Abstract:This article characterizes the thermal behavior of a commercialized silicon carbide (SiC) power MOSFET module with special concerns on high-temperature operating conditions as well as particular focuses on SiC MOSFET dies. A temperature-dependent Cauer-type thermal model of the SiC MOSFET is proposed and extracted based on offline finite-element simulations. This Cauer model is able to reveal the temperature-dependent thermal property of each packaging layer, and it is suitable for the high-temperature thermal-profile prediction with sufficient computational efficiency. Due to the temperature-dependent thermal properties of the SiC die and ceramic material, the junction-heatsink thermal resistance can be increased by more than 10% under high-temperature conditions (up to 200 °C), which can considerably worsen thermal estimations of the SiC die and its packaging materials. Furthermore, the experimental measurement of transient thermal impedance was conducted under operating temperature variations (with virtual junction temperature ranging from 60.5 °C to 199.6 °C), and the effectiveness of the proposed temperature-dependent Cauer model was fully validated.

What problem does this paper attempt to address?

This paper attempts to solve the problems of signal compression and transmission in low - power wireless neural recording tasks. Specifically, the paper focuses on how to reduce the amount of data transmission through Compressed Sensing (CS) technology without sacrificing signal recovery quality and classification accuracy, thereby prolonging battery life. The paper proposes a deep - learning - based Quantized Compressed Sensing framework (BW - NQ - DNN), aiming to optimize the design of measurement matrices, non - uniform quantizers and non - iterative recovery solvers to improve the compression ratio and signal recovery performance. ### Main problems 1. **Sparse representation of signals**: Neural impulses are not sparse on common dictionaries (such as the discrete cosine transform basis and the discrete wavelet transform basis), which leads to a serious decline in performance when using these dictionaries for signal recovery. 2. **Measurement matrix design**: The measurement matrix must satisfy the Restricted Isometry Property (RIP), so designing an appropriate measurement matrix is crucial for signal recovery. 3. **Quantization error**: The error introduced by quantization is inevitable, but reasonable use of quantization can significantly reduce the number of bits required for wireless transmission. 4. **Efficient recovery solvers**: Traditional recovery algorithms have high computational complexity and can only be used for offline processing, while real - time neural recording applications require efficient recovery algorithms. ### Solutions The BW - NQ - DNN framework proposed in the paper solves the above problems in the following ways: 1. **Binary measurement matrix**: Directly learn a binary measurement matrix, which not only improves performance but also maintains the efficiency of hardware implementation. 2. **Non - uniform quantizer**: Optimize the non - uniform quantizer, which is especially suitable for the case of low quantization bit - depth and is significantly superior to the uniform quantizer. 3. **Non - iterative recovery solver**: Learn a non - iterative recovery solver through a deep neural network, which significantly improves the recovery accuracy and computational efficiency. ### Experimental results The experimental results show that BW - NQ - DNN not only significantly reduces the number of transmitted bits, but also achieves high recovery performance and pulse classification accuracy in high - compression - ratio tasks. Compared with the existing state - of - the - art methods, this framework has significant advantages in recovery accuracy and computational complexity. ### Formula examples - **Basic formula of quantized compressed sensing**: \[ y=\Phi x \] where $\Phi\in\mathbb{R}^{M\times N}$ is the measurement matrix, $y\in\mathbb{R}^M$ is the compressed measurement value, and $x\in\mathbb{R}^N$ is the original signal. - **Optimization objective of non - uniform quantizer**: \[ \phi^*=\arg\min_{\phi}E(\|x - \hat{x}\|^2) \] where $\hat{x}$ is the signal recovered from the quantized measurement value $z$ by the CS recovery solver. - **Quality assessment of signal recovery**: \[ \text{SNDR}=20\log\left(\frac{\|x\|_2}{\|x - \hat{x}\|_2}\right) \] \[ \text{CA}=\frac{\text{Number of correctly classified pulses}}{\text{Total number of pulses}}\times100\% \] Through these methods and indicators, the paper demonstrates the effectiveness and superiority of the proposed framework in wireless neural recording.

Thermal Characterization of Silicon Carbide MOSFET Module Suitable for High-Temperature Computationally Efficient Thermal-Profile Prediction

High-Temperature Characterizations of a Half-Bridge Wire-Bondless SiC MOSFET Module

Comparison of High Voltage SiC MOSFET and Si IGBT Power Module Thermal Performance

Structure and the Thermal Management of a 3-D Silicon Carbide MOSFET Module

Analysis of Novel Packaging Techniques for High Power Electronics in SiC

Compact-Interleaved Packaging Method of Power Module With Dynamic Characterization of 4H-SiC MOSFET and Development of Power Electronic Converter at Extremely High Junction Temperature

Analysis and Modeling of Temperature Dependence of I-V behavior in Silicon Carbide MOSFETs

High Temperature Stability and the Performance Degradation of SiC MOSFETs

Modeling of SiC MOSFET with temperature dependent parameters and its applications

An Analytical Switching Loss Model for SiC MOSFET Considering Temperature-Dependent Reverse Recovery over an Extremely Wide High-Temperature Range

A Temperature-Dependent Analytical Model of SiC MOSFET Short-Circuit Behavior Considering Parasitic Parameters

High-Temperature Characterization of a 1.2-Kv SiC MOSFET Using Dynamic Short-Circuit Measurement Technique

Overview of Finite-Element Analysis in Simulation of SiC Power Device Packaging

Development of a 1200 V, 120 A SiC MOSFET module for high-temperature and high-frequency applications

Junction Temperature Estimation of a SiC MOSFET Module for 800V High-Voltage Application in Electric Vehicles

Dynamic Performance of 6.5kV/400A SiC MOSFET Module and its Gate Driver

Reliability Issues of SiC MOSFETs: A Technology for High-Temperature Environments

Development of a steady state electrothermal cosimulation model of SiC power modules

High temperature stability evaluation of SiC MOSFETs

Electrothermal Analysis of a Double-side Cooling Power Module with Large-area SiC MOSFETs

Electrical Performances and Physics Based Analysis of 10kv SiC Power MOSFETs at High Temperatures