Energy efficiency in signal processing hardware is vital for several reasons, in particular to reduce the environmental impact, to increase stand-by times, to make batteries smaller and lighter, and to simplify the cooling of circuits. The need for reducing the energy consumption is also accentuated by the trend of adding more and more functionality into electronic devices, which means that each functionality has to become more energy efficient if not the total energy consumption shall increase.
At the signal processing level, the energy efficiency of the hardware can be increased in basically two ways. The first one is to reduce the computational complexity (number of arithmetic operations per time unit) of the DSP algorithms. The second way is to use low-precision computations and hardware, which however also means more signal distortion. In this project, we work both on developing computationally efficient agile DSP algorithms, that can adapt to meet the system requirements with as low energy consumption as possible, and on developing DSP algorithms that can handle the increased signal distortion from energy-efficient low-precision hardware.
The first part of the project is to develop efficient agile DSP algorithms that can adapt to the changes in functionality and quality requirements in communication systems with time-varying operation modes. An example is a filter with adaptable bandwidths and center frequencies (functionality) as well as adaptable attenuations and word-lengths (quality requirements). In this way, one can meet different system requirements with minimum computational complexity and energy consumption in the corresponding hardware implementation. In addition, it is important that the adaptation is fast to enable real-time operation. The DSP algorithms in focus here are found in the physical layer of transceivers, like sampling rate conversion, resampling, channelization, carrier aggregation, multiplexing/transmultiplexing, frequency-band reallocation, spectrum sensing, channel equalization, multi-channel signal reconstruction, and channel and parameter estimation.
One important wireless communication technology, where energy-efficient hardware becomes indispensable, is massive MIMO. In massive MIMO, the base stations have to be equipped with hundreds of individually controllable radio chains, and each chain has to include all necessary hardware to modulate and demodulate radio-frequency signals. The overall energy consumption of the hardware would be enormous without energy-efficient solutions.
The second part of the project works on developing DSP algorithms for massive MIMO, that will make it possible to use energy-efficient hardware in base stations despite high signal distortion. The impact that the signal distortion has on the end communication performance is characterized in this part. The work is particularly focused on enabling the use of analog-to-digital converters with low resolutions, low-end power amplifiers that are allowed to operate close to saturation, and mitigation of phase noise in local oscillators. Only by enabling the use of such highly energy-efficient hardware is it possible to build highly complex massive MIMO system in a feasible and environmentally sustainable way.