Spelling suggestions: "subject:"junction.these thermal resistance"" "subject:"junction.the thermal resistance""
1 |
Packaging of Enhancement-Mode Gallium Nitride High-Electron-Mobility Transistors for High Power Density ApplicationsLu, Shengchang 27 June 2022 (has links)
Gallium nitride (GaN) high-electron-mobility transistors (HEMTs) are favored for their smaller specific on-resistance, lower switching losses, and higher theoretical temperature limits as compared to traditional silicon (Si) power switches. They have the potential to dramatically increase the power density and efficiency of power electronics systems by replacing traditional Si-based switches.
However, GaN HEMTs have a faster switching speed compared to their Si-based counterparts. Minimizing the parasitic loop inductances of the GaN HEMT package is crucial for reducing electromagnetic interference (EMI) noise and voltage spikes. Another concern with GaN HEMTs comes from their lower thermal conductivity and smaller die size. The HEMTs generally have a higher heat flux density, and accordingly, demand better heat dissipation. Thus, innovations are needed for making GaN HEMT packages with low parasitic inductances and higher thermal performances to further their applications in high-frequency, high-power-density converters.
To reduce loop inductance, other researchers have embedded GaN HEMTs in a printed circuit board (PCB) and used plated vias for interconnections and heat dissipation. However, this approach requires more complex manufacturing steps and has lower thermal performance.
This dissertation introduces different embedded packaging techniques for 650V, 150A GaN HEMTs; this method involves interconnecting the bare chips between direct-bonded copper (DBC) and a PCB or between two DBCs, as discussed in Chapter 2. Vertical interconnections by gold pins and silver rods are introduced and implemented in embedded packages to limit the parasitic loop inductance within 1.5 nH and parasitic resistances within 1.5 mΩ.
The thermal performance of the embedded GaN HEMT packages is experimentally verified in Chapter 2; then, the junction-to-case thermal resistance (RthJC) measurement is discussed in Chapter 3. The common temperature-sensitive electrical parameters (TSEPs) of a GaN HEMT for junction temperature measurement lack sufficient sensitivity or stability due to the electron-trapping effect. The non-uniform distribution of the case temperature and a large temperature gradient between the case and heatsink also make it difficult to accurately measure the case temperature. In Chapter 3, gate-to-gate resistance (Rg2g) is selected as the TSEP for junction temperature measurement. The stacked thermal interface material (TIM) technique was used to reduce errors in case temperature measurement. This technique was implemented in a custom GaN HEMT package and in embedded GaN HEMT packages for measuring junction-to-case thermal resistance. The discrepancy between measurement and simulation is less than 20%, and the junction-to-case thermal resistance for embedded packages is within 0.1 °C/W.
Chapter 4 evaluates the reliability of the GaN HEMT embedded packages developed in Chapter 2 by utilizing a power cycling test. Monitoring the junction temperature of the embedded packages online is challenging during the power cycling test. Other approaches have used the on-resistance as the TSEP in order to monitor junction temperature for GaN HEMTs but this is not accurate due to electron trapping. As discussed in Chapter 3, Rg2g is chosen as the TSEP to monitor the junction temperature without worrying about the influence of electron trapping, and this approach cycles the embedded packages at 75 A from 25°C to 125°C. The packages can endure 23,000 power cycles before failure.
This work is the first to develop, fabricate, and characterize embedded packages for 650V, 150A GaN HEMT bare chips. These embedded packages with high-power-rated GaN HEMT bare dice provide an opportunity to reduce the number of paralleled power switches, reduce the system's cooling size, and increase the system's power density. In addition, this work is the first to develop the junction-to-case thermal resistance measurement technique by gate-to-gate electrical resistance and stacked-TIM for GaN HEMT packages. The technique helps enable solid thermal design for power electronics systems. / Doctor of Philosophy / Power switches are everywhere in our daily life. They are the fundamental elements in power converters for converting power to electric vehicles. As global power demand for these applications continues to increase, high levels of both efficiency and power density are crucial for power switches. However, traditional silicon-based switches are already very mature, and their properties are very close to their theoretical limits. For further improvement, researchers have tried to replace traditional Si switches with wide-bandgap switches, which have much higher theoretical limits. Gallium nitride high-electron-mobility transistors (GaN HEMTs) are one of the candidates.
However, packaging these switches (GaN HEMTs) is challenging due to their initial properties. They naturally switch very quickly and have smaller sizes compared to traditional Si-based switches. The fast switching speed brings high dv/dt and di/dt during the switching period. It causes voltage spikes and electromagnetic interference (EMI) issues. And the smaller size contributes to higher heat flux density, thus requiring more efficient heat dissipation. To solve the challenge of packaging GaN HEMTs, this dissertation has developed embedded packaging techniques to achieve quiet switching and good heat dissipation. These packaging techniques enable GaN HEMTs' advantages and increase the power density and efficiency of power electronics systems.
To experimentally verify the thermal performance of the embedded packages developed a junction-to-case thermal resistance measurement technique was introduced. The thermal resistance of a custom GaN HEMT package was measured, as were those of the embedded packages CPES also developed. The simulation results and the experimental results are close to each other.
Finally, to further evaluate whether or not the newly developed embedded packages are reliable, power cycling tests were carried out at I = 75 A. The packages survived over 23,000 cycles before failure.
|
Page generated in 0.1221 seconds