Digital Predistortion With Sub-Sample Shift for In-Band Linearity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current distortion compensation methods for power amplifiers in wireless communication systems face challenges in achieving efficient distortion compensation, particularly in high-bandwidth scenarios like 5G submillimeter wave/millimeter wave bands, due to high operating frequencies required for DPD, which exceed the performance capabilities of existing hardware, and struggle with compensating for non-linear distortions within the carrier band.
Innovation Solution
A distortion compensation apparatus and method that operates at a lower sampling rate by incorporating a pseudo-interpolation/sub-sample-shift processing unit and a sub-sample delay filter, allowing for distortion compensation at sample and sub-sample points without the need for upsampling, thereby reducing the number of coefficients and operating frequencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If digital predistortion (DPD) is used to compensate for non-linear distortion in power amplifiers, then linearity of the output signal is improved and output distortion is suppressed, but the operating frequency required for DPD increases exponentially with bandwidth
Solution Approach 1:
The patent segments the distortion compensation task by separating in-band distortion compensation from out-of-band distortion compensation. The DPD apparatus compensates only for in-band distortion components within the carrier band, while out-of-band distortion is handled separately through filtering. This segmentation allows the DPD to operate at a lower sampling frequency (equal to the carrier band bandwidth) rather than requiring a frequency proportional to the total signal bandwidth, thereby resolving the contradiction between linearity improvement and operating frequency increase.
2Productivity
If the bandwidth of the input signal is increased to achieve high-capacity communication, then communication capacity is improved, but the memory effect in the power amplifier becomes more noticeable and DPD performance is limited
Solution Approach 1:
The patent extracts and removes out-of-band distortion components from the feedback signal before feeding it back to the DPD apparatus. By using a band-pass filter to eliminate out-of-band components and an offset frequency shift to remove DC offsets, the system ensures that the DPD only processes in-band distortion. This extraction approach prevents the memory effect from degrading DPD performance even when the input signal bandwidth is increased for high-capacity communication.
3Manufacturing precision
If the number of coefficients in the DPD model is increased to improve compensation performance, then distortion compensation accuracy is improved, but the complexity of the DPD apparatus and calculation burden increase
Solution Approach 1:
By segmenting the distortion compensation scope to only in-band components, the patent reduces the effective bandwidth over which the DPD must operate. This reduction in operational bandwidth decreases the number of coefficients required in the DPD model to achieve the same compensation accuracy, thereby reducing both the device complexity and calculation burden while maintaining distortion compensation performance.
4Manufacturing precision
If upsampling is performed before DPD to compensate for sub-sample memory effects, then compensation performance is improved, but the operating frequency and number of coefficients increase
Solution Approach 1:
Instead of upsampling the signal before DPD to achieve sub-sample resolution (the conventional approach), the patent inverts the approach by operating the DPD at the original sampling frequency and using fractional delay filters to achieve sub-sample memory effect compensation. This inversion allows the system to maintain the original sampling frequency while still compensating for sub-sample effects, thereby avoiding the increase in operating frequency and coefficient count that would result from upsampling.
Data Source
AI summary
A DPD (1) includes: a polynomial structure including a pseudo-interpolation/sub-sample-shift processing unit (101) configured to operate at a sampling rate for sampling an input signal not upsampled in a previous stage of the DPD (1), pseudo-interpolate a sample point between sample points of the input signal, and shift the pseudo-interpolated sample point by a sub-sample and a multiplexer (109) configured to select a combination of a sub-sample shift amount; and an FIR filter (107) configured to be provided in a subsequent stage of the polynomial structure and include a sub-sample delay filter delaying a sample point of the input signal by a sub-sample. The DPD (1) compensates for distortion due to a sample point of the input signal and compensates for distortion due to a sub-sample point between sample points of the input signal for the DPD (1), by using the polynomial structure and the FIR filter (107).