Abstract
In recent decades, the microelectronics industry has experienced a wide democratization of the use of telecommunication applications. The improved process design and manufacturing have produced complex and high performance analog, mixed and radio frequency circuits for these applications. However, the test cost of these integrated circuits still represents a large part of the manufacturing cost. Indeed, very often, analog testing is not just a functional test but needs measurements for specification validations. These measurements require the use of dedicated instruments expensive resources on standard industrial test equipment.One of the essential but costly specifications to validate in RF circuitry is the phase noise level. The currently used industrial technique consists in capturing the signal from the circuit under test using an RF tester channel equipped with a high performance analog to digital converter; a Fourier transform is then applied to the digitized signal and the phase noise is measured on the resulting spectrum.The approach proposed in this thesis is to achieve the phase noise measurement using solely digital low-cost resources. The basic idea is to perform 1-bit capture of the analog signal with a standard digital channel and develop post-processing algorithms dedicated for phase noise evaluation from the zero-crossings of the signal.Two methods are presented. The first method is based on an estimate of the instantaneous signal frequency and an analysis of their dispersion induced by phase noise. This method imposes a strong constraint on the sampling frequency to be used and proved to be sensitive to noise amplitude, limiting the range of possible measures. A second method is then proposed to overcome these limitations. From the binary capture of the analog signal, a reconstruction of the instantaneous phase of the signal is carried out, then filtered and characterized by a common tool of frequency stability assessment: the Allan variance. This technique, robust to amplitude noise and jitter, can be parametrized and enables efficient characterization of phase noise without fundamental constraint.In addition to the simulations, these techniques are subject to a stochastic study and are validated experimentally on different types of signals to be measured - artificially generated or from chips on the market - and with different measuring instruments - on oscilloscope or industrial tester, in laboratory and on a production line-. An On-chip implementation is also proposed and validated with a FPGA prototype.