Since they appeared, wireless technologies have deeply transformed our society. Today, wireless internet access and other wireless applications demandincreasingly more traffic. However, the continuous traffic increase can be unbearableand requires rethinking and redesigning the wireless technologies inmany different aspects. Aiming to respond to the increasing needs of wirelesstraffic, we are witnessing a rapidly evolving wireless technology scenario.This thesis addresses various aspects of the transmitters used in wireless communications.Transmitters present several hardware (HW) impairments thatcreate distortions, polluting the radio spectrum and decreasing the achievabletraffic in the network. Digital platforms are now flexible, robust and cheapenough to enable compensation of HW impairments at the digital base-bandsignal. This has been coined as ’dirty radio’. Dirty radio is expected in future transmitters where HW impairments may arise to reduce transmitter cost or to enhance power efficiency. This thesis covers the software (SW) compensation schemes of dirty radio developed for wireless transmitters. As describedin the thesis, these schemes can be further enhanced with knowledge of thespecific signal transmission or scenarios, e.g., developing cognitive digital compensationschemes. This can be valuable in today’s rapidly evolving scenarioswhere multiple signals may co-exist, sharing the resources at the same radiofrequency (RF) front-end. In the first part, this thesis focuses on the instrumentation challenges andHWimpairments encountered at the transmitter. A synthetic instrument (SI)that performs network analysis is designed to suit the instrumentation needs.Furthermore, how to perform nonlinear network analysis using the developedinstrument is discussed. Two transmitter HW impairments are studied: themeasurement noise and the load impedance mismatch at the transmitter, asis their coupling with the state-of-the-art digital compensation techniques.These two studied impairments are inherent to measurement systems and areexpected in future wireless transmitters. In the second part, the thesis surveys the area of behavioral modeling and digital compensation techniques for wireless transmitters. Emphasis is placed on low computational complexity techniques. The low complexity is motivated by a predicted increase in the number of transmitters deployed in the network, from base stations (BS), access points and hand-held devices. A modeling methodology is developed that allows modeling transmitters to achieve both reduced computational complexity and low modeling error. Finally, the thesis discusses the emerging architectures of multi-channel transmittersand describes their digital compensation techniques. It revises the MIMOVolterra series formulation to address the general modeling problem anddrafts possible solutions to tackle its dimensionality. In the framework of multi-channel transmitters, a technique to compensate nonlinear multi-carrier satellite transponders is presented. This technique is cognitive because it uses the frequency link planning and the pulse-shaping filters of the individual carriers. This technique shows enhanced compensation ability at reduced computational complexity compared to the state-of-the-art techniques and enables the efficient operation of satellite transponders.

Increasing demands for data rate, energy efficiency and reliability in wireless communications have resulted in the introduction of radio frequency (RF) multiple input multiple output (MIMO) transmitters. However, MIMO transmitters suffer from additional crosstalk impairments along with the power amplifier (PA) and I/Q imbalance distortions observed in single input single output (SISO) transmitters. Therefore, this thesis focuses on the characterization and compensation of these hardware impairments in RF SISO and MIMO transmitters.

PA distortions are often compensated using the Volterra series, but it suffers from high computational complexity. Therefore, a non-parametric method based on density estimation has been proposed in this thesis to estimate the PA transfer function, from which pruned Volterra models can be developed. The method is validated for a Doherty PA and achieves competitive error performance at a lower complexity than its competitors.

For MIMO transmitters, a characterization technique that uses multitone excitation signals has been proposed. Multitone signals yield non-overlapping tones at the outputs of the MIMO Volterra kernels. These kernel outputs are used to identify the dominant crosstalk impairments, from which block structure and base-band behavioral models are developed. The method is validated for 2x2 and 3x3 MIMO transmitters and it is shown that the derived models achieve a better complexity accuracy trade-off than the other pruned MIMO Volterra models considered in this thesis.

Finally, the thesis presents compensation models for joint static I/Q imbalance and MIMO PA distortions based on conjugate pair and real-valued basis functions. The models are augmented with sub-sample resolution to compensate for dynamic I/Q imbalance distortions. The proposed models are validated for a 2x2 RF MIMO transmitter and achieve a better complexity accuracy trade-off than the other state-of-the-art models considered in this thesis.