Thermal Resistance Modeling of Active Heat Sinks
 

Hi everyone,

I've been working with this problem for a bit now, and haven't found a straightforward method of attacking it which actually works for me yet. Maybe you have? I am in no means an expert with this software, but I'm really trying to learn.

I've been running CFD thermal analyses on several electronic packages using Floworks/FloSim (2010 now), and one of the biggest problems I've been running into is the appropriate thermal resistance modeling of active heat sinks/CPU coolers when the specs are not given by the manufacturer. I can't work the whole model into the larger assembly, as the mesh necessary to model 0.0075in fins is entirely too large for my computer. I have the "real" values as determined by bench tests, but the simulation values don't line up.

I've attempted to model the finned heat sink, fan, and CPU similar to what is in the tutorial and then calculate deg-C/W based on the numbers generated from that simulation and plug them into an appropriately sized "Heat Sink Simulation" in the larger model. As I write this, the simulated point and surface temperatures are about 35C less than measured (and other times they come out as melting solids). I'm using psi=(Tsurface-Tambient)/Power dissipated. How close to the heat sink should I be calculating the local ambient temperature if I'm using that formula? Or should I be looking at it from a completely different method?

I could go through an entire list of other things I've tried, but I'd love to know if anybody has an actual answer to tackling this than trying to come up with some nonsense workaround.

Thanks a bunch.

Use Blocking or Local Mesh
 

Hi Castle,

You might have tried this method, but even then I am re-iterating for future engg having this doubt. I am assuming you are using Flotherm/Icepak.

The heatsink, CPU and the fans can be modeled in the single assembly. You can use Grid constraints around this assembly to capture the features. Finer the mesh, accurate the results. I know, fin thickness is very less. But it has to be captured for accurate results.

More over, CFD is already an approximation from the continuous reality to Digitized (discontinuous) simulation. In this if you again approximate with the resistance values, the results are going to be very approximate like having 35 Deg delta.

Hope this helped.

Megatron

if you want to obtain accurate result from simulation, many factor needed to be considered. firstly, the boundary condition should be as closer as to real condition, and secondly, the proper setting of each components, for instance, fan curve, material properties. the third, make proper meshing for your system.

proper mesh should means that sufficient meshes around the components you concern, and meanwhile make less meshes around the area you do not concern..

so what I want to say is software is just a tool, which is able to help to tell you result in advance. how to make it get accurate result is strongly depending on how to use it..you need experence a lot to learn a lot from failure, ,,

Hi Castle,
the "heat sink simulation" is a simplified model and therfore has less accuracy as it doesn't represent the original geometry in detail. This is a block approach and based on emperical values calculated from the boundary condition settings such as the pressure drop and thermal resistance.
If you haven'r defined these values correctly the difference to the measurements will be even more in the worst case.

I don't know which tutorial you mean as in FloEFD there is no tutorial about the heat sink simulation. Maybe FloSimulation has different or additional tutorials.
But that's usually the way to do it, using a small scale detailed model to determine these values and then use them for the large scale model with the simplified approach of the heat sink simulation.

Regards,
Boris