Spreadsheet Based Thermal Resistance Networks
Click here to download the presentation.
Overview: First order, preliminary calculations play a critical role in any initial engineering design. In the case of electronics cooling, these calculations typically focus on identifying a primary thermal path and estimating the thermal resistance of the significant contributors within that path. As a design matures, designers need to account for thermal paths that include parallel and series resistances as well as multiple heat sources. This increased complexity generally transitions the calculations from the ‘back of the envelope’ to software tools such as finite element modeling (FEM) or computational fluid dynamics (CFD). This transition to more capable, but more complex, analysis tools may also lead to a transition in who is doing the analysis from the system designer to a thermal analyst who is more familiar with the software but perhaps less familiar with the specific design.
The presentation will present an approach for filling the gap between simple one-dimensional thermal calculations and FEM/CFD analyses. Excel has substantial analysis capabilities and can also provide relatively simple user interfaces. The tutorial will discuss methods to estimate the thermal resistances within a system and how to automatically populate those values into a thermal resistance matrix. Excel-based matrix analysis will then be used to solve for device temperatures. If there is sufficient time, the tutorial will also describe how the same analysis framework can be used to analyze the transient behavior of the system.
This tutorial is aimed at electrical, system and mechanical engineers who would like to have more ability to assess ‘what if’ scenarios with preliminary designs before turning to dedicated thermal design software. The analysis methods do not replace FEM or CFD in the design process. However, they can improve those tasks by allowing designers to better understand their systems and provide the thermal analysts with a more comprehensive design before the detailed analysis begins.
Ross Wilcoxon is a Principal Mechanical Engineer in the Rockwell Collins Advanced Technology Center. He conducts research and supports product development related to component reliability, electronics packaging and thermal management. He has contributed to the development of multiple systems for communication, processing, displays and radars for commercial and military avionics applications. His research areas have included rapid test methods for evaluating component reliability, heat pipes, liquid metal cooling, advanced composites and tin whisker mitigation. He is a past chair of the SEMI-THERM conference, has dozens of conference and journal publications and holds 24 US patents. Prior to joining Rockwell Collins in 1998, he was an assistant professor at South Dakota State University. He received B.S. and M.S. degrees in mechanical engineering from South Dakota State University and a Ph.D. in mechanical engineering from the University of Minnesota.
The following are questions presented to the speaker by the attendees during the webinar, along with answers to each.
Why do you need node 10 if you have node 5 at ambient?
Answer: If I was making a problem-specific spreadsheet, I wouldn’t need node 10. But this spreadsheet was set up to handle up to 10 nodes with node 10 being the ambient. Rather than change the notation for the network and the equations that I had shown initially, I just connected node 5 to my defined node 10 ambient.
Have you ever tried to solve a transient problem with a spreadsheet?
Answer: Yes, but I do a simplistic numerical integration to determine temperatures over time. You can use the thermal resistance matrix that you generate in this calculation. I discussed this at the 2012 IMAPS Thermal Management ATW and am providing portions of those slides to be posted if people are interested.
Please comment on how you decide which nodes should have a resistance defined and which do not. Also do all ambient facing nodes get a resistance to ambient?
Answer: The main assumption is that the entire node is at a uniform temperature, so you should try to define your resistance model in terms of blocks of material that will each be at a relatively uniform temperature. You may want to attend the short course on this topic at SEMI-THERM next March – I suspect that they will provide some insight for how to effectively define nodes.
Do you need to account for spreading resistance in the PWB from one node to another?
Answer: If that represents an important thermal path, yes. I have not come across a good way to do that yet, so I generally assume 1-D heat transfer if I need to account for coupling between components on a circuit board. I am planning on digging into that more in the future for an analysis that I need to do – perhaps I will have something ready for the next Thermal Live.