LED Simulation severly inconsistant with Real Life!

[Anonymous]

In order to ensure that the simulations I am producing on LED luminaires are correct, I have tried to simulate a real-life experiment consisting of an LED on a metal-core PCB stepped 10cm away from a Polyamide driver enclosure. Temperatures on the board and the driver case were measured. The Results are as follows:

PCB Temp Real Life = 138C

PCB Temp Simulation = 269C

Driver Case Temp Real Life = 60C

Driver Case Temp Simulation = 137C

As you can see, the difference is vast and after several different attempts i cannot seem to match real life!

The images below explain the real-life set-up and simulation.

Any help would be appreciated!

[Jon.Wilde]

You may need to consider radiation here, have you checked out the help guide on lighting?

[napoleonm]

Radiation is important in this simulation.

[Anonymous]

Hi, why is radiation important in this case, is it because of the close proximity of components? Or if radiation plays a key part, then why wouldn't it be important in all Natural Convection cases, or is it?

I would just like to understand radiation a bit more.

[Jon.Wilde]

High temperature differentials. Very generally speaking, once we have differentials above about 70C-80C, we should consider radiation. Obviously it is always having an effect but it is often minimal below this.

[napoleonm]

Radiation is an important mode of heat transfer at high temperature, and more from what I see your model is, I agree with wildej

[Royce_adsk]

As a side note, I know some people like to have a film coefficient on the top surface, but I prefer to only have a 0 pressure on the top.  Using just the 0 pressure is more inline with our standard guidelines.

Otherwise, I agree with the other comments.  Turn on radiation and assign temperature BCs to the 4 side walls of the air chimney. That should help.  Let us know where that takes you please.

Cheers,

[Anonymous]

Thanks for your help guys. Here are the results of a re-run with your comments and also a slight different way of modelling the MCPCB.

Any more suggestions to help closen the gap beween the model and real-life would be appreciated!

[Jon.Wilde]

This is far closer, 85%-95% accuracy is good. You could easily start to optimise the design from this point.

It might be possible to improve it further, probably through mesh refinement and ensuring that you have the emissivity values of each material 100% spot on.