MIKON 2024 Workshop
“Digital Predistortion for 5G/6G Wireless Transmitters“
Organized by Professor Anding Zhu, University College Dublin, Ireland
Digital predistortion (DPD) is a linearization technique that uses an inverse function in digital baseband to compensate for the distortion induced by power amplifiers at RF frequencies in wireless transmitters. By using DPD, the RF power amplifiers can be operated in high efficiency mode without losing linearity, which dramatically increases energy efficiency of the wireless systems. DPD has become is one of the most fundamental building blocks in wireless transmitters today. The use of DPD is expected to continue in 5G and 6G systems, but the shift from single antenna to multiple-input multiple-output (MIMO) phased arrays and the ever-increasing signal bandwidth have presented significant challenges for DPD designers in managing power consumption and meeting linearity requirements. This workshop will discuss recent research advances and future trends in using DPD in future broadband wireless systems, particularly focusing on potential applications of machine learning (ML) and artificial intelligent (AI) techniques.
Presenters:
Talk 1: Digital Predistortion for Broadband Transmitters Using Machine Learning
Abstract: DPD has been widely adopted to keep RF power amplifier operating with high efficiency without losing linearity in modern wireless systems, but the ever-increasing signal bandwidth have presented significant challenges for DPD designers in managing power consumption and meeting linearity requirements. This presentation will first discuss the limitations of conventional techniques and highlight the challenges we are facing in the new systems. Recent research advances using machine learning techniques will be presented, including new advanced model architectures, complexity reduction techniques and fast model adaptation algorithms.
Speaker’s Bio: Anding Zhu received the Ph.D. degree in electronic engineering from University College Dublin (UCD), Ireland, in 2004. He is currently a Professor with the School of Electrical and Electronic Engineering at UCD. His research interests include high-frequency nonlinear system modeling and device characterization techniques, high-efficiency power amplifier design, and nonlinear system identification algorithms. He has published more than 200 peer-reviewed journal and conference articles. Prof. Zhu is an IEEE Fellow. He serves as an Elected Member of Administrative Committee of IEEE Microwave Theory and Technology Society (MTT-S) since 2019 and is currently the Chair of the Technical Coordination and Future Directions Committee.
Talk 2: Artificial Neural Networks for Digital Predistortion Linearization of Multi-input Wideband Power Amplifiers
Abstract: In this talk we will discuss the use of artificial neural networks (ANNs) for digital predistortion (DPD) linearization of highly efficient power amplifiers (PAs) when operated with wideband OFDM-based communications signals. ANNs linearization capabilities will be compared to traditional polynomial-based DPD behavioral models and N-stage cascaded models. Experimental results considering the linearization of Analog Devices’ dual-input pseudo-Doherty load-modulated balanced amplifier will be provided. Finally, the ANN-based DPD implementation using a graphic processing unit (GPU) with compute unified device architecture (CUDA) units will be discussed.
Speaker’s Bio: Pere L. Gilabert received his degree in Telecommunication Eng. from the UPC in 2002. In February 2002 he visited the INFOCOM department of the University of Rome “La Sapienza” where he developed his MSc. thesis with an Erasmus exchange grant. He joined the department of Signal Theory and Communications (TSC) in 2003 and received his PhD from the UPC in February 2008. As a result of his research activity, he was awarded with the Extraordinary Doctorate Award 2007-08 in the field of ICT Engineering. In 2010 he received his degree in Humanities from the Universitat Oberta de Catalunya (UOC). Since 2022 he is the head of the Components and Systems for Communications (CSC) research group of the TSC department. His research activity is in the field of linearization techniques and digital signal processing solutions for highly efficient transmitter architectures.
Talk 3: Phase-Normalized Neural Networks for Linearization of Power Amplifiers
Abstract: Digital predistortion (DPD) is a key technique for maintaining the linearity of RF power amplifiers (PAs) in radio base stations. In the upcoming 6G era, signal quality will be critical to enable high throughput communications, while large transmission bandwidths complicate the linearization task and challenge state-of-the-art DPD models. Machine learning DPD using real-valued artificial neural networks (NNs) is considered as a potential solution to offer improved linearization performance, scalability, and flexibility. However, commonly adopted real-valued NN structures lack the ability to efficiently handle the base-band I/Q phase. This talk introduces phase normalization to unlock the missing potential in real-valued NNs and achieve a superior modelling and linearization ability with a favourable complexity to performance trade-off. Adoption of the proposed phase normalization to feed-forward as well as recurrent neural network structures are presented.
Speaker’s Bio: Arne Fischer-Bühner graduated as Diplom-Ingenieur (Dipl-Ing.) in electrical engineering from the Technical University Dresden in March 2021. He is currently a researcher at Nokia Bell Labs in Antwerp, Belgium, where he is pursuing a Marie-Curie funded industrial doctorate together with Tampere University, Finland. His research interests include AI/ML for modeling and compensation of nonlinear impairments in RF transmitters, DFE signal processing, and related processing architectures and systems.
Talk 4: AI-assisted Digital Predistortion of RF Power Amplifiers for 6G Flexible Spectrum Applications
Abstract: As wireless communication systems continuously evolve, the issue of spectrum scarcity becomes increasingly prominent. In 6G, adopting flexible spectrum technologies to enhance spectrum utilization is critical, which poses significant challenges for radio frequency (RF) power amplifiers (PAs). This will require the technological evolution to adapt to dynamic changes and ensure compatibility with versatile spectrum scenarios. In this talk, we will introduce novel digital predistortion techniques specifically designed for 6G flexible spectrum applications, with a focus on artificial intelligence (AI)-assisted digital predistortion technologies. These AI-assisted methods are employed to model and linearize the nonlinear characteristics of power amplifiers in high-efficiency, high-linearity signal transmission across varying conditions. Moreover, by combining these digital predistortion techniques with digital-assisted control mechanisms, the PA can be dynamically optimized to accommodate the flexible spectrum environment while sustaining high efficiency and linearity. The innovative approaches presented herein offer highly promising solutions for future wireless communication systems.
Speaker’s Bio: Chao Yu received the B.E. degree in information engineering and the M.E. degree in electromagnetic fields and microwave technology from Southeast University (SEU), Nanjing, China, in 2007 and 2010, respectively, and the Ph.D. degree in electronic engineering from University College Dublin (UCD), Dublin, Ireland, in 2014. He is currently a Professor with the School of Information Science and Engineering, SEU, Nanjing. His current research interests include microwave and millimeter wave power amplifier modeling and linearization, and 5G massive multiple-input-multiple-output (MIMO) RF system design. He has authored and co-authored over 200 academic journal articles and conference papers, with over 80 contributions appearing in prominent journals within the field and holds more than 30 authorized patents. He is a recipient of the 2021 IEEE MTT-S Microwave Prize.
Talk 5: Failures and Successes During the Development of a Custom DPD Solution for Cellular Base Stations
Abstract: DPD is a linearization technique that applies an inverse function in digital baseband to compensate for the nonlinear distortion induced in analog PAs. The concept is simple but making it work effectively and efficiently is far from being an easy task. In this presentation, we will discuss the practical issues encountered in DPD development for cellular base stations. We will first explain the design requirements according to the standards and then outline the experimental configuration. The experimental validation will be discussed, particularly focusing on the situation when the validation fails. The example issues can be follows: “Houston, we have a problem. The PA is dead.” – safety measures when experimenting with a DPD implementation, which is not well validated yet (sanity checks, proper system monitoring, power leveling, etc.). “There is hysteresis in our PA!” – a thermal-related issue discovered when measuring Pout versus Pin characteristics, “With this dummy load I cannot RX!” – an interesting issue with PIM (Passive Intermodulation) phenomena observed during validation of the FDD base station’s front end.
Speaker’s Bio: Przemysław Korpas received a M.Sc. degree in telecommunications engineering in 2010 and a PhD with honours in electronics in 2015 – both at the Warsaw University of Technology. His research interests include applications of Software Defined Radio technology in the field of telecommunications and RF measurement equipment design. Since 2017, he has been involved in collaboration with an SME designing hardware and software solutions for cellular base stations.