A digital predistortion system with dynamically configurable piecewise linear lookup table and index

Abstract

The present invention discloses a digital predistortion system with dynamically configurable piecewise linear lookup tables and indexes, belonging to the technical field of wireless communication design. The digital predistortion system comprises a digital / analog converter, an upconversion, a power amplifier, a gain attenuator, a downconversion, an analog / digital converter, a DPD trainer, and a DPD executor. The DPD trainer comprises a DPD inverse distortion unit, a 1dB compression point calculation unit, a piecewise linear fitting unit, and a lookup table index and storage allocation unit. The DPD executor comprises a delay alignment unit, a lookup table index configuration unit, a lookup table unit, and a multiplication-accumulation unit. The method structure can dynamically allocate the segmentation interval density according to the position of the 1dB compression point of the nonlinear curve, that is, the segmentation density is increased near the 1dB compression point of the PA to improve the linearization accuracy, and the segmentation can be sparsely spaced at other positions to reduce the amount of calculation.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication technology design, and in particular relates to a digital predistortion system with a dynamically configurable piecewise linear lookup table and index. Background Art

[0002] Digital predistortion (DPD) technology is widely used in wireless communications. It corrects for severe signal nonlinear distortion caused by the inherent memory nonlinearity of power amplifiers (PAs) operating in their high-efficiency range. To save hardware costs and ensure real-time transmission, DPD typically uses a lookup table to compensate for PA nonlinear distortion.

[0003] A common approach to traditional lookup tables is to uniformly sample the nonlinear curve, that is, the input data interval is the same, and the sampled values ​​are stored in the lookup table. However, this approach has two problems. First, if the input data interval is too small, the number of sampling points will inevitably increase. Although this will improve accuracy, the depth of the lookup table will increase, resulting in a large consumption of resources. Second, if the data interval is too large, the accuracy of the lookup table will inevitably decrease. Another lookup table approach uses non-uniform sampling, that is, the input data interval is not fixed. Although a close inverse distortion model can be obtained through the software least squares method, and then the lookup table is implemented through piecewise linear fitting, it is easy to fall into local optimal solutions. In addition, the non-uniform intervals in some designs are fixed and cannot be matched. This will have certain limitations on the stability, flexibility, and application of DPD in different scenarios. Summary of the Invention

[0004] The purpose of the present invention is to provide a digital predistortion system with a dynamically configurable piecewise linear lookup table and index. This method structure can dynamically allocate the segmented interval density based on the position of the 1dB compression point of the nonlinear curve. Specifically, the segmented density is increased near the 1dB compression point of the PA inverse distortion model to improve linearization accuracy, while other locations can be sparsely segmented to reduce the amount of computation. Based on the concept of piecewise linearization, the present invention can dynamically adjust the segmented intervals and the content of the lookup table in the chip hardware part. This structure and method not only help improve system accuracy, stability, and flexibility, but also can adapt to tracking the nonlinear changes of the PA in different scenarios, which can solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solutions: a digital predistortion system with a dynamically configurable piecewise linear lookup table and index, the digital predistortion system comprising a digital / analog converter, an upconverter, a power amplifier, a gain attenuator, a downconverter, an analog / digital converter, a DPD trainer, and a DPD actuator;

[0006] During operation, the PDD actuator receives an input signal x(n) and performs pre-distortion processing. The pre-distorted signal x(n) is converted into an analog signal through a digital-to-analog converter, then up-converted and modulated to the radio frequency. It is then amplified by a power amplifier to form the output signal y(n). At the same time, the feedback path of the output signal y(n) is activated. The output signal y(n) is then down-converted and demodulated to baseband via a gain attenuator. It is then digitized by an analog-to-digital converter to form a feedback signal, which is then input into the DPD trainer. The DPD trainer analyzes the error between the input signal x(n) and the feedback signal in real time, dynamically generates compensation parameters, and updates the DPD actuator to form a closed-loop control.

[0007] The DPD trainer includes a DPD inverse distortion unit, a 1dB compression point calculation unit, a piecewise linear fitting unit, and a lookup table index and storage allocation unit.

[0008] DPD inverse distortion unit, used for reverse modeling and training of PA devices;

[0009] 1dB compression point calculation unit, used to determine the input value corresponding to the 1dB compression point of the DPD inverse distortion unit;

[0010] The piecewise linear fitting unit is used to perform dense segmentation processing on points near the 1dB compression point input value, and perform sparse segmentation processing on other areas;

[0011] A lookup table index and storage allocation unit, used to ensure a one-to-one correspondence between the segment index value and the corresponding storage content lookup table segment index;

[0012] Among them, the DPD executor includes a delay alignment unit, a lookup table index configuration unit, a lookup table unit and a multiplication and accumulation unit;

[0013] The delay alignment unit is used to delay the input data to ensure that the index value delay is aligned with the input data after passing through the DPD trainer;

[0014] The lookup table index configuration unit is used to dynamically configure the lookup table index interval and the lookup table content, determine which segment index range the input data belongs to based on the modulus value, and map the actual lookup table address;

[0015] The lookup table unit is used to perform a lookup operation based on the address index and output the matching value;

[0016] The multiplication and accumulation unit is used to multiply and accumulate the values ​​in the lookup table with the memory items of different depths of the DPD and output them.

[0017] Preferably, the specific execution process steps of the DPD trainer are as follows:

[0018] a. The DPD inverse distortion unit reverse models and trains the PA device to obtain an inverse distortion model within the error range;

[0019] b. Based on the DPD model trained in step a, pass it through the 1dB compression point calculation unit to obtain the input value corresponding to the 1dB compression point;

[0020] c. After obtaining the input value corresponding to the 1dB compression point in step b, construct a window around the value, perform dense segmentation processing within the window, and perform sparse segmentation processing on the area outside the window;

[0021] d. After the piecewise linear fitting in step c is completed, the lookup table segment index value and the corresponding content are sent to the lookup table index and storage allocation unit, and then the segment index value is matched with the lookup table content value and sent to the DPD executor in the digital domain.

[0022] Preferably, the specific execution process steps of the DPD executor are as follows:

[0023] a. After the digital domain input data enters the DPD executor, it is processed by the delay alignment unit. The delayed data is an integer multiple of the digital domain clock period.

[0024] b. Send the DPD trainer data and the delayed data from step a to the lookup table index configuration unit. Dynamically configure the index boundary values ​​and lookup table contents. Perform a modulus calculation on the delayed aligned input data. This modulus value is matched with the actual index value in the lookup table to determine the segmented index range within which it falls and map it to the actual index value in the lookup table.

[0025] c. Use step b to find the table index value to complete the DPD table lookup function;

[0026] d. Perform multiplication and accumulation operations on the value found in step c and the DPD memory items of different depths, and output the DPD predistortion value.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The digital predistortion system with dynamically configurable piecewise linear lookup table and index can dynamically allocate the segment interval density according to the position of the 1dB compression point of the nonlinear curve. That is, the segment density is increased near the 1dB compression point of the PA to improve the linearization accuracy, and the other positions can be sparsely segmented to reduce the amount of calculation. It not only helps to improve the system accuracy, stability and flexibility, but also can adapt to and track the nonlinear changes of the PA in different working scenarios such as temperature changes, different chip manufacturing process angles, and PA load impedance, and make real-time corrections to them. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a typical DPD digital predistortion system block diagram.

[0030] Figure 2 This is a block diagram of the DPD trainer and DPD actuator modules of a digital predistortion system with a dynamically configurable piecewise linear lookup table and index.

[0031] Figure 3 It is the nonlinear curve diagram of the DPD inverse distortion model.

[0032] Figure 4 A block diagram of the DPD executor configuration storage implementation of the present invention.

[0033] Figure 5 This is a diagram illustrating the implementation of multiplication and accumulation operations for different memory items of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0035] DPD technology is a PA linearization compensation technology commonly used in current base station transmitters. Its main idea is to configure a digital predistorter in the digital domain that is opposite to the PA's nonlinear characteristics, thereby ensuring a linear response when the PA outputs at high efficiency.

[0036] The block diagram of a typical DPD digital predistortion system is as follows: Figure 1 As shown, the dotted box is the improved part of the present invention. The present invention improves the update method of the lookup table in the DPD trainer and the implementation method of the digital domain DPD actuator, and adopts a dynamic configurable piecewise linear lookup table and index digital predistortion system, such as Figure 2 Compared with the traditional structural method, the DPD trainer in the present invention adds a unit for calculating the 1dB compression point and its corresponding lookup table index and storage allocation unit, and the DPD executor makes a dynamic configuration improvement on the lookup table index.

[0037] Figure 2This is a block diagram of the DPD trainer and DPD actuator modules of the digital pre-distortion system with dynamically configurable piecewise linear lookup tables and indexes of the present invention. The DPD inverse distortion unit in the DPD trainer is used for reverse modeling and training of PA devices; the 1dB compression point calculation unit is used to determine the input value corresponding to the 1dB compression point of the DPD inverse distortion unit; the piecewise linear fitting unit is used to perform dense segmentation processing on points near the 1dB compression point input value and sparse segmentation processing on other areas. The lookup table index and storage allocation unit is used to ensure that the segment index value corresponds one-to-one with the corresponding storage content lookup table segment index. In the DPD actuator, the delay alignment unit delays the input data to ensure that the index value delay remains aligned with the input data after passing through the DPD trainer. The lookup table index configuration unit dynamically configures the lookup table index interval and the lookup table content, determines which segment index range the input data modulus value is in, and maps the actual lookup table address. The lookup table unit performs a lookup operation based on the address index and outputs the matching value. The multiplication and accumulation unit is used to multiply and accumulate the value in the lookup table with the memory items of different depths of the DPD and output them.

[0038] The Volterra series is the fundamental model used in PA memory nonlinearity research. Most derived models are simplified versions of the Volterra series. The MP (Memory Polynomial) model has been widely researched and applied, becoming the most common behavioral model in digital predistortion (DPD). The MP model for PAs is shown below:

[0039]

[0040] The DPD model is the inverse form of the PA model, and the corresponding model is as follows:

[0041]

[0042] 1. The DPD trainer implementation process is as follows:

[0043] The key point of the DPD inverse distortion model is to find a pm The value of a pm The number of coefficients is P × M. The most common method is to use the least squares method. By sampling the PA model, multiple groups [x(n), y(n)] are obtained. The polynomial coefficient a can be obtained according to the Cholesky decomposition method. pm .

[0044] After the DPD inverse distortion model is determined, the 1dB compression point [x(n 1dB ),y(n 1dB )],like Figure 3 As shown, and temporarily stored. 1dB) is constructed around a window of size [a,b], where x(n 1dB )∈[a,b], where the values ​​of a and b are adjustable. Once the window size is determined, piecewise linear fitting is performed. Because each memory term in the DPD model corresponds to its own polynomial, from different time points, DPD is essentially the superposition of nonlinear combinations in different time domains. Therefore, to facilitate the description of the piecewise linear fitting process, only the DPD when m=0 is used for piecewise linear fitting. The processing method for other time domains is the same. When m=0, the DPD model becomes as follows:

[0045]

[0046] The above nonlinear formula can be fitted with a piecewise linear function, as follows:

[0047]

[0048] Among them, q represents the rank of the segmented interval, and Q is the total number of interval segments. Another form is as follows:

[0049]

[0050] According to the calculation, x(n 1dB ), dense segmentation is performed within the [a, b] window size range, and sparse segmentation is performed in other areas. After the piecewise linear fit is completed, the different segment intervals and the corresponding slope K values ​​are processed and saved one by one and sent to the digital domain DPD executor.

[0051] 2. The DPD actuator implementation process is as follows:

[0052] Because the data collected by the DPD trainer is obtained from the PA feedback loop, this loop must have a delay, so a delay alignment unit is required to delay the input x(n) data. This is easy to implement in the digital domain. You only need to define multi-level registers and perform beat processing.

[0053] After the input x(n) is delayed and aligned, x(n delay ), we need to judge the corresponding modulus value |x(n delay )|In which index segment interval is it located? The segment interval index comes from the DPD trainer. The lookup table index configuration unit in DPD configures and stores the boundary values ​​corresponding to each segment interval. Its implementation diagram is as follows: Figure 4 ,a0_reg,a1_reg,a n _reg corresponds to the boundary value of the piecewise linear interval, and these values ​​can be dynamically configured by the processor via the bus. delay )|For which interval, it is mapped to the corresponding lookup table address Addr.

[0054] After the lookup table address Addr is determined, the content in the table is queried according to the address. The content in the table is the slope K and the y value corresponding to the lower boundary of the corresponding segment interval. Here, y is essentially the intercept, which has been determined by the piecewise linear processing process in the DPD trainer. In order to facilitate processing in the digital domain, the K value needs to be scaled. In this part, |x(n delay )|The corresponding value y(n delay ) query and calculation, assuming |x(n delay )|In the piecewise linear interval [a1, a2), the corresponding value is calculated as follows after looking up the table:

[0055]

[0056] The above is the implementation process of a memory depth item. The calculation methods of other memory depth items are essentially the same. In the DPD pre-distortion model, a lookup table method is used. Since the nonlinear memory DPD model has several different memory items, there are corresponding lookup tables. For different memory items, multiplication and accumulation operations need to be performed. This operation is implemented as follows Figure 5 As shown, the shaded part is the corresponding multiplication and accumulation operation, Z -1 The tapping operation corresponding to one clock cycle in the digital domain multiplies the input modulus value by the different memory items and finally accumulates them to obtain the final predistortion value Y(n). Due to its simple structure, it is easy to implement in the digital domain.

[0057] The digital pre-distortion system with a dynamically configurable piecewise linear lookup table and index not only helps improve system accuracy, stability, and flexibility, but also can adapt to and track nonlinear changes in the PA under different operating scenarios such as temperature changes, different chip manufacturing process angles, and PA load impedance, and make real-time corrections to them.

[0058] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0059] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A digital predistortion system with a dynamically configurable piecewise linear lookup table and index, characterized in that: The digital predistortion system includes a digital / analog converter, an up-conversion, a power amplifier, a gain attenuator, a down-conversion, an analog / digital converter, a DPD trainer, and a DPD actuator; During operation, the PDD actuator receives an input signal x(n) and performs pre-distortion processing. The pre-distorted signal x(n) is converted into an analog signal through a digital-to-analog converter, then up-converted and modulated to the radio frequency. It is then amplified by a power amplifier to form the output signal y(n). At the same time, the feedback path of the output signal y(n) is activated. The output signal y(n) is then down-converted and demodulated to baseband via a gain attenuator. It is then digitized by an analog-to-digital converter to form a feedback signal, which is then input into the DPD trainer. The DPD trainer analyzes the error between the input signal x(n) and the feedback signal in real time, dynamically generates compensation parameters, and updates the DPD actuator to form a closed-loop control. The DPD trainer includes a DPD inverse distortion unit, a 1dB compression point calculation unit, a piecewise linear fitting unit, and a lookup table index and storage allocation unit. DPD inverse distortion unit, used for reverse modeling and training of PA devices; 1dB compression point calculation unit, used to determine the input value corresponding to the 1dB compression point of the DPD inverse distortion unit; The piecewise linear fitting unit is used to perform dense segmentation processing on points near the 1dB compression point input value, and perform sparse segmentation processing on other areas; A lookup table index and storage allocation unit, used to ensure a one-to-one correspondence between the segment index value and the corresponding storage content lookup table segment index; Among them, the DPD executor includes a delay alignment unit, a lookup table index configuration unit, a lookup table unit and a multiplication and accumulation unit; The delay alignment unit is used to delay the input data to ensure that the index value delay is aligned with the input data after passing through the DPD trainer; The lookup table index configuration unit is used to dynamically configure the lookup table index interval and the lookup table content, determine which segment index range the input data belongs to based on the modulus value, and map the actual lookup table address; The lookup table unit is used to perform a lookup operation based on the address index and output the matching value; The multiplication and accumulation unit is used to multiply and accumulate the values ​​in the lookup table with the memory items of different depths of the DPD and output them.

2. The digital predistortion system with dynamically configurable piecewise linear lookup table and index according to claim 1, characterized in that: The specific execution process steps of the DPD trainer are as follows: a. The DPD inverse distortion unit reverse models and trains the PA device to obtain an inverse distortion model within the error range; b. Based on the DPD model trained in step a, pass it through the 1dB compression point calculation unit to obtain the input value corresponding to the 1dB compression point; c. After obtaining the input value corresponding to the 1dB compression point in step b, construct a window around the value, perform dense segmentation processing within the window, and perform sparse segmentation processing on the area outside the window; d. After the piecewise linear fitting in step c is completed, the lookup table segment index value and the corresponding content are sent to the lookup table index and storage allocation unit, and then the segment index value is matched with the lookup table content value and sent to the DPD executor in the digital domain.

3. The digital predistortion system with dynamically configurable piecewise linear lookup table and index according to claim 2, characterized in that: The specific execution process steps of the DPD executor are as follows: a. After the digital domain input data enters the DPD executor, it is processed by the delay alignment unit. The delayed data is an integer multiple of the digital domain clock period. b. Send the DPD trainer data and the delayed data from step a to the lookup table index configuration unit. Dynamically configure the index boundary values ​​and lookup table contents. Perform a modulus calculation on the delayed aligned input data. This modulus value is matched with the actual index value in the lookup table to determine the segmented index range within which it falls and map it to the actual index value in the lookup table. c. Use step b to find the table index value to complete the DPD table lookup function; d. Perform multiplication and accumulation operations on the value found in step c and the DPD memory items of different depths, and output the DPD predistortion value.

Images