In the realm of power electronics, the need for enhanced performance and efficiency is becoming increasingly paramount, driven by the ubiquitous presence of power systems in modern technology. A recent presentation at the PCIM Europe 2024 has highlighted a novel methodology focusing on the virtual optimization of power electronic systems, which relies heavily on the creation of accurate semiconductor models. These models are essential for effective simulations of switching behaviour, current sharing, and over-voltage characteristics of power electronic converters.
The significance of power electronics spans various sectors, including consumer electronics, renewable energy, and transportation systems. An integral element influencing the overall efficiency of these systems is the power semiconductor's capability, which is crucial for determining electrical characteristics, managing thermal aspects, and simplifying control complexity.
With the advent of virtual design methodologies, engineers now have opportunities to optimise power modules and complete converter systems without the necessity of building physical prototypes. Yet, this process is contingent on the precision of semiconductor models that must encompass a broad range of operational variables, such as dynamic and static current sharing among parallel semiconductor dies as well as over-voltage scenarios under different operating conditions.
Despite the role of simulations in modern development, existing design workflows often present a myriad of challenges due to a lack of seamless integration between the characterisation of semiconductor dies and models for power modules. Conventional virtual design workflows lean heavily on transient electrical simulation to replicate semiconductor switching behaviour accurately, aiding in both electrical and thermal optimisation. However, achieving these simulations hinges on the availability of highly accurate semiconductor models. Existing vendor models often have limitations, as they must be validated for each unique design, and the inherent complexity of semiconductor devices means that simplified models may lead to significant inaccuracies.
Collaboration between simulation tool providers, such as Keysight Technologies, has initiated advancements in methodologies that enhance model accuracy. A study conducted by Stefan Haensel and colleagues from Siemens and Keysight Technologies revealed the extraction and validation process for IGBT (Insulated Gate Bipolar Transistor) and diode models using Keysight's tools, yielding promising results without prior knowledge of the devices’ internal architecture.
The extraction process itself necessitates detailed data that is seldom found in standard device datasheets. Key parameters include the IGBT's transfer characteristic, collector-emitter voltage-dependent input/output capacitance, and reverse transfer capacitance—alongside similar parameters for diodes, such as output characteristics and the voltage-dependent capacitance. Accounting for temperature variations is another critical aspect, as semiconductor performance can significantly fluctuate with thermal changes.
Through the use of a power device analyser, the team gathered essential data for model fitting. Both static and dynamic measurements were employed, utilising double-pulse tests to capture transient behaviours like reverse recovery in diodes and the switching characteristics of IGBTs. Their research produced promising alignments between the measured data and the models, indicating satisfactory model accuracy across a variety of operating points.
The evaluation of model accuracy is based on several criteria, focusing not only on switching losses but also on other critical parameters, such as over-voltage conditions and parasitic elements, which can impact overall device behaviour. The study articulated thirteen distinct criteria that assess the transient simulation accuracy of IGBTs and diodes.
Findings from the study revealed that the models maintained an error margin for switching losses generally within ±7% for high-power applications and ±25% for low-power conditions. However, discrepancies were noted in the di/dt error across voltage ranges, showcasing the complexities of modelling transient behaviours in extreme conditions.
As a conclusion, this advancement in semiconductor modelling contributes to the future of power electronic system design, allowing for comprehensive comparisons between various module designs and conditions. While current models have shown promise, there is an acknowledgement of the need for continuous refinement—particularly concerning switching losses, their dynamic behaviour, and adaptability for gate driver configurations.
Looking to the horizon, the expectation remains that semiconductor manufacturers will deliver more detailed, flexible models that integrate seamlessly into existing simulation workflows. Such developments could greatly reduce the necessity for elaborate and costly hardware iterations, embodying a methodological shift towards expedited virtual design and optimisation in power electronics.
Source: Noah Wire Services