Simulation physique des sources de bruit basse fréquence, Cyclostationnaire, dans les composants semi-conducteurs: Modélisation compacte, pour la CAO des circuits non linéaires. – CYCLOMOD

INTRODUCTION: Communications systems require complex modulation formats such as 256 QAM modulation format, and faster symbol rate. Today, in order to fulfil these characteristics, equipment manufacturers currently require monolithic VCOs whose main features are: - A centre frequency between 2 GHz and 20 GHz. - Output power: 10dBm - Tuning bandwidth: 11% - Phase noise spectrum in the whole bandwidth -95 dBc @10 KHz of the carrier -115 dBc @100 KHz of the carrier The only monolithic devices currently allowing to obtain such joint features (Tuning bandwidth ' Phase noise spectrum) are the HBT transistors and varactors manufactured in InGaP-GaAs technology. New improvements are yet required for future systems. For example a phase noise improvement by over 10dB (improvement of one order of magnitude!) for identical tuning bandwidths will be needed: - -105 dBc @10 KHz of the carrier - -125 dBc @100 KHz of the carrier On the other hand, the centre frequencies range will be widened far beyond 20 GHz. At present, monolithic VCO circuit designers are using compact models of semi-conductor devices allowing to foresee rather accurately the deterministic characteristics (Power, frequency, Tuning bandwidth) of VCO oscillators. Nevertheless Cyclostationary low frequency noise sources are the main sources accounting for phase noise characteristics. However cyclostationary low frequency noise modeling is currently done in a totally empirical, uncertain way, which results in very uncertain predictive phase noise: This project aims at developing a compact predictive model based on a theoretical study of low frequency noise sources of diodes and bipolar transistors under large signal operation. Our ambition is to end at a technological breakthrough in the field of compact noise predictive modelling of semiconductor devices, for non linear CAD of RF microwaves and millimetre waves circuits. The proposed modelling, currently non-existent as far as we know, is based on physical simulations and measurements which results will be continuously compared during the project. The final results will enable us: - To definitely improve the accuracy of monolithic InGaP-GaAs VCOs predictive simulation in terms of phase noise; - On the other hand, to improve by one order of magnitude the phase noise of monolithic VCO oscillators used in telecommunications. - Finally, to improve, by retroaction, the devices (diodes and transistors) to be used in VCOs for systems which will be designed in the forthcoming decade. PROJECT PLANNING : - UMS, who is not partner in order to alleviate the financial budget, will provide test vehicles: TLM resistors, homojunction and heterojunction diodes, and also HBT transistors in InGaP-GaAs technology) and oscillator circuits (VCOs) - Two tasks will then be carried out in parallel: - Physically based simulation of the semiconductor devices under ATLAS software, in order to extract deterministic distributed electrical model, nonlinear of HBTs (Alcatel-Thalès III-V Lab). - Deterministic electrical compact modeling of the simulated devices (XLIM Lab). This modelling will be achieved from physically based simulation performed previously and pulsed S parameters measurement under a specific measurement set-up developed at XLIM. -Measurements of input/output stationary low frequency equivalent current noise sources at the accesses of the devices(XLIM Lab). - Physically-based noise simulation of the semiconductor devices will be undertaken under ATLAS software in order to fit the static noise sources measured on the test vehicles (III-V Lab, XLIM Lab) - A formalism will then be developed, in order to determine a relation between low frequency noise sources under DC and under large signal operating conditions (XLIM Lab). This formalism should enable us to extract the cyclostationarity function of low frequency equivalent current noise sources at the accesses of the devices. - The cyclostationary low frequency noise compact models will then be extracted and introduced into deterministic electrical compact models previously developed. - These whole compact models will be implemented in a nonlinear circuit ADS simulator (XLIM Lab). - Dynamic low frequency noise measurements under RF large signal operation will be done with the noise measurement set-up whose capabilities have been extended to this aim (XLIM Lab). Comparisons will then be made between the results obtained: - On the one hand with the ADS simulator software. - On the other hand on the low frequency noise experimental set-up under large signal operation. - A critical discussion will conclude the project. The methodology developed in order to extract cyclostationary compact noise models will be made available to the scientific and industrial communities.