Multiband Radio Predistortion Using Lower-Dimensional LUTs

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Traditional digital predistortion (DPD) methods for multiband radios face scalability issues due to high computational complexity and unmanageable memory requirements as the number of frequency bands increases beyond two, making them costly and inefficient.

Innovation Solution

The implementation of reduced dimensionality separate digital predistortion (S-DPD) systems that use B-q dimensional hybrid Look-Up Tables (LUTs, where q={1, 2, ..., B-2}, and in some cases, one-dimensional LUTs, to decompose the B-dimensional LUT problem into lower-dimensional LUT problems, reducing computational cost and memory requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Speed

If traditional LUT based methods are used for linearization, then fast computation is achieved, but memory requirements become unmanageable as the number of frequency bands increases

Engineering Contradiction:
Improvecomputation speedVSAvoidmemory requirements
Core Design Contradiction:
SpeedVSQuantity of substance

Solution Approach 1:

The patent segments the B-dimensional LUT into multiple lower-dimensional LUTs by decomposing the multiband predistortion function into separate band-specific components. Each frequency band is processed independently using its own LUT, avoiding the need for a single large B-dimensional structure. This segmentation maintains fast lookup computation while reducing memory requirements from exponential to linear scaling with the number of bands.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent transitions from a B-dimensional LUT structure to multiple 1-dimensional LUTs by changing the dimensional organization. Instead of storing predistortion values for all B bands simultaneously in a high-dimensional space, the system uses B separate 1-dimensional LUTs, each handling one band. This dimensionality reduction transforms an unmanageable memory structure into a scalable solution.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Adaptability or versatility

If separate digital predistortion with B-dimensional LUTs is used to support B frequency bands, then each band is processed independently, but device complexity increases significantly

Engineering Contradiction:
Improvefrequency band supportVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The system segments the multiband predistortion processing into B independent single-band processing chains, each with its own LUT and control logic. This segmentation allows each band to be handled separately with simple, identical processing modules, reducing overall system complexity compared to a monolithic B-dimensional approach while maintaining full frequency band support.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent creates a universal processing template that can be instantiated B times, once for each frequency band. Each band processor uses the same LUT-based predistortion logic, making the system modular and easier to implement. This universal approach reduces design complexity while supporting any number of frequency bands.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Measurement precision

If high order polynomials are evaluated for multiband systems, then accurate predistortion is achieved, but computational cost becomes equally expensive as traditional LUT methods

Engineering Contradiction:
Improvepredistortion accuracyVSAvoidcomputational cost
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent pre-computes and stores predistortion values in LUTs during an offline training phase, copying the complex polynomial evaluation results into a lookup structure. During runtime, instead of evaluating high-order polynomials, the system simply copies pre-computed values from the LUT based on input amplitude and phase. This copying approach maintains predistortion accuracy while dramatically reducing computational cost.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The system performs preliminary polynomial evaluation and predistortion computation during an offline training phase, storing results in LUTs before actual operation. This preliminary action moves the computationally expensive operations from runtime to setup time, allowing fast lookup during actual predistortion application while maintaining the accuracy benefits of high-order polynomial modeling.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentUS20240063760A1Practical predistortion architectures for multiband radios
Publication Date: 2024.02.22 TELEFONAKTIEBOLAGET LM ERICSSON (PUBL)
  • US20240063760A1 patent drawing
  • US20240063760A1 patent drawing
  • US20240063760A1 patent drawing

AI summary

Systems and methods for separate digital predistortion (S-DPD) for multi-band radios are disclosed. In one embodiment, a method of operation of a digital predistortion (DPD) actuator system for a radio node for a wireless network comprises receiving a plurality of input signals (x1, . . . , xB) for a plurality of frequency bands (b1, . . . , bB), respectively. The method further comprises, for each frequency band (bl) of the plurality of frequency bands (b1, . . . , bB), generating a plurality of predistorted signals for the frequency band (bl) based on the plurality of input signals (x1, . . . , xB) and a plurality of Look-Up Tables (LUTs) each having less than B dimensions and combining the plurality of predistorted signals for the frequency band (bl) to provide a combined predistorted signal for the frequency band (bl). In this manner, lower dimensionality LUTs are used, which in turn reduces cost and complexity of the DPD system.