Session: 06-01 Two-Phase Cooling

Paper Number: 97172

97172 - Comparative Analysis of Direct and Indirect Cooling of Wide-Bandgap Power Modules and Performance Enhancement of Jet Impingement-Based Direct Substrate Cooling 

With the development of high power and torque machines, requirements for high power density electronics are increasing. Thermal management of such system requires high heat extraction. Wide band gap (WBG) devices permit high temperature operations of power modules, but miniaturizations of device increase heat flux density. Conventional air-cooling-based heat sinks-and cold plate based liquid cooling has been used for power electronics applications. For high power density applications, within the size constraint air-based cooling is mostly obsolete. Liquid cooling is the most popular choice for majority of power electronics applications where power modules are cooled generally by two different configurations. First one is direct substrate cooling where the power module is directly exposed with the coolant. The second one is indirect cooling, where power module is attached to a heat sink which is actively cooled by the coolant. In indirect cooling, the heat sink is attached to the substrate by a thermal interface material (TIM). Electrically insulated TIM is not as thermally conductive as solder and creates a “choke point” in the heat flow path from the chip to the coolant.  Solder based attachment is prone to delamination and eventually failures due to thermal cycling. For directly cooled device, no TIM is needed.

            In this paper high-performance jet impingement based direct cooling has been implemented for a SiC based DBC substrate for various power losses. Conventional jet impingement cooling found in literature shows high local heat transfer coefficient (HTC) with small liquid volume, but lower global/average HTC due to low fluid coverage in the impingement surface. In this paper, the limitations of conventional jet impingement-based cooling system have been addressed and an array of design iterations has been proposed to overcome the bottlenecks. Primary results show that with increasing nozzles per chip from 6 to 12, reduces the device temperature 14 %. A novel diverging nozzle design has been proposed which reduces the device temperature further 10% and keeps the coolant pressure drop under 2 psi. Finally, a novel topology has been implemented which deploys 60 nozzles per chip with non-uniform diameter. The distribution of diameter has been chosen by implementing population-based multi objective optimization algorithm genetic algorithm (GA). The optimized design archives 7% less pressure drop than the unform distribution of diameter while 4% further reduction of chip temperature. Transient analysis shows, jet impingement based direct cooling reaches steady state faster than indirect cooling.    

Presenting Author: Himel Barua Oak Ridge National Lab