Power modelling and optimization of future wireless transceivers

Over the last decades, communication networks have grown exponentially, starting with low-rate applications with GSM and now moving to high-speed mobile internet based on 5G deployment and future extensions towards 6G. While expanding networks and embracing new frequency bands helps meeting user throughput demands, this also increases the network power consumption, raising environmental, economic and battery lifetime concerns.

An interesting approach is to predict the power consumption of future devices before building them. By doing so for many possible architectures and configurations, we can select the most appropriate options, dimension them optimally, and propose ways to control their run-time operation to maximize their energy efficiency. Power modelling of wireless transceivers has been successfully investigated at IMEC for more than 10 years. However, considering the diversity of communication devices to design and enhance in future networks, many challenges lie ahead and require fundamental investigations. More specifically, you will investigate the following points in this PhD:

• New frequency bands are opened towards 6G, e.g., FR3 (7 - 24 GHz). This comes with new transceiver architectures and technologies, with a critical need to investigate the most energy-efficient solutions.
• New communication networks will also support radar-like sensing applications, based on Joint Communications and Sensing. This opens the door to interesting trade-offs by modelling the power consumption of hardware components supporting both sensing and communication features.
• Distributed architectures are also proposed. As compared to centralized base stations, they bring unsolved problems on the most energy-efficient strategy between local processing and data exchange to the central point.
• While high-level power modelling and optimization has been efficiently used for many years, the accuracy of this approach still deserves fundamental research. One aspect to improve is the benchmarking with state-of-the-art, extracting trends from power consumption data in recent designs. Even more crucial, modelling the digital power consumption based on operation count is known to be of limited accuracy. This high-level approach requires a stronger connection to the digital hardware to identify the best trade-off between model complexity and accuracy.
• Based on a refined system power consumption model, you will be able to assess advanced algorithms based on the benefits they bring vs. their power/complexity. For instance, Digital Pre-Distortion can enhance power amplifier efficiency but at the cost of extra digital complexity. Your model will determine the best solution.
• While traditionally based on understanding each sub-component, a system power model may also benefit from AI solutions, e.g., browsing through power consumption datasheets to extract relevant patterns and improve models and optimal solutions. Additionally, assessing neural network power consumption itself can complement the model for systems where traditional signal processing is complemented by machine learning.
• Finally, you will be able to put the energy efficiency at the transceiver hardware level in a broader perspective of sustainability and life cycle analysis, where run-time communication power consumption is only a part of the total picture.

To support this critical research, you will combine models from diverse sources and develop a flexible simulation environment allowing to model and optimize the power consumption of wireless transceivers. As a PhD student, you will be part of a large IMEC community working on the research, implementation and prototyping of future communications systems with experts in wireless communication, signal processing, digital, analog and mm-wave design, and machine learning. This is a unique opportunity to develop innovative, multi-disciplinary technology and shape future wireless networks. You will publish your research in top-level journals and conferences.