Les MOSFET en SiC sont appelées à remplacer les IGBT en Silicium pour des applications de demandant une plus forte vitesse de commutation. Cependant, les MOSFET en SiC ont encore quelques problèmes de fiabilité, tels que la robustesse de la diode interne ou bien la robustesse de l'oxyde de grille. Cette dernière est liée à l’oxyde de grille des composants du type MOSFET. Des instabilités de la tension de seuil sont aussi signalées. Cette thèse aborde ces deux sujets sur des MOSFET commerciaux 1200 V. L'étude de la diode interne met en évidence que les caractéristiques I-V (de la diode intrinsèque) demeurent stables après l'application d'un stress. Cependant, une dérive surprenante de la tension de seuil apparaît. Des tests complémentaires, en stressant le canal à la place de la diode, avec les mêmes contraintes n'ont pas montré de dérive significative de la tension de seuil. Donc, l'application d'un stress en courant quand le composant est en mode d'accumulation semble favoriser l'apparition des instabilités de la tension de seuil. La robustesse de l'oxyde de grille concerne les instabilités de la tension de seuil, mais aussi l'estimation de la durée de vie à des conditions d'opération nominales. Les résultats obtenus montrent que la durée de vie de l'oxyde de grille n'est plus un problème. Pourtant, le suivi du courant de grille pendant les tests ainsi que les caractérisations de la capacité de grille mettent en évidence des translations de la courbe C(V) à cause des phénomènes d’injection des porteurs et de piégeage, mais aussi la possible présence d’ions mobiles. Aussi, une bonne analyse des dégradations et dérives liées à l’oxyde de grille doit être réalisée. / SiC power MOSFETs are called to replace Si IGBT for some medium and high power applications (hundreds of kVA). However, even if crystallographic defects have been drastically reduced, SiC MOSFETs are always concerned by some robustness issues such as the internal diode robustness or the robustness of the gate oxide. The last one especially affects MOSFETs devices and is linked to the apparition of instabilities in the threshold voltage. This thesis focuses on these two issues. The study of the internal diode robustness highlighted that the I-V curve (of the intrinsic diode) remains stable after the application of a current stress in static mode, but also with the DUT placed in a converter with inductive switchings. These are the most stressful conditions. However, a surprising drift in the threshold voltage has been observed when some devices operates under these conditions; in static mode or in a converter. Complementary tests stressing the channel instead of the internal diode in the same temperature and dissipated power, have not resulted in a drift of the threshold voltage. Thus, the application of a current stress when the device is in accumulation regime could favour the apparition of instabilities in the threshold voltage. The study of the gate oxide focus in the instabilities of the threshold voltage, but also on the expected lifetime of the oxide at nominal conditions. Results obtained shown that the expected lifetime (TDDB) of the oxide is no longer a problem. Indeed, tests realized in static mode, but also in a converter under inductive switching conditions resulted in expected lifetimes well above 100 years. However, the monitoring of the gate current during the test and gate capacitance characterizations C(V) highlighted a shift in the capacitance due to carrier injection and trapping phenomena and probably to the presence of mobile-ions. Still regarding the instabilities of the threshold voltage, classic tests resulted in no significant variations of the threshold voltage at 150 _C. However, at 200 _C the drift observed for some manufacturers is higher than +30%. This is unacceptable for high-temperature applications and evidence that the quality of the gate oxide and the SiC=SiO2 interface must continue to be improved, together with the manufacturing methods to minimize the presence of mobile ions in the substrate.

The field of infrared (IR) detector technology has shown vast improvements in terms of speed and performance over the years. Specifically the dynamic range (DR) and sensitivity of detectors showed significant improvements. The most commonly used technique of implementing these IR detectors is the use of charge-coupled devices (CCD). Recent developments show that the newly investigated bipolar complementary metal-oxide semiconductor (BiCMOS) devices in the field of detector technology are capable of producing similar quality detectors at a fraction of the cost. Prototyping is usually performed on low-cost silicon wafers. The band gap energy of silicon is 1.17 eV, which is too large for an electron to be released when radiation is received in the IR band. This means that silicon is not a viable material for detection in the IR band. Germanium exhibits a band gap energy of 0.66 eV, which makes it a better material for IR detection. This research is aimed at improving DR and sensitivity in IR detectors. CCD technology has shown that it exhibits good DR and sensitivity in the IR band. CMOS technology exhibits a reduction in prototyping cost which, together with electronic design automation software, makes this an avenue for IR detector prototyping. The focus of this research is firstly on understanding the theory behind the functionality and performance of IR detectors. Secondly, associated with this, is determining whether the performance of IR detectors can be improved by using silicon germanium (SiGe) BiCMOS technology instead of the CCD technology most commonly used. The Simulation Program with Integrated Circuit Emphasis (SPICE) was used to realise the IR detector in software. Four detectors were designed and prototyped using the 0.35 µm SiGe BiCMOS technology from ams AG as part of the experimental verification of the formulated hypothesis. Two different pixel structures were used in the four detectors, which is the silicon-only p-i-n diodes commonly found in literature and diode-connected SiGe heterojunction bipolar transistors (HBTs). These two categories can be subdivided into two more categories, which are the single-pixel-single-amplifier detectors and the multiple-pixel-single-amplifier detector. These were needed to assess the noise performance of different topologies. Noise influences both the DR and sensitivity of the detector. The results show a unique shift of the detecting band typically seen for silicon detectors to the IR band, accomplished by using the doping feature of HBTs using germanium. The shift in detecting band is from a peak of 250 nm to 665 nm. The detector still accumulates radiation in the visible band, but a significant portion of the near-IR band is also detected. This can be attributed to the reduced band gap energy that silicon with doped germanium exhibits. This, however, is not the optimum structure for IR detection. Future work that can be done based on this work is that the pixel structure can be optimised to move the detecting band even more into the IR region, and not just partially. / Dissertation (MEng)--University of Pretoria, 2013. / Electrical, Electronic and Computer Engineering / unrestricted

A 5 GHz continuous unlicensed bandwidth is available at millimeter-wave (mm-wave) frequencies around 60 GHz and offers the prospect for multi gigabit wireless applications. The inherent atmospheric attenuation at 60 GHz due to oxygen absorption makes the frequency range ideal for short distance communication networks. For these mm-wave wireless networks, the low noise amplifier (LNA) is a critical subsystem determining the receiver performance i.e., the noise figure (NF) and receiver sensitivity. It however proves challenging to realise high performance mm-wave LNAs in a silicon (Si) complementary metal-oxide semiconductor (CMOS) technology. The mm-wave passive devices, specifically on-chip inductors, experience high propagation loss due to the conductivity of the Si substrate at mm-wave frequencies, degrading the performance of the LNA and subsequently the performance of the receiver architecture. The research is aimed at realising a high performance mm-wave LNA in a Si BiCMOS technology. The focal points are firstly, the fundamental understanding of the various forms of losses passive inductors experience and the techniques to address these issues, and secondly, whether the performance of mm-wave passive inductors can be improved by means of geometry optimising. An associated hypothesis is formulated, where the research outcome results in a preferred passive inductor and formulates an optimised passive inductor for mm-wave applications. The performance of the mm-wave inductor is evaluated using the quality factor (Q-factor) as a figure of merit. An increased inductor Q-factor translates to improved LNA input and output matching performance and contributes to the lowering of the LNA NF. The passive inductors are designed and simulated in a 2.5D electromagnetic (EM) simulator. The electrical characteristics of the passive structures are exported to a SPICE netlist which is included in a circuit simulator to evaluate and investigate the LNA performance. Two LNAs are designed and prototyped using the 13μ-m SiGe BiCMOS process from IBM as part of the experimental process to validate the hypothesis. One LNA implements the preferred inductor structures as a benchmark, while the second LNA, identical to the first, replaces one inductor with the optimised inductor. Experimental verification allows complete characterization of the passive inductors and the performance of the LNAs to prove the hypothesis. According to the author's knowledge, the slow-wave coplanar waveguide (S-CPW) achieves a higher Q-factor than microstrip and coplanar waveguide (CPW) transmission lines at mm-wave frequencies implemented for the 130 nm SiGe BiCMOS technology node. In literature, specific S-CPW transmission line geometry parameters have previously been investigated, but this work optimises the signal-to-ground spacing of the S-CPW transmission lines without changing the characteristic impedance of the lines. Optimising the S-CPW transmission line for 60 GHz increases the Q-factor from 38 to 50 in simulation, a 32 % improvement, and from 8 to 10 in measurements. Furthermore, replacing only one inductor in the output matching network of the LNA with the higher Q-factor inductor, improves the input and output matching performance of the LNA, resulting in a 5 dB input and output reflection coefficient improvement. Although a 5 dB improvement in matching performance is obtained, the resultant noise and gain performance show no significant improvement. The single stage LNAs achieve a simulated gain and NF of 13 dB and 5.3 dB respectively, and dissipate 6 mW from the 1.5 V supply. The LNA focused to attain high gain and a low NF, trading off linearity and as a result obtained poor 1 dB compression of -21.7 dBm. The LNA results are not state of the art but are comparable to SiGe BiCMOS LNAs presented in literature, achieving similar gain, NF and power dissipation figures. / Dissertation (MEng)--University of Pretoria, 2012. / Electrical, Electronic and Computer Engineering / unrestricted

Tato disertační práce se zabývá využitím programovatelných hradlových polí (FPGA) v diagnostice měničů, využívajících spínaných IGBT tranzistorů. Je zaměřena na budiče těchto výkonových tranzistorů a jejich struktury. Přechodné jevy veličin, jako jsou IG, VGE, VCE během procesu přepínání (zapnutí, vypnutí), mohou poukazovat na degradaci IGBT. Pro měření a monitorování těchto veličin byla navržena nová architektura budiče IGBT. Rychlé měření a monitorování během přepínacího děje vyžaduje vysokou vzorkovací frekvenci. Proto jsou navrhovány paralelní vysokorychlostní AD převodníky (> 50 MSPS). Práce je zaměřena převážně na návrh zařízení s FPGA včetně hardware a software. Byla navržena nová deska plošných spojů s FPGA, která plní požadované funkce, jako je řízení IGBT pomocí vícenásobných paralelních koncových stupňů, monitorování a diagnostiku, a propojení s řídicí jednotkou měniče.

Der thermische Widerstand ist eine wichtige Kenngröße von Silicium-Germanium-Hetero-Bipolartransistoren (SiGe-HBTs). Bisher kam es bei der quantitativen Bestimmung der thermischen Widerstände von SiGe-HBTs zu deutlichen Abweichungen zwischen Simulation und Messung. Der Unterschied zwischen Simulation und Messung betrug bei den untersuchten HBTs mehr als 30 Prozent. Diese Arbeit widmet sich der Aufklärung und Beseitigung der möglichen Ursachen hierfür. Zu diesem Zweck werden als erstes die Messmethoden analysiert. Es zeigt sich, dass die bisher verwendete Extraktionsmethode sensitiv auf den Early-Effekt (Basisweitenmodulation) reagiert. Im Rahmen der Untersuchungen wurde ein neues Extraktionsverfahren entwickelt. Die neue Extraktions­methode ist unempfindlich gegenüber dem Early-Effekt. Mit Bauelemente­simulationen wird erstmalig die Wirkung des Seebeck-Effektes (Thermospannungen) auf die elektrisch extrahierten thermischen Widerstände demonstriert. Der Seebeck-Effekt bewirkt, dass die elektrisch extrahierten thermischen Widerstände der untersuchten HBTs nahezu 10 Prozent kleiner als die erwarteten Werte sind. Dieser Effekt wurde bisher nicht beachtet und wird hier erstmals nachgewiesen. Weiterhin wird die Abhängigkeit des thermischen Widerstandes vom Arbeitspunkt untersucht. Dabei hat sich gezeigt, dass bis zu einer Basis-Emitter-Spannung von 0,91 Volt die geometrische Form des Wärme abgebenden Gebietes unabhängig vom Arbeitspunkt ist. Anhand von Messungen wird gezeigt, dass die Dotierung die spezifische Wärmeleitfähigkeit von Silicium reduziert. Die Abnahme wird für Dotierungen größer als 1*1019 cm‑3 deutlich sichtbar. Ist die Dotierung größer als 1*1020 cm‑3, beträgt die Abnahme der spezifischen Wärmeleitfähigkeit mehr als 75 Prozent. Mithilfe einer Simulatorkalibrierung wird die spezifische Wärmeleitfähigkeit als Funktion der Dotierung bestimmt. Die erhaltene Funktion kann künftig beim thermischen Entwurf von HBTs verwendet werden. Somit können zukünftig genauere Vorhersagen zum thermischen Widerstand der HBTs gemacht werden. Dies ermöglicht zuverlässigere Aussagen darüber, wie Änderungen des Transistordesigns zur Minimierung des thermischen Widerstandes beitragen. / The thermal resistance is an important parameter of silicon-germanium heterojunction bipolar transistors (SiGe HBTs). Until now, the quantitative determination of the thermal resistance showed significant differences between measurements and simulations. The difference between simulation and measurement of the investigated HBTs was more than 30 percent. This thesis devotes the clarification and elimination of potential sources for it. For this purpose, the measurement methods are analyzed at first. It is shown, that the currently used extraction method is sensitive to the Early effect (basewidth modulation). A now extraction method was developed, which is not sensitive to the Early effect. For the first time, the influence of the Seebeck effect (thermoelectric voltages) on the electrically extracted thermal resistance is shown by device simulations. The Seebeck effect leads to a 10 percent lower extracted thermal resistances compared to the expected values of the investigated HBTs. This effect was not taken into account up to now and is demonstrated here for the first time. Furthermore, the dependence of the thermal resistance on the operating point was investigated. The results show that the shape of the heat source is independent of the operating point if the base emitter voltage is smaller than 0.91 volt. The thermal conductivity of silicon is decreased by increasing doping concentrations. This is shown by measurements. The reduction of the thermal conductivity is well observable for doping concentrations higher than 1*1019 cm‑3. For doping concentration higher than 1*1020 cm‑3 the reduction amounts to more than 75 percent. The thermal conductivity was determined as a function of the doping concentration with the aid of a simulator calibration. This function can be used in the future thermal design of HBTs. It facilitates the optimization of the HBTs with respect to a minimal thermal resistance.

Les procédés bipolaires semi-conducteurs complémentaires à oxyde de métal (BiCMOS) peuvent être considérés comme étant la solution la plus généralepour les produits RF car ils combinent la fabrication sophistiquée du CMOSavec la vitesse et les capacités de conduction des transistors bipolaires silicium germanium(SiGe) à hétérojonction (HBT). Les HBTs, réciproquement, sontles principaux concurrents pour combler partiellement l'écart de térahertzqui décrit la plage dans laquelle les fréquences générées par les transistors etles lasers ne se chevauchent pas (environ 0.3 THz à 30 THz). A_n d'évaluerles capacités de ces dispositifs futurs, une méthodologie de prévision fiable estsouhaitable. L'utilisation d'un ensemble hétérogène d'outils et de méthodes desimulations permet d'atteindre successivement cet objectif et est avantageusepour la résolution des problèmes. Plusieurs domaines scientifiques sont combinés, tel que la technologie de conception assistée par ordinateur (TCAO),la modélisation compacte et l'extraction des paramètres.Afin de créer une base pour l'environnement de simulation et d'améliorerla confirmabilité pour les lecteurs, les modèles de matériaux utilisés pour lesapproches hydrodynamiques et de diffusion par conduction sont introduits dèsle début de la thèse. Les modèles physiques sont principalement fondés surdes données de la littérature basées sur simulations Monte Carlo (MC) ou dessimulations déterministes de l'équation de transport de Boltzmann (BTE).Néanmoins, le module de TCAO doit être aussi étalonné sur les données demesure pour une prévision fiable des performances des HBTs. L'approchecorrespondante d'étalonnage est basée sur les mesures d'une technologie depointe de HBT SiGe pour laquelle un ensemble de paramètres spécifiques àla technologie du modèle compact HICUM/L2 est extrait pour les versionsdu transistor à haute vitesse, moyenne et haute tension. En s'aidant de cesrésultats, les caractéristiques du transistor unidimensionnel qui sont généréesservent de référence pour le profil de dopage et l'étalonnage du modèle. Enélaborant des comparaisons entre les données de références basées sur les mesureset les simulations, la thèse fait progresser l'état actuel des prévisionsbasées sur la technologie CAO et démontre la faisabilité de l'approche.Enfin, une technologie future de 28nm performante est prédite en appliquantla méthodologie hétérogène. Sur la base des résultats de TCAO, leslimites de la technologie sont soulignées. / Bipolar complementary metal-oxide-semiconductor (BiCMOS) processescan be considered as the most general solution for RF products, as theycombine the mature manufacturing tools of CMOS with the speed and drivecapabilities of silicon-germanium (SiGe) heterojunction bipolar transistors(HBTs). HBTs in turn are major contenders for partially filling the terahertzgap, which describes the range in which the frequencies generated bytransistors and lasers do not overlap (approximately 0.3THz to 30 THz). Toevaluate the capabilities of such future devices, a reliable prediction methodologyis desirable. Using a heterogeneous set of simulation tools and approachesallows to achieve this goal successively and is beneficial for troubleshooting.Various scientific fields are combined, such as technology computer-aided design(TCAD), compact modeling and parameter extraction.To create a foundation for the simulation environment and to ensure reproducibility,the used material models of the hydrodynamic and drift-diffusionapproaches are introduced in the beginning of this thesis. The physical modelsare mainly based on literature data of Monte Carlo (MC) or deterministicsimulations of the Boltzmann transport equation (BTE). However, the TCADdeck must be calibrated on measurement data too for a reliable performanceprediction of HBTs. The corresponding calibration approach is based onmeasurements of an advanced SiGe HBT technology for which a technology specific parameter set of the HICUM/L2 compact model is extracted for thehigh-speed, medium-voltage and high-voltage transistor versions. With thehelp of the results, one-dimensional transistor characteristics are generatedthat serve as reference for the doping profile and model calibration. By performingelaborate comparisons between measurement-based reference dataand simulations, the thesis advances the state-of-the-art of TCAD-based predictionsand proofs the feasibility of the approach.Finally, the performance of a future technology in 28nm is predicted byapplying the heterogeneous methodology. On the basis of the TCAD results,bottlenecks of the technology are identified. / Bipolare komplementäre Metall-Oxid-Halbleiter (BiCMOS) Prozesse bietenhervorragende Rahmenbedingungen um Hochfrequenzanwendungen zurealisieren, da sie die fortschrittliche Fertigungstechnik von CMOS mit derGeschwindigkeit und Treiberleistung von Silizium-Germanium (SiGe) Heterostruktur-Bipolartransistoren (HBTs) verknüpfen. Zudem sind HBTs bedeutendeWettbewerber für die teilweise Überbrückung der Terahertz-Lücke, derFrequenzbereich zwischen Transistoren (< 0.3 THz) und Lasern (> 30 THz).Um die Leistungsfähigkeit solcher zukünftigen Bauelemente zu bewerten, isteine zuverlässige Methodologie zur Vorhersage notwendig. Die Verwendungeiner heterogenen Zusammenstellung von Simulationstools und Lösungsansätzenerlaubt es dieses Ziel schrittweise zu erreichen und erleichtert die Fehler-_ndung. Verschiedene wissenschaftliche Bereiche werden kombiniert, wie zumBeispiel der rechnergestützte Entwurf für Technologie (TCAD), die Kompaktmodellierungund Parameterextraktion.Die verwendeten Modelle des hydrodynamischen Simulationsansatzes werdenzu Beginn der Arbeit vorgestellt, um die Simulationseinstellung zu erläuternund somit die Nachvollziehbarkeit für den Leser zu verbessern. Die physikalischenModelle basieren hauptsächlich auf Literaturdaten von Monte Carlo(MC) oder deterministischen Simulationen der Boltzmann-Transportgleichung(BTE). Für eine zuverlässige Vorhersage der Eigenschaften von HBTs muss dieTCAD Kon_guration jedoch zusätzlich auf der Grundlage von Messdaten kalibriertwerden. Der zugehörige Ansatz zur Kalibrierung beruht auf Messungeneiner fortschrittlichen SiGe HBT Technologie, für welche ein technologiespezifischer HICUM/L2 Parametersatz für die high-speed, medium-voltage undhigh-voltage Transistoren extrahiert wird. Mit diesen Ergebnissen werden eindimensionaleTransistorcharakteristiken generiert, die als Referenzdaten fürdie Kalibrierung von Dotierungspro_len und physikalischer Modelle genutztwerden. Der ausführliche Vergleich dieser Referenz- und Messdaten mit Simulationengeht über den Stand der Technik TCAD-basierender Vorhersagenhinaus und weist die Machbarkeit des heterogenen Ansatzes nach.Schlieÿlich wird die Leistungsfähigkeit einer zukünftigen Technologie in28nm unter Anwendung der heterogenen Methodik vorhergesagt. Anhand derTCAD Ergebnisse wird auf Engpässe der Technologie hingewiesen.