Software-defined Bluetooth Low Energy transceiver for reconfigurable IoT smart sensors

Technology is more and more present around us. In particular, the Internet of Things (IoT) has been emerging for a few years and is now in huge expansion. Along with it comes a massive deployment of smart sensors. This is full of promise for a lot of domains (social, economical, etc.). However, it raises some sustainability concerns, particularly from an environmental point of view. Firstly, this huge deployment puts a lot of pressure on the natural resources. Secondly, the ecotoxicity of the sensors battery is an issue, especially as their lifetime is very limited. This short lifetime comes from the fact that the wireless communication standards that these sensors use are constantly evolving, but due to their lack of reconfigurability the sensors cannot keep up with this evolution. A possible way towards sustainability is the use of a software-defined radio (SDR) for the wireless communication of those smart sensors. An SDR has the advantage of being reconfigurable through over-the-air update, and of being able to support multiple communication standards, which makes it versatile for a large variety of IoT applications. The big challenge about SDRs is their power consumption, which should remain as low as possible from the sustainability point of view. With this in mind, this thesis investigates the low-power implementation of a software-defined transceiver following the Bluetooth Low Energy (BLE) communication standard. BLE is the fourth generation of Bluetooth and has been designed to meet the requirements of the IoT applications. This thesis proposes a receiver implementing non-coherent differential detection, which offers a very good tradeoff between low-complexity and performance. Moreover, the receiver mitigates non-idealities such as symbol timing offset and carrier frequency offset with very low-cost algorithms, based on an IIR filter and a Taylor series approximation respectively. This design allows to stay under the 0.1-% bound of bit error rate at a 13-dB signal to noise ratio (SNR) with a probability of 99.3 %, and with a probability of 99.9 % at a 14-dB SNR.