre-distortion effect achieved by the data obtained by the related art DPD table generation method.
  • FIG. 1 is a schematic structural diagram of a DPD table generating apparatus according to a first embodiment of the present application
  • FIG. 2 is a flowchart of a method for generating a DPD table according to a second embodiment of the present application
  • FIG. 3 is a schematic structural diagram of a DPD system according to a third embodiment of the present application.
  • the DPD table generating apparatus 1 is a schematic structural diagram of a DPD table generating apparatus according to a first embodiment of the present application.
  • the DPD table generating apparatus 1 provided by the present application includes an evaluation module 11, an acquisition module 12, and optimization. Module 13 and generation module 14, wherein
  • the evaluation module 11 is configured to perform, according to the nonlinear mathematical model of the device to be processed, all the working parameters affecting the DPD table of the device to be processed, using the collection module 12 to perform full traversal data collection for all working parameters of the product samples of the device to be processed, and obtain evaluation data. Generating a data collection use case and a full range of data generation method according to the evaluation data, for use by the acquisition module 12 and the generation module 14;
  • the acquisition module 12 is configured to use the data acquisition use case obtained by the evaluation module 11 to treat the device Collecting data, obtaining optimization data, and feeding back to the optimization module 13;
  • the optimization module 13 is configured to perform detection and optimization of the digital predistortion effect on the optimization data collected by the acquisition module 12, and the optimization result is used as a model coefficient of the nonlinear mathematical model;
  • the generation module 14 is arranged to generate a DPD table based on the model coefficients using the full range data generation method obtained by the evaluation module 11.
  • the DPD table generating apparatus 1 in the above embodiment further includes: a calling module, configured to: after generating the DPD table, store the DPD table, and search for a corresponding DPD table according to real-time working parameters of the device to be processed and The output is DPD.
  • a calling module configured to: after generating the DPD table, store the DPD table, and search for a corresponding DPD table according to real-time working parameters of the device to be processed and The output is DPD.
  • the evaluation module 11 in the above embodiment is configured to analyze the evaluation data using a clustering and fitting method, and obtain a valid data collection use case and a full range by combining the finite values of all the working parameters.
  • the data generation method forms a data file by the data generation method and the full range data generation method.
  • the clustering and fitting involved in the present application are commonly used algorithms, and are generally described below.
  • Cluster analysis also known as group analysis, is a statistical analysis method for the classification of research (samples or indicators) and an important algorithm for data mining.
  • Cluster analysis consists of several patterns. Usually, a pattern is a vector of measures or a point in a multidimensional space. Cluster analysis is based on similarity, with more similarities between patterns in one cluster than patterns that are not in the same cluster.
  • Cluster analysis includes the following four main methods:
  • the k-means algorithm accepts the input quantity k; then divides the n data objects into k clusters so that the obtained clusters are satisfied: the object similarity in the same cluster is higher; and the object similarity in different clusters Smaller.
  • Cluster similarity is calculated by using a "central object” (gravitational center) obtained by the mean of the objects in each cluster.
  • k objects are arbitrarily selected from n data objects as the initial cluster center; and for other objects remaining, they are assigned to the most similar ones according to their similarity (distance) with these cluster centers ( Clustering represented by the clustering center;
  • the mean square error is generally used as the standard measure function.
  • the k clusters have the following characteristics: each cluster itself is as compact as possible, and the clusters are separated as much as possible.
  • K-MEANS has its drawbacks: the size of the generated classes is not very different and is sensitive to dirty data.
  • K-medoids method select an object called mediod instead of the center of the above, such a medoid identifies the class.
  • the difference between K-medoids and K-means lies in the selection of the center point.
  • K-means we take the center point as the average of all the data points in the current cluster.
  • K-medoids algorithm we will The cluster selects such a point - it has the smallest sum of distances to all other points (in the current cluster) - as the center point.
  • Step 1 Arbitrarily select K objects as medoids (O1, O2, ... Oi...Ok).
  • Step 2 divide the remaining objects into classes (based on the principle closest to medoid);
  • Step 3 For each class (Oi), select an Or in order, and calculate the consumption -E(Or) after replacing Oi with Or. Select the E with the smallest E to replace Oi. Thus K medoids change, and then go to step 2.
  • Step 4 this cycle until K medoids are fixed.
  • This algorithm is not sensitive to dirty data and abnormal data, but the amount of calculation is obviously larger than the K-means, and generally only suitable for small data.
  • the K-medoids algorithm mentioned above is not suitable for large data volume calculations. Clara algorithm, a sampling-based method that can process large amounts of data.
  • Clara algorithm The idea of the Clara algorithm is to replace the entire data with the actual data, and then use the K-medoids algorithm to get the best medoids on the sampled data. Clara algorithm from actual number According to the sampling of multiple samples, the corresponding (O1, O2...Oi...Ok) is obtained by K-medoids algorithm on each sample, and then the smallest one of E is selected as the final result.
  • the Clarans algorithm is proposed.
  • the difference from the Clara algorithm is that the sampling is constant in the process of finding the best medoids by Clara algorithm.
  • the Clarans algorithm uses different samples for each cycle. Unlike the process of finding the best medoids described above, the number of cycles must be artificially defined.
  • the fitting refers to a number of discrete function values ⁇ f1, f2, ..., fn ⁇ of a certain function.
  • a number of undetermined coefficients f( ⁇ 1, ⁇ 2, ..., ⁇ n) in the function By adjusting a number of undetermined coefficients f( ⁇ 1, ⁇ 2, ..., ⁇ n) in the function, the function and the known point are made. The difference in set (least squares meaning) is minimal.
  • the pending function is linear, it is called linear fitting or linear regression (mainly in statistics). If the pending function is not linear, it is called nonlinear fitting or nonlinear regression.
  • the expression can also be a piecewise function, in which case it is called a spline fit.
  • the numerical statistics of a set of observations are in agreement with the numerical statistics of the corresponding set of values.
  • the image says that fitting is to connect a series of points on a plane with a smooth curve. Because there are countless possibilities for this curve, there are various fitting methods.
  • Polyfit can be used to fit polynomials in MATLAB. Fitting and interpolation and approximation are the three basic tools for numerical analysis. In the common sense, the difference is that the fitting is a known point sequence, which is close to them as a whole; the interpolation is a known point column and passes through the point column completely; Approximation is a known curve, or a list of points, by which the constructed functions are brought infinitely close to them.
  • the acquisition module 12 in the above embodiment is configured to perform digital predistortion testing on the radio frequency signal of the device to be processed according to the data collection use case; display the digital predistorted radio frequency signal in a spectrum form, and scan the frequency step by step. A set of power values is obtained to generate optimization data.
  • the optimization module 13 in the above embodiment is configured to compare the pre-distorted signal with the standard signal, and detect the DPD effect until the power difference optimal value is found, within the range of the DPD table. First fix the imaginary part of the DPD, perform the real part traversal, determine the optimal real part data, fix the real part of the DPD to the optimal real part data, perform the imaginary part traversal, determine the optimal imaginary part data, and cycle to the optimal value. .
  • the optimization module 13 in the above embodiment is further configured to alternately perform large step optimization and small step optimization, and cycle to an optimal value.
  • the present application also provides a DPD system including the DPD table generated by the DPD table generating apparatus 1 provided by the present application.
  • the method for generating a DPD table includes the following steps:
  • S201 According to the nonlinear mathematical model of the device to be processed, perform full traversal data collection, obtain evaluation data, and generate data collection according to the evaluation data for all working parameters affecting the DPD table of the device to be processed, all working parameters of the product samples of the device to be processed. Use cases and full range of data generation methods;
  • S204 Generate a DPD table according to the model coefficients by using a full range data generation method.
  • the step of generating a data collection use case and a full range of data generation method according to the evaluation data in the foregoing embodiment includes: analyzing the evaluation data by using a clustering and fitting method to obtain a limited extraction of all working parameters.
  • An effective data collection use case and a full range of data generation methods are combined to form a data file for data collection use cases and a full range of data generation methods.
  • the step of obtaining the optimization data in the foregoing embodiment includes: performing a digital predistortion test on the radio frequency signal of the device to be processed according to the data collection use case; displaying the digital predistorted radio frequency signal in a spectrum form, according to the frequency
  • the step scan obtains a set of power values to generate optimized data.
  • the optimization data in the above embodiment performs the detection and optimization of the digital predistortion effect, and the optimization result is used as the model coefficient of the nonlinear mathematical model.
  • the step of generating the DPD table includes: predistorting The signal is compared with the standard signal, and the DPD effect is detected until the optimal value of the power difference is found.
  • the imaginary part of the DPD is fixed first, and the real part is traversed to determine the optimal real data, and the DPD is The part is fixed to the optimal real part data, and the imaginary part is traversed. Determine the optimal imaginary data and cycle to the optimal value.
  • the step of detecting and optimizing the DPD effect on the data in the foregoing embodiment includes: performing large step optimization and small step optimization alternately, and looping to an optimal value.
  • the present application provides a system and method for automatically generating a DPD table, which is directed to a mathematical model of the nonlinear characteristics of the device, using an automated acquisition system, and acquiring the DPD model coefficients required by the mathematical model based on the principle of probability statistics, in order to achieve The feasibility, feasibility and equipment production efficiency of data acquisition, these DPD model coefficients are collected under limited samples and screening factors. On this basis, in order to ensure the accuracy and predistortion effect of the data, the acquired DPD model The coefficients are traversed and optimized for the predistortion effect. After the optimized model coefficients are written into the DPD table, the basic data obtained by these optimizations are finally used, and the complete data in the selected factors and condition ranges are calculated to generate a complete data.
  • the DPD table is written into the device's EEPROM (Electrically Erasable Programmable Read-Only Memory).
  • the device is running with real-time conditions (such as specific operating parameters) as an index to query the EEPROM DPD table. Calling the corresponding DPD model coefficients and substituting them into the mathematical model for signal predistortion calculation. No pre-distortion.
  • the present application applies a method for generating a DPD table on the premise of a known nonlinear mathematical model of a device.
  • the accuracy, effectiveness, and predistortion effect of the data acquired by using this method are improved compared with the related techniques; Includes the following steps:
  • the first step sample evaluation. Select a limited product sample, and perform full traversal data acquisition in the data acquisition module under the five factors of bandwidth, power, frequency, modulation mode and temperature.
  • the collected data is fed back to the sample evaluation module, and the sample evaluation module uses clustering and quasi-study. Combining methods for statistical analysis, the finite combination of five factors is obtained, which is called data collection use case.
  • the method of generating the full range of data of five factors is analyzed, and the data collection use case and the method of generating the full range data are formed into data.
  • the files are respectively input as a data acquisition module and a DPD table generation module, specifically, the data collection use case is used as an input of the data acquisition module, and the full range data generation method is used as an input of the DPD table generation module;
  • the second step data collection.
  • Data acquisition use case obtained by sample evaluation, through control module The instrument control module in the middle sets the spectrum analyzer, the device control module sets the transmitting device, the DPD data goes to the default value, enables the DPD function, and performs DPD on the emitted RF signal.
  • the interface board control module in the control module controls the switch on the interface board to connect the transmitting device and the spectrum analyzer, so that the radio frequency signal passing through the DPD is displayed in the spectrum form in the spectrum.
  • the control spectrum analyzer is stepped by frequency to obtain a set of power values, which are fed back to the data acquisition module, and the data acquisition module inputs the data into the data processing module for analysis and data extraction;
  • the third step data processing.
  • the data input to the data acquisition module is traversed in the DPD data range given by the device, and the DPD data value traversed each time is fed back to the data acquisition module to control the DPD of the transmitted signal, and the pre-distorted signal and the standard signal are used for power. Contrast, detect the DPD effect until the optimal value of the power difference is found, the effect of the transmitted signal is optimized, and the optimal value and range of the DPD data at this time are recorded. Based on this, the signal source is set to receive the signal through the instrument control module.
  • the interface board control module controls the switch to connect the receiving device and the signal source, detects the receiving signal of the receiving device, performs the DPD effect verification and optimization of the received signal, and the receiving signal optimization process is also the same as the data collecting module. Perform a feedback loop until the optimal value is found. Repeat steps 2 and 3 until you have traversed all the use cases given by the evaluation.
  • the fourth step DPD table generation.
  • the optimal value obtained by data processing is input into the DPD table generation module, which uses the full range data generation method obtained by the sample evaluation to generate complete DPD data related to the five influencing factors, forms a data file, and data is obtained through the device control module in the control module.
  • the file is written to the transmitting device, and after writing, the data written to the transmitting device is read back through the device control module for verification.
  • the system for offline generation of a DPD table provided by this embodiment includes the following modules:
  • the measurement device 35 is divided into an instrument control module 351, an interface board control module 352, and a device control module 353 according to different control objects; the system for offline generation of the DPD table further includes: a radio frequency connection line, a network port connection line, and a serial port connection. line.
  • the instrument control module 351 and the device control module 353 respectively connect the radio frequency meter and the device under test through the network port connection line, and complete the transmission and interaction of the control signal and the digital signal.
  • the interface board control module 352 connects the interface board through the serial port connection line, sends a control signal to the interface board, and reads back the control state.
  • Device under test The radio frequency connection line is connected with the radio frequency meter through the interface board to perform RF signal exchange; the interface board is designed with a radio frequency switch, and the RF signal switching between each device and the instrument is completed under the control of the interface board control module; It is mainly responsible for receiving and displaying RF signals, and receiving signals from receiving devices.
  • the functional block diagram of the above system is shown in Figure 3.
  • the connecting arrows in the figure indicate the signal direction.
  • the connecting lines from coarse to fine indicate different signal types. From coarse to fine, the data stream, RF transmitting signal, RF receiving signal and control signal are in turn. , wherein the control signals are both bidirectional, indicating that after the signal is sent to the controlled device, the controlled device needs to feed back and be read and controlled.
  • the method for generating a DPD table includes:
  • the first step transmitting a signal.
  • the transmitting and receiving devices are set by the device control module 353 in the control module 35, and then the switching device is connected to the transmitting device through the interface board control module 352 in the control module 35.
  • a spectrum analyzer that causes the transmitting device to emit the required signal
  • the second step real traversal optimization.
  • the imaginary part of the DPD data is fixed, traversing the real part, enabling the DPD function through the interface board control module 352 of the control module 35, and then reading and analyzing the spectrum data from the spectrum analyzer through the meter control module 351. Find the DPD data with the best DPD effect;
  • the third step imaginary traversal optimization. In the range of DPD data given by the device, use the real data obtained in the second step, traverse the imaginary part, and analyze the spectrum optimization in the same way as the second step;
  • the fourth step the second part of the real optimization.
  • use the imaginary part data obtained in the third step traverse the real part, and analyze the spectrum optimization in the same way as the second step;
  • Step 5 Repeat the second to fourth steps, and compare the transmitted signal with the standard signal until the optimal value is found and recorded, and the search range of the DPD data is also recorded;
  • Step 6 Configure the received signal.
  • the signal source is configured by the meter control module 351 in the control module 35.
  • the device control module 353 controls the transmitting device to enable the DPD function, and then switches the interface board switch through the interface board control module 352 of the control module 35 to connect the transmitting device and the signal source, and the signal. a source and a receiving device, such that the receiving device receives the required signal;
  • the seventh step large stepping to find the best.
  • the DPD data is traversed by a large step, and the received signal on the receiving device is read by the device control module 353 to optimize the DPD data;
  • the eighth step small stepping optimization. Large stepping optimization results output a data range, input small step optimization, use the same method in the seventh step to further optimize in a small range;
  • the ninth step repeating the seventh to eighth steps until the DPD model coefficients corresponding to the two consecutive optimal received signals are the same, ending the reception signal optimization, and recording the optimization result;
  • Step 10 Repeat steps 1 through 9 until the data collection use case is all optimized, record the final result of the optimization, and output the final result of the optimization as a data file.
  • This application mainly implements the system and method for automatically generating DPD tables on the PC.
  • This method considers various conditions that affect the performance of DPD during device operation. And factors, based on the nonlinear characteristic model of the power amplifier, the DPD data is collected in a targeted manner, and the correctness and validity of the collected data are immediately tested and verified in the equipment operating environment, and further optimized according to the DPD effect, optimization The latter data is written and generated into the DPD table, ensuring the correctness and effectiveness of the entire method and system.
  • the method is versatile. For different devices and devices, only different nonlinear mathematical models need to be given, as well as the conditions and factors affecting the DPD.
  • the method and system described in this application can be used to collect DPD data and generate DPD table.
  • DPD chips are divided into manual generation data and automatic implementation of DPD chips.
  • the cost of DPD chips is high, and the DPD acquisition number, device nonlinear model establishment, DPD training, and DPD table generation all require system overhead, and the signal requirements are not Very high, or the nonlinear effects of amplifiers are not very demanding, DPD chip technology is too costly and lacks flexibility.
  • the DPD table generation method given in this application has a predistortion effect close to or even better than that of the DPD chip, which has been practiced and verified in the application of microwave products. Therefore, the present application takes into consideration the flexibility and low cost of the manual method, and can achieve the predistortion effect of the DPD chip under the same conditions, and is a simple, efficient and low-cost integrated method.
  • x is the input signal
  • y is the output signal
  • g1, g3, and g5 are polynomial model coefficients.
  • the values of the three model coefficients determine the DPD effect under different conditions and factors. It has been proved by experiments that only g3 of the three model coefficients play a decisive role. Therefore, only the g3 value is found when counting. Digital chip manual The given g3 range is -2 to +2.
  • the first step sample evaluation. Selecting 4 product samples, due to the limited value of the modulation method, is determined by the evaluation results of other factors. Bandwidth evaluation, power estimation and frequency estimation are performed first. In the data acquisition module, the full traversal data acquisition is performed, and the sample evaluation module uses the methods of clustering and fitting to perform statistical analysis on the traversed data, and obtains the values of bandwidth, power and frequency, and uses these values respectively under high and low temperature. Perform a temperature assessment.
  • the bandwidth is 28M
  • the power is 16dBm, 21dBm and 23dBm
  • the modulation modes are 128QAM, 512QAM and 1024QAM
  • the frequency points are 21200MHz, 21800MHz, 22400MHz, 23000MHz and 23600MHz
  • the normal temperature is 33°
  • the output data file is InterpolationData_23G.ini. High temperature and low temperature 55°, 5°, -15°, -33° and -40° separately output a file as access database table usr_Coefficients_23G.
  • the second step data collection.
  • the interface control module of the RDC acquisition tool is also controlled by the interface control module of the automatic control and acquisition software to display the transmitted RF signal in the form of spectrum on the spectrum analyzer.
  • the control spectrum analyzer is stepped by frequency to obtain a set of power values, which are fed back to the data acquisition module, and the data acquisition module inputs the data into the data processing module for analysis and data extraction;
  • the third step data processing.
  • the data input by the data acquisition module traverses the g3 value within the range of the DPD model coefficient g3, and the g3 value of each traversal is fed back to the data acquisition module to control the transmission signal for DPD, and the signal after the DPD is compared with the standard signal defined by the European standard. Detecting the DPD effect until the optimal value of the power difference is found, ending the effect of the transmitted signal, and recording the optimal value and range of the DPD model coefficient g3 at this time, based on which the signal source SMF100A is set by the instrument control module.
  • Receiving a signal, and controlling the switch to connect the receiving device and the signal source through the interface board control module, detecting the receiving signal of the receiving device, performing DPD effect verification and optimization of the received signal, and the same as the transmitting signal optimization, the receiving signal optimization process is also The data acquisition module performs a feedback loop until an optimal value is found. Repeat steps 2 and 3 until you have traversed all the use cases given by the evaluation.
  • the fourth step DPD table generation.
  • the optimal value obtained by data processing is input into the DPD table generation module, which uses the full range data generation method obtained by the sample evaluation to generate five influencing factors related to the completion.
  • the whole DPD data is formed into a data file, and the data file is written into the transmitting device through the device control module of the automatic control and the counting software, and after the writing is completed, the data written to the transmitting device is read back through the device control module for verification.
  • the DPD data optimization algorithm used in the specific application is as follows:
  • the first step is to transmit a signal.
  • Read the use case given in the InterpolationData_23G.ini file send the control commands through the automation control and the acquisition software to set the transmitting and receiving devices, and then send the control commands through the automatic control and the counting software to switch the RF switch connected transmitting device of the RDC counting tooling. And the spectrum analyzer FSV, so that the transmitting device sends out the required signal;
  • the second step the real part of the traversal optimization.
  • the imaginary part of the DPD model coefficient g3 is fixed, the real part is traversed, the DPD function is enabled by the automation control and the software, and then the software is controlled by the automation and the software.
  • the spectrum analyzer reads and analyzes the spectrum data to find the DPD data with the best DPD effect;
  • the third step the imaginary part traverses to find the best.
  • the real data obtained in the second step traverse the imaginary part, and analyze the spectrum optimization in the same way as the second step;
  • the fourth step the real part of the second optimization.
  • use the imaginary part data obtained in the third step traverse the real part, and analyze the spectrum optimization in the same way as the second step;
  • the second to fourth steps are repeated, and the transmitted signal is compared with the standard signal until the optimal value is found and recorded, and the search range of the DPD data is also recorded;
  • the sixth step is to configure the receiving signal.
  • the automatic control and acquisition software configures the signal source, and the automation control and acquisition software control the transmitting device to enable the DPD function.
  • the interface board switch is connected to the transmitting device and the signal source, as well as the signal source and the receiving device. So that the receiving device receives the required signal;
  • the seventh step large stepping to find the best.
  • the DPD model coefficient g3 is traversed by 0.1 as a large step, the MSE value of the received signal on the receiving device is read by the 53 module, and the DPD data is optimized by the MSE value;
  • the eighth step small stepping optimization.
  • the large stepping optimization result outputs a data range, and in the range, the same method in the seventh step is used to further optimize the small range in a small step of 0.02;
  • Step 10 Repeat steps 1 through 9 until the data collection use case is all optimized. Record the final result of the optimization and output the final result as an excel file.
  • the file is mostly a product barcode command, for example, 219053719224.xlsm.
  • the present application provides a method for generating a DPD table, considering all the influencing factors, using statistical methods to evaluate and analyze all factors, obtaining a combination of influencing factors of the number of uses, and using these use cases for the number of acquisitions, which solves the discrete conditions. Selecting the problem and using automation also ensure the efficiency of data collection and generation. At the same time, the acquisition algorithm and the complete DPD data generation algorithm are also improved compared with the existing algorithms.
  • the related technology DPD table generation method whether it is offline writing The incoming DPD table or the DPD table obtained in real-time training will not perform pre-distortion effect detection and optimization on the data in the table.
  • the DPD table generation method given in this application has data in detecting and optimizing pre-distortion.
  • the effect is obtained under the premise of obtaining the DPD, and the data is used to enable the DPD, and a receiving and transmitting link is configured and connected, the signal of the receiving end device is detected, and the DPD data obtained by the preceding signal is further traversed, verified and optimized by the receiving signal, thereby ensuring
  • the data in the DPD table is the data after the pre-distortion effect is optimized, and thus the data in the DPD table in this application is achieved.
  • DPD list generation method is better than the real acquired data related art pre-distortion effect achieved.
  • Embodiments of the present invention further provide a computer readable storage medium storing computer executable instructions that are implemented when the computer executable instructions are executed.
  • a program to instruct related hardware eg, a processor
  • a computer readable storage medium such as a read only memory, a disk, or CD, etc.
  • all or part of the steps of the described embodiments may also be implemented using one or more integrated circuits.
  • each module/unit in the embodiment may be implemented in the form of hardware, such as an integrated circuit to implement its corresponding function, or may be implemented in the form of a software function module, for example, executed by a processor and stored in a memory.
  • Embodiments of the invention are not limited to any specific form of combination of hardware and software.
  • the method for generating a DPD table the data is collected under the premise of detecting and optimizing the pre-distortion effect, and the DPD is enabled after the data is obtained, and a receiving and transmitting link is configured and connected, and the receiving end device is detected.
  • the signal further traverses, verifies, and optimizes the previously obtained DPD data by using the received signal, thereby ensuring that the data in the DPD table is the data after the predistortion effect is optimized, thereby making the data in the DPD table in the present application reach the pre
  • the distortion effect is better than the predistortion effect achieved by the data acquired by the related art DPD table generation method.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Amplifiers (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

公开了一种DPD表生成方法、装置及DPD系统，该方法包括：根据待处理装置的非线性数学模型，对待处理装置的产品样本的所有工作参数进行全遍历数据采集，获取评估数据，根据评估数据生成数据采集用例及全范围的数据生成方法；使用数据采集用例对待处理装置采集数据，获取寻优数据；对寻优数据进行数字预失真效果的检测和寻优，寻优结果作为非线性数学模型的模型系数；使用全范围的数据生成方法,根据模型系数生成DPD表。通过上述方案，使用统计学方法对影响待处理装置的DPD表的所有工作参数进行评估分析，同时，对表中的数据进行预失真效果检测和寻优，使得DPD表具备较佳的数字预失真效果。

Description

数字预失真表生成方法、装置及数字预失真系统 技术领域
本申请涉及但不限于移动通信领域，特别是一种数字预失真表生成方法、装置及数字预失真系统。
背景技术
由于影响器件非线性特性的因素较多，原因复杂，而且很多影响因素的变化在设备实际运行时具有连续性，DPD(Digital Pre-Distortion，数字预失真)表生成时，表中数据的采集只能在这些连续变化的因素中选取几个离散条件，这种离散条件如何选取，设备运行时，每种运行条件下进行预失真时所使用的DPD数据不一定都能在DPD表中得到，这些DPD表中获取不到的数据如何通过表中的有限数据计算得到，如何保证采集及计算得到的DPD数据的准确性、有效性和预失真的最佳性能，是相关技术的各种DPD模块和算法的关注点。
但是，相关的DPD表生成方法只考虑了一种影响因素，如仅根据放大器特性或者反馈信号进行预失真训练以及表的生成，并且，没有检测和评估DPD表中数据的预失真效果；即，相关的DPD表生成方法考虑的影响因素单一、且没有对DPD表的数据进行检测及评估，效果差。
因此，如何提供一种可以解决相关技术的DPD表影响因素单一、且没有对DPD表的数据进行检测及评估导致的数字预失真效果差的DPD表生成方法，是本领域技术人员亟待解决的技术问题。
发明内容
以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。
本申请提供了一种数字预失真表生成方法、装置及数字预失真系统，以解决相关技术的DPD表影响因素单一、且没有对DPD表的数据进行检测及 评估导致的数字预失真效果差的问题。
本申请提供了一种DPD表生成方法，其包括：
根据待处理装置的非线性数学模型，针对影响待处理装置DPD表的所有工作参数，对待处理装置的产品样本的所有工作参数进行全遍历数据采集，获取评估数据，根据评估数据生成数据采集用例及全范围的数据生成方法；
使用数据采集用例对待处理装置采集数据，获取寻优数据；
对寻优数据进行数字预失真效果的检测和寻优，寻优结果作为非线性数学模型的模型系数；
使用全范围的数据生成方法,根据模型系数生成DPD表。
可选地，根据评估数据生成数据采集用例及全范围的数据生成方法包括：使用聚类、拟合的方法对评估数据进行分析，得到所有工作参数的有限取值组合而成的有效的数据采集用例以及全范围的数据生成方法。
可选地，获取寻优数据包括：根据数据采集用例，对待处理装置的射频信号进行数字预失真测试；将数字预失真的射频信号以频谱形式显示，按频率步进扫描得到一组功率值，生成寻优数据。
可选地，对寻优数据进行数字预失真效果的检测和寻优，将寻优结果作为非线性数学模型的模型系数，生成DPD表包括：将预失真后的信号与标准信号进行功率对比，检测DPD效果，直到找到功率差值最优值，在DPD表的范围内，先固定DPD虚部，进行实部遍历，确定最优的实部数据，将DPD实部固定为最优的实部数据，进行虚部遍历，确定最优的虚部数据，循环至最优值。
可选地，对寻优数据进行数字预失真效果的检测和寻优包括：大步进寻优(large-step optimization)与小步进寻优(small-step optimization)交替执行，循环至最优值。
本申请另外提供一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令被执行时实现所述方法。
本申请提供了一种DPD表生成装置，其包括评估模块、采集模块、寻优模块及生成模块，其中，
评估模块设置为根据待处理装置的非线性数学模型，针对影响待处理装置DPD表的所有工作参数，使用采集模块对待处理装置的产品样本的所有工作参数进行全遍历数据采集，获取评估数据，根据评估数据生成数据采集用例及全范围的数据生成方法；
采集模块设置为使用数据采集用例对待处理装置采集数据，获取寻优数据；
寻优模块设置为对寻优数据进行数字预失真效果的检测和寻优，寻优结果作为非线性数学模型的模型系数；
生成模块设置为使用所述全范围的数据生成方法,根据所述模型系数生成DPD表。
可选地，评估模块设置为使用聚类、拟合的方法对评估数据进行分析，得到所有工作参数的有限取值组合而成的有效的数据采集用例及全范围的数据生成方法。
可选地，采集模块设置为根据数据采集用例，对待处理装置的射频信号进行数字预失真测试；将数字预失真的射频信号以频谱形式显示，按频率步进扫描得到一组功率值，生成寻优数据。
可选地，寻优模块设置为将预失真后的信号与标准信号进行功率对比，检测DPD效果，直到找到功率差值最优值，在DPD表的范围内，先固定DPD虚部，进行实部遍历，确定最优的实部数据，将DPD实部固定为最优的实部数据，进行虚部遍历，确定最优的虚部数据，循环至最优值。
可选地，寻优模块还设置为大步进寻优与小步进寻优交替执行，循环至最优值。
同时，本申请也提供了一种DPD系统，其包括本申请提供的DPD表生成装置生成的DPD表。
本申请的有益效果：
本申请提供了一种DPD表生成方法，考虑了影响待处理装置的DPD表的所有工作参数，使用统计学方法对所有工作参数进行评估分析，得到采数的影响因素组合用例，使用这些用例进行采数，既解决了离散条件选取问题， 利用自动化也保证了数据的采集和生成效率；同时，在采数算法和完整的DPD数据生成算法上较相关技术的算法也有所改进，相关技术的DPD表生成方法，无论是离线写入的DPD表，还是实时训练得到的DPD表，均不会对表中的数据进行预失真效果检测和寻优，本申请给出的DPD表生成方法，其数据是在检测和寻优预失真效果前提下采集得到，得到后使用该数据使能DPD，并且配置和连通一条收、发链路，检测接收端设备信号，用接收信号进一步遍历、验证和寻优前面得到的DPD数据，从而保证DPD表中的数据都是预失真效果寻优后的数据，进而使得本申请中DPD表中的数据所达到的预失真效果要好于相关技术的DPD表生成方法获取的数据所达到的预失真效果。
在阅读并理解了附图和详细描述后，可以明白其他方面。
附图概述
图1为本申请第一实施例提供的DPD表生成装置的结构示意图；
图2为本申请第二实施例提供的DPD表生成方法的流程图；
图3为本申请第三实施例提供的DPD系统的结构示意图。
本发明的较佳实施方式
现通过具体实施方式结合附图的方式对本申请做出进一步的诠释说明。
第一实施例：
图1为本申请第一实施例提供的DPD表生成装置的结构示意图，由图1可知，在本实施例中，本申请提供的DPD表生成装置1包括评估模块11、采集模块12、寻优模块13及生成模块14，其中，
评估模块11设置为根据待处理装置的非线性数学模型，针对影响待处理装置DPD表的所有工作参数，使用采集模块12对待处理装置的产品样本的所有工作参数进行全遍历数据采集，获取评估数据，根据评估数据生成数据采集用例及全范围的数据生成方法，供采集模块12及生成模块14使用；
采集模块12设置为使用评估模块11得到的数据采集用例对待处理装置 采集数据，获取寻优数据，并反馈给寻优模块13；
寻优模块13设置为对采集模块12采集得到的寻优数据进行数字预失真效果的检测和寻优，寻优结果作为非线性数学模型的模型系数；
生成模块14设置为使用评估模块11得到的全范围的数据生成方法,根据所述模型系数生成DPD表。
在一些实施例中，上述实施例中的DPD表生成装置1还包括：调用模块，设置为在生成DPD表之后，存储DPD表，并根据待处理装置的实时工作参数，查找对应的DPD表并输出进行DPD。
在一些实施例中，上述实施例中的评估模块11设置为使用聚类、拟合的方法对评估数据进行分析，得到所有工作参数的有限取值组合而成的有效的数据采集用例以及全范围的数据生成方法，将数据生成方法及全范围的数据生成方法形成数据文件，本申请所涉及的聚类、拟合是常用的算法，下文进行大致说明。
聚类分析又称群分析，它是研究(样品或指标)分类问题的一种统计分析方法，同时也是数据挖掘的一个重要算法。聚类(Cluster)分析是由若干模式(Pattern)组成的，通常，模式是一个度量(Measurement)的向量，或者是多维空间中的一个点。聚类分析以相似性为基础，在一个聚类中的模式之间比不在同一聚类中的模式之间具有更多的相似性。
聚类分析包括以下4种主要方法：
1、K-MEANS；
k-means算法接受输入量k；然后将n个数据对象划分为k个聚类以便使得所获得的聚类满足：同一聚类中的对象相似度较高；而不同聚类中的对象相似度较小。聚类相似度是利用各聚类中对象的均值所获得一个“中心对象”(引力中心)来进行计算的。
k-means算法的工作过程说明如下：
首先从n个数据对象任意选择k个对象作为初始聚类中心；而对于所剩下其它对象，则根据它们与这些聚类中心的相似度(距离)，分别将它们分配给与其最相似的(聚类中心所代表的)聚类；
然后再计算每个所获新聚类的聚类中心(该聚类中所有对象的均值)；不断重复这一过程直到标准测度函数开始收敛为止。
一般都采用均方差作为标准测度函数.k个聚类具有以下特点：各聚类本身尽可能的紧凑，而各聚类之间尽可能的分开。
2、K-MEDOIDS；
K-MEANS有其缺点：产生类的大小相差不会很大，对于脏数据很敏感。
改进的算法：k—medoids方法。这儿选取一个对象叫做mediod来代替上面的中心的作用，这样的一个medoid就标识了这个类。K-medoids和K-means不一样的地方在于中心点的选取，在K-means中，我们将中心点取为当前cluster中所有数据点的平均值，在K-medoids算法中，我们将从当前cluster中选取这样一个点——它到其他所有(当前cluster中的)点的距离之和最小——作为中心点。
k-medoids方法的步骤：
步骤1，任意选取K个对象作为medoids(O1,O2,…Oi…Ok)。
以下是循环的：
步骤2，将余下的对象分到各个类中去(根据与medoid最相近的原则)；
步骤3，对于每个类(Oi)中，顺序选取一个Or，计算用Or代替Oi后的消耗—E(Or)。选择E最小的那个Or来代替Oi。这样K个medoids就改变了，下面就再转到步骤2。
步骤4，这样循环直到K个medoids固定下来。
这种算法对于脏数据和异常数据不敏感，但计算量显然要比K均值要大，一般只适合小数据量。
3、Clara；
上面提到K-medoids算法不适合于大数据量的计算。Clara算法，这是一种基于采样的方法，它能够处理大量的数据。
Clara算法的思想就是用实际数据的抽样来代替整个数据，然后再在这些抽样的数据上利用K-medoids算法得到最佳的medoids。Clara算法从实际数 据中抽取多个采样，在每个采样上都用K-medoids算法得到相应的(O1,O2…Oi…Ok)，然后在这当中选取E最小的一个作为最终的结果。
4、Clarans；
Clara算法的效率取决于采样的数量，一般不太可能得到最佳的结果。
在Clara算法的基础上，又提出了Clarans的算法，与Clara算法不同的是：在Clara算法寻找最佳的medoids的过程中，采样都是不变的。而Clarans算法在每一次循环的过程中所采用的采样都是不一样的。与上面所讲的寻找最佳medoids的过程不同的是，必须人为地来限定循环的次数。
而拟合是指已知某函数的若干离散函数值{f1,f2,…,fn}，通过调整该函数中若干待定系数f(λ1,λ2,…,λn)，使得该函数与已知点集的差别(最小二乘意义)最小。如果待定函数是线性，就叫线性拟合或者线性回归(主要在统计中)，如果待定函数不是线性的，则叫作非线性拟合或者非线性回归。表达式也可以是分段函数，这种情况下叫作样条拟合。一组观测结果的数字统计与相应数值组的数字统计吻合。形象的说,拟合就是把平面上一系列的点，用一条光滑的曲线连接起来。因为这条曲线有无数种可能，从而有各种拟合方法。拟合的曲线一般可以用函数表示，根据这个函数的不同有不同的拟合名字。
在MATLAB中可以用polyfit来拟合多项式。拟合以及插值还有逼近是数值分析的三大基础工具，通俗意义上它们的区别在于：拟合是已知点列，从整体上靠近它们；插值是已知点列并且完全经过点列；逼近是已知曲线，或者点列，通过逼近使得构造的函数无限靠近它们。
在一些实施例中，上述实施例中的采集模块12设置为根据数据采集用例，对待处理装置的射频信号进行数字预失真测试；将数字预失真的射频信号以频谱形式显示，按频率步进扫描得到一组功率值，生成寻优数据。
在一些实施例中，上述实施例中的寻优模块13设置为将预失真后的信号与标准信号进行功率对比，检测DPD效果，直到找到功率差值最优值，在DPD表的范围内，先固定DPD虚部，进行实部遍历，确定最优的实部数据，将DPD实部固定为最优的实部数据，进行虚部遍历，确定最优的虚部数据，循环至最优值。
在一些实施例中，上述实施例中的寻优模块13还设置为大步进寻优与小步进寻优交替执行，循环至最优值。
同时，本申请也提供了一种DPD系统，其包括本申请提供的DPD表生成装置1生成的DPD表。
第二实施例：
图2为本申请第二实施例提供的DPD表生成方法的流程图，由图2可知，在本实施例中，本申请提供的DPD表生成方法包括以下步骤：
S201：根据待处理装置的非线性数学模型，针对影响待处理装置DPD表的所有工作参数，对待处理装置的产品样本的所有工作参数进行全遍历数据采集，获取评估数据，根据评估数据生成数据采集用例及全范围的数据生成方法；
S202：使用数据采集用例对待处理装置采集数据，获取寻优数据；
S203：对寻优数据进行数字预失真效果的检测和寻优，寻优结果作为非线性数学模型的模型系数；
S204：使用全范围的数据生成方法,根据模型系数生成DPD表。
在一些实施例中，上述实施例中的根据评估数据生成数据采集用例及全范围的数据生成方法的步骤包括：使用聚类、拟合的方法对评估数据进行分析，得到所有工作参数的有限取值组合而成的有效的数据采集用例以及全范围的数据生成方法，将数据采集用例以及全范围的数据生成方法形成数据文件。
在一些实施例中，上述实施例中的获取寻优数据的步骤包括：根据数据采集用例，对待处理装置的射频信号进行数字预失真测试；将数字预失真的射频信号以频谱形式显示，按频率步进扫描得到一组功率值，生成寻优数据。
在一些实施例中，上述实施例中的寻优数据进行数字预失真效果的检测和寻优，将寻优结果作为非线性数学模型的模型系数，生成DPD表的步骤包括：将预失真后的信号与标准信号进行功率对比，检测DPD效果，直到找到功率差值最优值，在DPD表的范围内，先固定DPD虚部，进行实部遍历，确定最优的实部数据，将DPD实部固定为最优的实部数据，进行虚部遍历， 确定最优的虚部数据，循环至最优值。
在一些实施例中，上述实施例中的对数据进行DPD效果的检测和寻优的步骤包括：大步进寻优与小步进寻优交替执行，循环至最优值。
现结合具体应用场景对本申请做进一步的诠释说明。
第三实施例：
本申请给出了一种自动生成DPD表的系统和方法，该方法针对器件的非线性特性数学模型，使用自动化采数系统，基于概率统计学的原理采集数学模型需要的DPD模型系数，为了实现数据采集的可行性、可实施性以及设备生产效率，这些DPD模型系数在有限样本和筛选因素下采集得到，在此基础上，为了保证数据的准确性和预失真效果，对采集得到的DPD模型系数进行预失真效果的遍历和寻优，将这些优化后的模型系数写入DPD表后，最终使用这些寻优得到的基础数据，计算生成所选因素和条件范围内的完整数据，生成完整的DPD表，并写入设备的EEPROM(Electrically Erasable Programmable Read-Only Memory，电可擦可编程只读存储器)中，设备运行时以实时条件(如具体的工作参数)为索引，查询EEPROM的DPD表，调用对应的DPD模型系数，代入数学模型中进行信号预失真计算，对信号进行预失真。
本申请在已知器件非线性数学模型的前提下申请了一种生成DPD表的方法，使用这种方法采集得到的数据准确性、有效性以及预失真效果较相关技术的方法都有所改进；包括以下步骤：
第一步、样本评估。选择有限的产品样本，在带宽、功率、频率、调制方式和温度五种因素下在数据采集模块中进行全遍历数据采集，采集到的数据反馈给样本评估模块，样本评估模块使用聚类、拟合等方法进行统计分析，得到五种因素的有限取值组合，称之为数据采集用例，同时分析出五种因素全范围数据的生成方法，将数据采集用例及全范围数据的生成方法形成数据文件，分别作为数据采集模块和DPD表生成模块的输入，具体的为，数据采集用例作为数据采集模块的输入，全范围数据的生成方法作为DPD表生成模块的输入；
第二步、数据采集。利用样本评估得到的数据采集用例，通过控制模块 中的仪表控制模块设置频谱仪，设备控制模块设置发射设备，DPD数据去默认值，使能DPD功能，对发出的射频信号进行DPD。同样通过控制模块中的接口板控制模块控制接口板上的开关连通发射设备和频谱仪，使得经过DPD的射频信号以频谱形式显示在频谱上。控制频谱仪按频率步进扫描得到一组功率值，反馈给数据采集模块，数据采集模块将数据输入数据处理模块进行分析和数据提取；
第三步、数据处理。对数据采集模块输入的数据在器件给定的DPD数据范围内进行DPD数据遍历，每次遍历的DPD数据值反馈给数据采集模块控制对发射信号进行DPD，预失真后的信号与标准信号进行功率对比，检测DPD效果，直到找到功率差值最优值，结束发射信号效果寻优，记录此时的DPD数据最优值及范围，以此为基础，通过仪表控制模块设置信号源发出接收信号，并通过接口板控制模块控制开关连通接收设备和信号源，检测接收设备的接收信号，进行接收信号的DPD效果验证和寻优，与发射信号寻优一样，接收信号寻优过程也与数据采集模块进行反馈循环，直到找到最优值。重复第二和第三步，直到遍历完成评估给出的所有用例。
第四步、DPD表生成。数据处理得到的最优值输入DPD表生成模块，该模块利用样本评估得到的全范围数据生成方法生成五种影响因素相关的完整DPD数据，形成数据文件，通过控制模块中的设备控制模块将数据文件写入发射设备，写完后在通过设备控制模块回读写入发射设备的数据进行校验。
现结合图3对寻优算法进行说明，如图3所示，本实施例提供的离线生成DPD表的系统包括以下模块：
样本评估模块31、数据采集模块32、数据处理模块33、DPD表生成模块34、控制模块35、接口板36、包含频谱仪和信号源的射频仪表37、包含发射设备38和接收设备39的被测设备，其中控制模块35根据控制对象的不同分为仪表控制模块351、接口板控制模块352和设备控制模块353；离线生成DPD表的系统还包括：射频连接线、网口连接线、串口连接线。其中仪表控制模块351和设备控制模块353均通过网口连接线分别连接射频仪表和被测设备，完成控制信号和数字信号的发送和交互。接口板控制模块352通过串口连接线连接接口板，向接口板发送控制信号并回读控制状态。被测设备 和射频仪表之间使用射频连接线通过接口板间接连接，进行射频信号交互；接口板上设计有射频切换开关，在接口板控制模块控制下完成各设备和仪表之间的射频信号切换；射频仪表则主要负责射频信号的接收和显示，以及接收设备的接收信号转入。以上系统的功能框图参见附图3，图中连线箭头表示信号方向，连线从粗到细表示不同的信号类型，由粗到细依次为数据流、射频发射信号、射频接收信号和控制信号，其中控制信号均为双向，表示向被控设备发送信号之后，被控设备要反馈和被读取被控状态。
基于上图，本实施例提供的DPD表生成方法包括：
第一步：发射信号。使用评估时选定的五种影响因素取值和信号参数，通过控制模块35中的设备控制模块353设置发射和接收设备，然后通过控制模块35中的接口板控制模块352切换开关连通发射设备和频谱仪，使得发射设备发出要求的信号；
第二步：实部遍历寻优。在器件给定的DPD数据范围内，固定DPD数据虚部，遍历实部，通过控制模块35的接口板控制模块352使能DPD功能，然后通过仪表控制模块351从频谱仪读取并分析频谱数据，寻找DPD效果最优的DPD数据；
第三步：虚部遍历寻优。在器件给定的DPD数据范围内，使用第二步得到的实部数据，遍历虚部，采用与第二步相同的方法分析频谱寻优；
第四步：实部二次寻优。在器件给定的DPD数据范围内，使用第三步得到的虚部数据，遍历实部，采用与第二步相同的方法分析频谱寻优；
第五步：重复第二到四步，发射信号与标准信号对比，直到找到最优值并记录，同时记录的还有DPD数据的寻优范围；
第六步：配置接收信号。通过控制模块35中的仪表控制模块351配置信号源，设备控制模块353控制发射设备使能DPD功能，然后通过控制模块35的接口板控制模块352切换接口板开关连通发射设备和信号源，以及信号源和接收设备，使得接收设备接收到要求的信号；
第七步：大步进寻优。在上面得到的DPD寻优范围内，大步进遍历DPD数据，通过设备控制模块353读取接收设备上的接收信号，寻优DPD数据；
第八步：小步进寻优。大步进寻优结果输出一个数据范围，输入小步进寻优，使用第七步同样的方法进一步小范围内寻优；
第九步：重复第七到第八步，直到连续两次最优接收信号对应的DPD模型系数相同，结束接收信号寻优，记录寻优结果；
第十步：重复第一到九步，直到数据采集用例全部寻优结束，记录寻优最终结果，并输出寻优最终结果为一个数据文件。
以上算法均在计算机上使用自动化控制和采数软件自动实现，有益效果：本申请主要在PC机上实现了DPD表自动生成的系统和方法，该方法考虑了设备运行时各种影响DPD性能的条件和因素，基于功率放大器的非线性特性模型有针对性的采集DPD数据，并且采集得到的数据正确性和有效性立刻在设备运行环境中进行测试和验证，且进一步根据DPD效果寻优，寻优后的数据才写入和生成DPD表，保证了整个方法和系统的正确性和有效性。同时该方法具有通用性，针对不同的器件和设备仅需要给出不同的非线性数学模型，以及影响DPD的条件和因素，就可以使用本申请所阐述的方法和系统进行DPD数据采集，进而生成DPD表。
另外，相关技术分为手动生成数据和DPD芯片自动化实现，DPD芯片所需成本高，而且DPD的采数、器件非线性模型建立、DPD训练以及DPD表生成都需要系统开销，对信号的要求不是很高，或者放大器非线性影响不是很大的需求，DPD芯片技术成本过高，也欠缺灵活性。最重要的是，本申请给出的DPD表生成方法，其预失真的效果接近甚至优于DPD芯片的效果，这一点已经在微波产品的应用中得到实践和验证。所以本申请兼顾手动方法的灵活性和低成本，在条件相同的情况下，又可以达到DPD芯片的预失真效果，是一种简单高效低成本的综合方法。
现在以微波产品为待检测装置为例进行说明，微波产品使用的数字芯片给出的预失真数学模型为：；
y＝x(g1+g3|x|2+g5|x|4)；
其中x为输入信号，y为输出信号，g1、g3和g5为多项式模型系数，这3个模型系数的取值决定了不同条件和因素下的DPD效果。经过试验证明3个模型系数中只有g3起决定作用。因此采数时仅寻优g3值。数字芯片手册 给出的g3范围为-2到+2。
流程部分的处理步骤如下：
第一步、样本评估。选择4个产品样本，由于调制方式取值有限，由其他因素的评估结果决定。先进行带宽评估，功率评估和频率评估。在数据采集模块中进行全遍历数据采集，样本评估模块使用聚类、拟合等方法对遍历得到的数据进行统计分析，得出带宽、功率和频率取值后，用这些值分别在高低温下进行温度评估。最后得到带宽为28M，功率为16dBm、21dBm和23dBm，调制方式为128QAM、512QAM和1024QAM，频点为21200MHz、21800MHz、22400MHz、23000MHz和23600MHz，常温33°，输出的数据文件为InterpolationData_23G.ini。高温和低温55°、5°、-15°、-33°和-40°单独输出一个文件为access数据库表usr_Coefficients_23G。
第二步、数据采集。利用样本评估得到的数据采集用例，通过自动化控制和采数软件的仪表控制模块设置频谱仪FSV，自动化控制和采数软件的设备控制模块设置微波射频单元设备SRU或AOU，使能DPD功能，并使其发出射频信号。同样通过自动化控制和采数软件的接口板控制模块控制RDC采数工装上的开关连通发射设备和频谱仪，使得发送的射频信号以频谱形式显示在频谱仪上。控制频谱仪按频率步进扫描得到一组功率值，反馈给数据采集模块，数据采集模块将数据输入数据处理模块进行分析和数据提取；
第三步、数据处理。对数据采集模块输入的数据在DPD模型系数g3的范围内遍历g3值，每次遍历的g3值反馈给数据采集模块控制发射信号进行DPD，DPD后的信号与欧标定义的标准信号进行功率对比，检测DPD效果，直到找到功率差值最优值，结束发射信号效果寻优，记录此时的DPD模型系数g3的最优值及范围，以此为基础，通过仪表控制模块设置信号源SMF100A发出接收信号，并通过接口板控制模块控制开关连通接收设备和信号源，检测接收设备的接收信号，进行接收信号的DPD效果验证和寻优，与发射信号寻优一样，接收信号寻优过程也与数据采集模块进行反馈循环，直到找到最优值。重复第二和第三步，直到遍历完成评估给出的所有用例。
第四步、DPD表生成。数据处理得到的最优值输入DPD表生成模块，该模块利用样本评估得到的全范围数据生成方法生成五种影响因素相关的完 整DPD数据，形成数据文件，通过自动化控制和采数软件的设备控制模块将数据文件写入发射设备，写完后在通过设备控制模块回读写入发射设备的数据进行校验。
具体应用中使用的DPD数据寻优算法如下：
第一步、发射信号。读取InterpolationData_23G.ini文件给出的用例，通过自动化控制和采数软件发送控制命令设置发射和接收设备，然后通过自动化控制和采数软件发送控制命令，切换RDC采数工装的射频开关连通发射设备和频谱仪FSV，使得发射设备发出要求的信号；
第二步、实部遍历寻优。在器件给定的DPD模型系数g3范围-2到+2内，固定DPD模型系数g3虚部，遍历实部，通过自动化控制和采数软件使能DPD功能，然后通过自动化控制和采数软件从频谱仪读取并分析频谱数据，寻找DPD效果最优的DPD数据；
第三步、虚部遍历寻优。在器件给定的DPD数据范围内，使用第二步得到的实部数据，遍历虚部，采用与第二步相同的方法分析频谱寻优；
第四步、实部二次寻优。在器件给定的DPD数据范围内，使用第三步得到的虚部数据，遍历实部，采用与第二步相同的方法分析频谱寻优；
第五步、重复第二到四步，发射信号与标准信号对比，直到找到最优值并记录，同时记录的还有DPD数据的寻优范围；
第六步、配置接收信号。通过自动化控制和采数软件配置信号源，自动化控制和采数软件控制发射设备使能DPD功能，然后通过自动化控制和采数软件切换接口板开关连通发射设备和信号源，以及信号源和接收设备，使得接收设备接收到要求的信号；
第七步、大步进寻优。在上面得到的DPD寻优范围内，以0.1为大步进遍历DPD模型系数g3，通过53模块读取接收设备上的接收信号的MSE值，用MSE值寻优DPD数据；
第八步、小步进寻优。大步进寻优结果输出一个数据范围，在该范围内，以0.02为小步进使用第七步同样的方法进一步小范围内寻优；
第九步、重复第七到第八步，直到连续两次最优接收信号对应的DPD 模型系数相同，结束接收信号寻优，记录寻优结果；
第十步、重复第一到九步，直到数据采集用例全部寻优结束，记录寻优最终结果，并输出寻优最终结果为一个excel文件，文件多以产品条码命令，例如219053719224.xlsm。
综上可知，通过本申请的实施，至少存在以下有益效果：
本申请提供了一种DPD表生成方法，考虑了所有的影响因素，使用统计学方法对所有因素进行评估分析，得到采数的影响因素组合用例，使用这些用例进行采数，既解决了离散条件选取问题，利用自动化也保证了数据的采集和生成效率；同时，在采数算法和完整的DPD数据生成算法上较现有的算法也有所改进，相关技术的DPD表生成方法，无论是离线写入的DPD表，还是实时训练得到的DPD表，均不会对表中的数据进行预失真效果检测和寻优，本申请给出的DPD表生成方法，其数据是在检测和寻优预失真效果前提下采集得到，得到后使用该数据使能DPD，并且配置和连通一条收、发链路，检测接收端设备信号，用接收信号进一步遍历、验证和寻优前面得到的DPD数据，从而保证DPD表中的数据都是预失真效果寻优后的数据，进而使得本申请中DPD表中的数据所达到的预失真效果要好于相关技术的DPD表生成方法获取的数据所达到的预失真效果。
本发明实施例另外提供一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令被执行时实现所述方法。
本领域普通技术人员可以理解所述方法中的全部或部分步骤可通过程序来指令相关硬件(例如处理器)完成，所述程序可以存储于计算机可读存储介质中，如只读存储器、磁盘或光盘等。可选地，所述实施例的全部或部分步骤也可以使用一个或多个集成电路来实现。相应地，所述实施例中的各模块/单元可以采用硬件的形式实现，例如通过集成电路来实现其相应功能，也可以采用软件功能模块的形式实现，例如通过处理器执行存储于存储器中的程序/指令来实现其相应功能。本发明实施例不限制于任何特定形式的硬件和软件的结合。
以上仅是本申请的具体实施方式而已，并非对本申请做任何形式上的限制，凡是依据本申请的技术实质对以上实施方式所做的任意简单修改、等同 变化、结合或修饰，均仍属于本申请技术方案的保护范围。
工业实用性
本申请给出的DPD表生成方法，其数据是在检测和寻优预失真效果前提下采集得到，得到后使用该数据使能DPD，并且配置和连通一条收、发链路，检测接收端设备信号，用接收信号进一步遍历、验证和寻优前面得到的DPD数据，从而保证DPD表中的数据都是预失真效果寻优后的数据，进而使得本申请中DPD表中的数据所达到的预失真效果要好于相关技术的DPD表生成方法获取的数据所达到的预失真效果。

Claims (11)

  1. 一种数字预失真表生成方法，包括：
    根据待处理装置的非线性数学模型，针对影响所述待处理装置数字预失真表的所有工作参数，对所述待处理装置的产品样本的所有工作参数进行全遍历数据采集，获取评估数据，根据所述评估数据生成数据采集用例及全范围的数据生成方法；
    使用所述数据采集用例对所述待处理装置采集数据，获取寻优数据；
    对所述寻优数据进行数字预失真效果的检测和寻优，寻优的结果作为非线性数学模型的模型系数；
    使用所述全范围的数据生成方法,根据所述模型系数生成数字预失真表。
  2. 如权利要求1所述的数字预失真表生成方法，其中，所述根据所述评估数据生成所述数据采集用例及所述全范围的数据生成方法的步骤包括：使用聚类以及拟合的方法对所述评估数据进行分析，得到所述所有工作参数之中的有限取值组合而成的有效的所述数据采集用例及所述全范围的数据生成方法。
  3. 如权利要求1或2所述的数字预失真表生成方法，其中，所述获取寻优数据的步骤包括：根据所述数据采集用例，对所述待处理装置的射频信号进行数字预失真测试；将数字预失真的射频信号以频谱形式显示，按频率步进扫描得到一组功率值，生成所述寻优数据。
  4. 如权利要求3所述的数字预失真表生成方法，其中，所述对所述寻优数据进行数字预失真效果的检测和寻优，将寻优结果作为所述非线性数学模型的模型系数，生成数字预失真表的步骤包括：将预失真后的信号与标准信号进行功率对比，检测数字预失真效果，直到找到功率差值最优值，在数字预失真表的范围内，先固定数字预失真虚部，进行实部遍历，确定最优的实部数据，将数字预失真实部固定为最优的实部数据，进行虚部遍历，确定最优的虚部数据，循环至最优值。
  5. 如权利要求4所述的数字预失真表生成方法，其中，所述对所述寻优数据进行数字预失真效果的检测和寻优的步骤包括：交替执行大步进寻优与小步进寻优，循环至最优值。
  6. 一种数字预失真表生成装置，包括评估模块、采集模块、寻优模块及生成模块，其中，
    所述评估模块设置为根据待处理装置的非线性数学模型，针对影响所述待处理装置数字预失真表的所有工作参数，使用所述采集模块对所述待处理装置的产品样本的所有工作参数进行全遍历数据采集，获取评估数据，根据所述评估数据生成数据采集用例及全范围的数据生成方法；
    所述采集模块设置为使用所述数据采集用例对所述待处理装置采集数据，获取寻优数据；
    所述寻优模块设置为对所述寻优数据进行数字预失真效果的检测和寻优，寻优结果作为非线性数学模型的模型系数；
    所述生成模块设置为使用所述全范围的数据生成方法,根据所述模型系数生成数字预失真表。
  7. 如权利要求6所述的数字预失真表生成装置，其中，所述评估模块设置为使用聚类以及拟合的方法对所述评估数据进行分析，得到所述所有工作参数之中的有限取值组合而成的有效的所述数据采集用例以及所述全范围的数据生成方法。
  8. 如权利要求6或7所述的数字预失真表生成装置，其中，所述采集模块设置为根据所述数据采集用例，对所述待处理装置的射频信号进行数字预失真测试；将数字预失真的射频信号以频谱形式显示，按频率步进扫描得到一组功率值，生成所述寻优数据。
  9. 如权利要求8所述的数字预失真表生成装置，其中，所述寻优模块设置为将预失真后的信号与标准信号进行功率对比，检测数字预失真效果，直到找到功率差值最优值，在数字预失真表的范围内，先固定数字预失真虚部，进行实部遍历，确定最优的实部数据，将数字预失真实部固定为最优的实部数据，进行虚部遍历，确定最优的虚部数据，循环至最优值。
  10. 如权利要求9所述的数字预失真表生成装置，其中，所述寻优模块设置为交替执行大步进寻优与小步进寻优，循环至最优值。
  11. 一种数字预失真系统，包括如权利要求6至10任一项所述的数字预失真表生成装置生成的数字预失真表。
PCT/CN2016/094674 2015-11-30 2016-08-11 数字预失真表生成方法、装置及数字预失真系统 WO2017092399A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510863670.0 2015-11-30
CN201510863670.0A CN106817328B (zh) 2015-11-30 2015-11-30 数字预失真表生成方法及装置

Publications (1)

Publication Number Publication Date
WO2017092399A1 true WO2017092399A1 (zh) 2017-06-08

Family

ID=58796169

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/094674 WO2017092399A1 (zh) 2015-11-30 2016-08-11 数字预失真表生成方法、装置及数字预失真系统

Country Status (2)

Country Link
CN (1) CN106817328B (zh)
WO (1) WO2017092399A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020149526A1 (en) * 2019-01-15 2020-07-23 Samsung Electronics Co., Ltd. Electronic device for reducing noise in frequency band of rx signal and method for the same

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111226397B (zh) * 2017-11-10 2022-05-17 上海诺基亚贝尔股份有限公司 一种在波束赋形中进行数字预失真处理的方法和装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1335699A (zh) * 2000-07-20 2002-02-13 华为技术有限公司 一种宽带发射机的自适应数字预失真方法和装置
CN101162913A (zh) * 2006-10-11 2008-04-16 大唐移动通信设备有限公司 预失真装置及方法
US20080144539A1 (en) * 2006-12-14 2008-06-19 Texas Instruments Incorporated Simplified Digital Predistortion in a Time-Domain Duplexed Transceiver
CN101478514A (zh) * 2009-01-08 2009-07-08 福建邮科通信技术有限公司 一种实现数字功率预失真的方法
CN103051574A (zh) * 2013-01-16 2013-04-17 大唐移动通信设备有限公司 数字预失真处理方法及系统

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103138686A (zh) * 2011-12-05 2013-06-05 佳律通信设备(上海)有限公司 一种功率放大器的非线性失真补偿装置
CN103051293B (zh) * 2012-12-03 2016-03-16 广东省电信规划设计院有限公司 射频放大器预失真处理方法及其系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1335699A (zh) * 2000-07-20 2002-02-13 华为技术有限公司 一种宽带发射机的自适应数字预失真方法和装置
CN101162913A (zh) * 2006-10-11 2008-04-16 大唐移动通信设备有限公司 预失真装置及方法
US20080144539A1 (en) * 2006-12-14 2008-06-19 Texas Instruments Incorporated Simplified Digital Predistortion in a Time-Domain Duplexed Transceiver
CN101478514A (zh) * 2009-01-08 2009-07-08 福建邮科通信技术有限公司 一种实现数字功率预失真的方法
CN103051574A (zh) * 2013-01-16 2013-04-17 大唐移动通信设备有限公司 数字预失真处理方法及系统

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020149526A1 (en) * 2019-01-15 2020-07-23 Samsung Electronics Co., Ltd. Electronic device for reducing noise in frequency band of rx signal and method for the same
US11063735B2 (en) 2019-01-15 2021-07-13 Samsung Electronics Co., Ltd. Electronic device for reducing noise in frequency band of Rx signal and method for the same

Also Published As

Publication number Publication date
CN106817328A (zh) 2017-06-09
CN106817328B (zh) 2020-08-21

Similar Documents

Publication Publication Date Title
CN109753762B (zh) 基于类别修正的配电网两阶段网络拓扑识别方法及装置
WO2017092399A1 (zh) 数字预失真表生成方法、装置及数字预失真系统
CN112272036B (zh) 用于射频接收机的温度补偿装置和方法、射频接收机
EP3667952A1 (en) Method, device, and storage medium for locating failure cause
CN113255778A (zh) 基于多模型融合的焊点质量检测方法、装置及存储介质
JP2022186677A (ja) 試験測定システム
CN114389751B (zh) 一种用于误码测试设备的自动均衡方法和系统
CN108768550A (zh) 基于动态多核带宽广义回归神经网络算法的宽频段发信机非线性建模方法
CN116423003A (zh) 一种基于数据挖掘的锡焊智能评估方法及系统
CN109257952A (zh) 以减少的数据量进行数据传输的方法、系统和程序产品
CN113591390A (zh) 一种接收机射频链路非线性效应的模型筛选方法、平台
CN112953961A (zh) 配电房物联网中设备类型识别方法
CN106597098B (zh) 一种频谱分析仪的数据处理方法和装置
CN104363066B (zh) 一种基于网络化的微波中继合路器自动测试方法
US20170160714A1 (en) Acquisition of high frequency data in transient detection
Choi et al. High-speed link analysis with system identification approach
JP3787840B2 (ja) 波形解析装置
CN114358121A (zh) 基于变电站设备的监测方法、装置和终端设备
US20170286611A1 (en) System and method for using data obtained from a group of geographically dispersed magnetic resonance systems to optimize customer-specific clinical, operational and/or financial performance
CN113054919B (zh) 一种数字预失真信号反馈电路的测试方法、装置及设备
WO2024066784A1 (zh) 测试设备稳定性的监控方法、装置、电子设备和存储介质
CN114608705B (zh) 一种光谱信号数据采样和峰值检测方法
CN204422642U (zh) 电焊机电压模拟前端信号检测电路
CN117369422B (zh) 一种控制器出厂参数校准方法、参数校准装置及存储介质
CN112884201B (zh) 基于人工智能算法的非介入式负荷监测方法、装置及系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16869710

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16869710

Country of ref document: EP

Kind code of ref document: A1