Open source thermal simulation and analysis with 3D visualizations
Published:
Topics: Open hardware, Open simulation, Open source tools
Thermal simulation and analysis is an invaluable tool in hardware design and integration, providing critical information about thermal performance of a device (such as temperature distribution across electrical and mechanical components) even before creating physical prototypes. It’s especially important in the case of complex, multi-layer designs such as the ones Antmicro develops for their customers, since the progressing miniaturization of hardware components also means a higher chance of overheating due to heat concentration. To avoid faulty operation or even failure of the device, thermal analysis is a standard verification step across many industries, including industrial machinery, automotive, space and consumer electronics.
In space applications specifically, vacuum testing is an important validation step required to ensure the device will remain operational while deployed to the orbit. Vacuum testing is arranged typically once the initial version of the space equipment becomes available, but the sooner the thermal phenomena gets simulated and addressed, the faster the device can be cleared for takeoff.
In this article we describe an open source-driven workflow for thermal simulation and analysis with Blender visualizations, complementing our hardware design testing process which already includes Signal Integrity simulation and electromagnetic field scanning. Antmicro’s thermal simulation solution has already been used to verify the thermal efficiency of some of our enclosure and radiator designs, including in vacuum, which is showcased in the following paragraphs.
Open source-driven thermal simulation and analysis
Antmicro’s thermal simulation workflow uses the Finite Element Method (FEM) for numerically solving differential equations, which involves dividing a complex shape into simple polygons (a mesh). FEM uses a representation of physical dependencies in the form of a set of third order base equations, and is a common choice for heat transfer simulations.
The analysis is performed using the following open source tools:
For data processing, the flow uses:
- ccx2paraview - CalculiX to Paraview converter
- ParaView - results previewer
The results can be then presented in the form of Blender animations, as shown later in the article.
The diagram above shows the general workflow of thermal simulation and visualization. First, you need to import or create a 3D model representing the element that you want to simulate. You also need to define material and environmental constraints for the simulation, such as initial temperature, heat flux, heat source and temperature boundary condition. Next, the 3D model needs to be converted into a mesh, using gmsh or netgen. The simulation itself is computed by the CalculiX solver. The results can be viewed in ParaView; they can be also exported to Blender in order to create 3D renders and animations.
These steps can be invoked from the command line which makes this method a perfect match for a CI-driven automation flow.
The detailed simulation procedure is described in this README.
Convection-based simulation and verification of Antmicro’s Raspberry Pi 5 cluster enclosure
In order to illustrate the process, we’ll use our open hardware design of the Raspberry Pi 5 cluster enclosure that has been recently published on GitHub.
We assumed two simulation scenarios: Raspberry Pi 5 idle state, with a CPU power consumption of 2.5W, and under maximum CPU stress, operating at 8W. In idle state, the averaged film coefficient for the enclosure was calculated to 5.14 W/(m^2 K), and for the CPU stress it was set to 6.61 W/(m^2 K). The ambient and initial temperature in both cases were set to 25 *C. The enclosure simulation time was set to 100 minutes, during which the temperature stabilized. At this point equilibrium was attained between the power supplied to the enclosure and the heat dissipated into the environment.
In order to verify the simulation steps described above, we measured the surface temperature of the enclosure part in question while it was attached to a GBR-618-12-20-2 thick film heating element emitting 2.5W and 8W respectively. The temperature was measured with a TM-321N thermometer with a K type thermocouple.
As illustrated in the graphs above, Antmicro’s thermal simulation workflow allows us to predict the maximum temperature that the enclosure will reach under a specified heat power load. By identifying potential thermal issues early in the design process, we can optimize cooling solutions, reduce energy consumption, and ensure stable performance across various operational conditions.
Simulation and verification in a vacuum environment
Another example involves predicting thermal performance in a vacuum environment, where specialized equipment is required for testing. Initial simulations accelerate development, allowing us to improve our designs early and reduce the need for extensive, iterative physical testing.
In the example below, we repeated the simulation scenarios described above, but this time in a vacuum environment. The emissivity coefficient for black-anodized aluminum was set to 0.77 and the simulation results were verified using a vacuum chamber with an internal pressure of -0.98 bar.
Thermal management for size-constrained devices
Antmicro offers comprehensive hardware design, integration and verification services. We employ CI-driven visualization, verification and simulation routines to identify all the design challenges as soon as possible, allowing our customers to reduce iteration time and development costs.
For use cases requiring size-constrained devices, such as UAVs, VR/AR, robotics and space, we provide miniaturization options as well as simulation-based Signal Integrity analysis, electromagnetic interference scanning and now also thermal analysis.
If you’d like to learn more about how Antmicro’s services can improve your hardware development workflow at any stage of the process, don’t hesitate to contact us at contact@antmicro.com.
