Digital predistortion (DPD) is one widely adopted approach to linearize radio transmitters, particularly the involved power amplifier (PA) units. Such methods help to push the PAs towards their saturation region, where power efficiency is good, while still providing clean transmit signal at antenna interface. Classical DPD solutions typically rely on high-order polynomial models, combined with filters, to obtain a very accurate estimate of the power amplifier-induced nonlinear distortion. Another important ingredient is the parameter learning path, which typically includes complex model extraction methods. In modern radio systems, such as the 5G New Radio (NR) deployments at millimeter-wave bands, transmitter linearization is impacted by the following new aspects. For one, transmitters build on active antenna arrays with multiple parallel PA units. For another, the transmit signal quality requirements, especially in terms of unwanted emissions at out-of-band region, have been relaxed, compared to legacy networks. Both of these aspects impact the DPD research, and are reflected in the new methods reported in the thesis.
Secondly, the so-called in-band full-duplex radio is a relatively new duplexing and transceiver technique, being currently under intensive study in the research community. Full-duplex aims at improving the flexibility and efficiency of the radio spectrum utilization, by allowing the devices to transmit and receive radio signals at the same time and over the same frequency band. The intrinsic problem of such full-duplex approach is the resulting self-interference, referring to the coupling of the powerful transmit signal to the coexisting and simultaneously operating receiver. Such self-interference signal needs to be efficiently modeled and cancelled, in real-time, to allow the receiver to demodulate and detect the actual incoming signals, while transmitting simultaneously at the same frequency. Creating accurate cancellation signals calls for sophisticated modeling of the self-interference and its dependence on the linear and nonlinear coupling responses between the RF transmitter and receiver. At the same time, the strict real-time requirements of the cancellation calls for computationally efficient processing and parameter estimation solutions. Developing and demonstrating such methods is the other main research topic of the thesis work.
Pablo Pascual Campo proposed several low-complexity algorithms and methods which can be used both in the context of digital predistortion and in-band full-duplex. These algorithms and methods comprise various underlying techniques – such as spline-based interpolation, sign algorithms, and low-complexity covariance matrix manipulation – in order to provide computationally efficient yet high-performance DPD and digital self-interference cancellation solutions. All the methods are also systematically validated through extensive RF measurement experiments, covering both sub-6GHz as well as 28GHz millimeter-wave examples.
The doctoral dissertation of M.Sc. (Tech) Pablo Pascual Campo in the field of Communications Engineering entitled Low-complexity Algorithms for Digital Predistortion and Self-interference Cancellation will be publicly examined as a virtual event on Friday 28.01.2022, at 12.00 noon. The Opponent will be Professor Daniel Rönnow from University of Gävle, Sweden. The Custos will be Professor Mikko Valkama from Tampere University, Tampere, Finland. The work has been co-supervised by Dr. Lauri Anttila from Tampere University.
The event can be followed via remote connection in MS Teams.
The dissertation is available online at https://urn.fi/URN:ISBN:978-952-03-2236-6