The increasing demand for higher power device utilization and reliability in power electronic systems is driving the integration of condition monitoring and active control in power electronic systems. With thermal heat dissipation being the limiting factor of module lifetime and performance, thermal real-time monitoring is key in this transition. However, this requires the availability of highly accurate and high-bandwidth temperature information with minimal phase lag. This paper reviews key methods for temperature extraction in power modules: temperature sensing, thermal estimators and thermal observers. While previous research has examined individual techniques of these methods in great detail, this paper presents a methodological overview that provides insight to how different technologies may contribute to next generation thermal monitoring solutions. In this context, the paper discusses how different technologies can be effectively combined to improve their overall performance. It finally discusses key challenges that must be addressed such that minimally invasive temperature sensing and thermal monitoring become an industrial practice in the future of power electronics.

To optimize the lifetime of switching power semiconductors, this paper presents a methodology to control power device junction temperature T _j and its change during power cycles ΔT _j at thermal boundaries. This paper proposes a supervisory state machine to interrupt nominal system-level control only when temperature bounds are exceeded, and coordinates smooth transitions as T _j (k) and ΔT _j (k) approach their respective boundaries. To ensure that thermal states are regulated via precise and independent modulation of conduction and switching loss elements, decoupling methods are proposed. Also proposed is a ΔT _j control law that closes a control loop on the rate of change state Ṫ _j , and introduces active thermal capacitance and conductance into the closed-loop thermal system dynamics. Experimental evaluation of the proposed system illustrates well damped T _j (k) and ΔT _j (k) responses, and gradual adjustment of the manipulated inputs switching frequency and duty ratio. Finally, comparison with a current limit-based T _j regulation method illustrates how the proposed system allows power converters to push harder against their thermal limits.

This paper introduces new technologies for monitoring thermal response and life-time varying parameters of power electronic modules in real-time without interrupting normal converter operation. The technologies are embedded in an adaptive thermal observer that integrates temperature measurements and electrothermal models to estimates device losses and temperatures at reliability-critical locations of the power module. The observer structure includes a model reference adaptive system (MRAS) for adaptively calibrating device switching-loss models. This improves thermal monitoring accuracy, avoids time-consuming off-line calibrations and creates the opportunity for efficiency estimations during power module operation. The adaptive observer together with small-signal loss-excitation and system-identification technologies extract the thermal impedance of the power module in magnitude and phase over multiple frequency decades. This in-situ thermal impedance spectroscopy (ITIS) has the unique feature of continuously providing information on the thermal interface of the power devices without interrupting normal inverter operation. The extracted information, i.e. life-time varying electrical and thermal parameters, enable diagnosing power module state-of-health with improved accuracy compared to prior technologies. It even allows separating different degradation effects, such as bond wire lift-off, device and base-plate solder delamination as well as deterioration of the convection process. With these new features, next generation monitoring systems can extract thermal response and degradation and thus protect power modules from severe thermal stress and overload. Furthermore, the obtained state-of-health information can be selectively used to reduce thermal stress and to schedule predictive maintenance leading to a safer and more reliable operation of future converter systems.

This paper develops impedance spectroscopy for characterizing transient heat transfer in assembled power semiconductor converter systems. The method addresses limits associated with traditional methods, e.g. data sheet transient thermal impedance curves are step responses only characterizing the junction-to-case spatial domain segment. The presented method utilizes an actively switching semiconductor as a controlled modulator of dynamic loss (heat) injections, along with multiple temperature measurements, to form spatial electrothermal impedance frequency response function (FRF) models. Complementary investigations of numerical and analytical modeling methods yield a general, compact thermal model template, for interpreting measured FRFs. A pulse-width modulated power converter with integrated die and heat sink temperature sensing was characterized with the spectroscopy method. Obtained ambient-referred and relative FRFs describe transient heat transfer, including fast (die-level) and slow (sink-level) modes, over frequency decades with coherence. The FRFs were parameterized and interpreted using the derived compact thermal model.

This paper develops techniques for actively tracking response shifts, induced by thermal-mechanical degradation over components lifetimes’, in spatiotemporal thermal response dynamics. The methodology considers both boundary (e.g. cooling) and internal (e.g. voiding) sources of degradation, and it is considered in the context of power semiconductor devices and packages, including power modules. The paper quantifies transient thermal response sensitivity to degradation using electrothermal impedances viewed with key frequency response function (FRF) metrics over wide dynamic ranges. A developed sensitivity analysis is applied using models and experimental evaluation to relate temperature sensing spatial location, harmonic content of semiconductor device losses, and source of degradation in terms of normalized FRFs that are rapidly interpreted. Complementary loss model parameter sensitivity analysis reveals advantages in tracking variation in thermal FRF phase delay, rather than amplitude. Last, to complement pure-FRF approaches for sensing degradation, which use non-parametric data, a method for directly estimating degradation-sensitive physical parameters, in real-time, is developed. Experiments demonstrate automatic estimation of a thermal resistance parameter.

This paper presents a method for designing converter systems embedding power modules for a specified degradation monitoring performance in real-time applications. It builds upon prior work which identified specific variations in three-dimensional transient heat transfer with interface and interconnect state-of-health. This paper contributes a joint, module-converter concurrent design method to quantify opportunities and limits in using chip-based temperature sensing to integrate electrothermal impedance spectroscopy in situ. Challenges intrinsic to modules, such as many, closely packed heat sources, are quantified. The paper also examines spatial degrees-of-freedom for placing additional degradation sensing temperature detectors. Experimental measurements of transient thermal frequency response show trends that can be exploited to push temperature sensors against their signal-to-noise ratio limits. Overall, the design method relies on circuit topology and modulation details; it is directed toward early-stage converter development.To that end, an included harmonic loss modeling step immediately provides a designer with feedback about the feasibility of a degradation sensing concept.