Skip to main content


Self-learning self-test and self-calibration for integrated millimeter-wave systems


Keywords: Radiofrequency, Integrated Systems, Machine learning

Abstract: Applications of radio frequency circuits (RF) operating at frequencies above 100GHz have been explored at R&D level for about ten years. Application scenarios include automotive radars, 5G telecommunications circuits (and the 6G future at maturity). 2030), monitoring (in particular for medical applications), etc. The advantage for these applications of operating at very high frequencies concerns the spatial resolution for radar and monitoring applications, this resolution being linked to the frequency bandwidth of the systems. This interest is linked to the speed in GBits / s for telecommunications. The difficulty of designing circuits at frequencies higher than a hundred GHz is important and currently constitutes a real challenge, because on the one hand the electric models of transistors, diodes and varactors are much less reliable than at lower frequencies where they have already been validated, and on the other hand the drift in circuit performance is more sensitive to technological inaccuracies due to dimensions which decrease when the frequency increases. The challenge is therefore important for the development of reliable circuits beyond 100GHz. In this scenario, the production yield is low and very costly test and calibration procedures are necessary to guarantee the functionality of the circuit. The proposed thesis work wishes to answer the problems set out above. With this in mind, we propose to exploit the power of advanced machine learning algorithms to simplify the testing and calibration of a complex integrated system in millimeter range. In particular, our objective is to demonstrate self-calibration and self-test techniques on circuits located at the heart of RF systems, namely an innovative “Low Noise Amplifier” / LNA system. »For receiving signals at a frequency between 120GHz and 140GHz, with self-calibration and self-test in-situ. Test and calibration algorithms will be developed on the basis of characteristics selection techniques developed in the context of the Exploratory Project of the LabEx PersyvalCAFE & Test (2016-17, piloted by Manuel J. Barragan) which will allow us to identify the causes The physical demonstrator with the complete system will be integrated in 55 nm BiCMOS technology made available by our partner STMicroelectronics in the context of lasting collaborations.


Thesis director: Philippe FERRARI (TIMA - RMS)
Thesis started on: Oct. 2020
Thesis defence : 12/02/2024
Doctoral school: EEATS

Submitted on January 12, 2022

Updated on April 8, 2024