8 Replies Latest reply on Aug 21, 2013 4:44 AM by Mark Keown

    Flow simulation for LED

    CW Sew

      hi guys,


      i did a flow simulation to study heat dissipation for a LED in a casing. compare with fluke thermal imager, the result is not even close. So wondering do i have mistake in the setting, therefore would like to ask :


      1. normally how you guys define the LED material? or which material you use to define LED?

      2. for the heat source of the LED, should i choose surface or volume source for heat generation rate?

      3. Out of these 2 ( LED Heat, LED Optical Power) data, should i go for LED heat or LED optical power?


      Looking forward for some replies. thanks!



        • Re: Flow simulation for LED
          Bill McEachern

          I would use a volume source

          I would use silicon as the material.

          I would make sure that the thermal dissipation matches what really happens and it is typically not anywhere close to the input power - which is the usual problem.

            • Re: Flow simulation for LED
              Jared Conway

              don't forget about radiation!


              CW, have you thought about what the output is from our thermal imager and how it would compare to what you're getting out of solidworks?


              also, have you taken a step back and done a controlled experiment? something you can calculate, something you simulate and something that you can physically test with minimal variables? 90% of the time when CFD doesn't match real life, it is the setup and assumptions that are made.


              i'd also consider breaking down what the LED does to make sure that you have all its "physics" encorporated.

                • Re: Flow simulation for LED
                  Jared Conway

                  also to just agree with bill, right component size/shape, materials as close as possible, volume source and correct loads (thermal + radiation).


                  when it comes to the component, you can breakt it down all the way to its elements, but you have to ask yourself, is that really necessary?

                • Re: Flow simulation for LED
                  CW Sew

                  thanks for your suggestion. I try and hopefully the result is close.

                • Re: Flow simulation for LED
                  Harold Brunt

                  Another thing to consider when collecting thermal data from LEDs is that a thermal imager is best used in applications where the target is fairly large (like a bulb heatsink or fixture). The accuracy of the measurement will depend on the emissivity of the selected surface and the proximity of the surface to the heat source. For a trial use a surface that is fairly warm and is unfinished or polished aluminum, make a measurement and then slap some blue masking tape on it and remeasure.


                  Most LED MCPCBs have a thermal via that is a direct path to the die. You can assume the die is running up to 10C hotter in some applications but most are fairly close. Use a probe that is attached as firmly as possible and you should get a decent idea of the source temperature.


                  I don't run the thermals here, just the lighting simulations but getting the correct inputs is half the battle in any type of simulation.

                  • Re: Flow simulation for LED
                    Kyu-Hyung Kim

                    Typically LED component apply Two-resistor model (don't need material property, Juet Need thermal resistor(thetaJB or thetaJC).


                    Because LED heat flow path mainly go down MPCB, 2R model is fit for LED compoent.


                    Supplier is provided 2R data, Maybe you can easy find data


                    Total Power = thermal power + optical power,


                    Because Optical power is convert to light, We should apply thermal power.

                    • Re: Flow simulation for LED
                      Mark Keown

                      One thing that you can do to help understand the problem is to make a thermal-resistor diagram for the heat flow.  Evaluate each thermal-resistor value to determine where the problem may be and your confidence level in each value.  Spreading resistance is difficult to calculate but CFD will calculate this correctly.  The interface resistance MPCB to heat sink is also a difficult value.


                      Look at each element from lab test and see where the differences are between lab and CFD.  This will identify what elements need more attention / detail etc.  For a nice example of thermal-resistance please see the link.