Thibaut MURET

Abstract: The exponential growth of the Internet of Things (IoT) is accompanied by a critical increase in the need for energy efficiency, particularly in the analog RF circuits used for wireless communications. However, the energy benefits traditionally associated with the downscaling of CMOS components, according to Moore's Law, are reaching their limits beyond the 22 nm technology node. This stagnation is mainly due to the reduction in supply voltages (around 0.6 to 0.7 V), the increasing impact of parasitic capacitances, and the stagnation in gate width reductions (around 12 to 16 nm), thereby preventing a substantial decrease in the energy consumption of analog RF circuits.
In this context, it becomes imperative to explore new disruptive architectural approaches, different from traditional LNA-First architectures. The approach considered in this research project specifically relies on two promising technological elements: passive N-Path structures and low-noise analog-to-digital converters (LNADCs).
The Mixer-First architecture based on N-Path filters reduces consumption by eliminating the LNA stage. N-Path filters replace traditional inductors with banks of switched capacitors, providing programmable RF filtering around the clock frequency with a very high quality factor (Q), without static consumption or bulky inductors. Combined with advanced techniques such as implicit capacitive stacking, these structures can also provide significant passive gain, ranging from 13 to 18 dB. This high passive gain significantly reduces the need for downstream active amplification.
LNADCs, on the other hand, enable early digitization of the RF signal by avoiding traditional linear analog amplifications, which are particularly energy-intensive at high frequencies. These converters now achieve a Schreier FoM close to the fundamental thermodynamic limit (approximately 185 dB), indicating that they are extremely close to the minimal energy floor set by thermal noise.