AnsweredAssumed Answered

Improving Heatsink for LED module

Question asked by Charles Fried on Jun 20, 2014
Latest reply on Jun 26, 2014 by Jared Conway

Like • Show 0 Likes0
Comment • 12

Hi there,

This is the first time I'm using heat simulations to improve a design so I am after some input on how it is best acheived.

I have an LED module which reaches a max temp of 75 degrees after 30 min, this sits onto the heatsink with some heatsink grease.

I'd like to do a study to find out wether tweaking the heatsink in some places can improve heat dissapation.

Can someone point me in the right direction? I am aware there is two types of studies steady and transient?

Which type of thermal load should I use?

Many thanks,

Charles

No one else has this question

Outcomes

12 Replies

Jared Conway Jun 20, 2014 5:37 PM
Mark Correct
Correct Answer
tutorials in the software should get you started
then flow simulation training from yoru var www.hawkridgesys.com if you're on the west coast or in canada or ok with online
we also have application specific training if you're interested

i'd say you start with a steady state analysis
thermal load, start with heat generated on the LED and then after that youcan go to adding radiation as appropriate
Like • Show 0 Likes0
Actions
- Charles Fried @ Jared Conway on Jun 23, 2014 5:00 AM
  Mark Correct
  Correct Answer
  Thanks for the reply.
  Can you give me a little more information as to where it would be appropriate to add some radiation?
  
  I started by setting all my contact areas with the resistance value of my thermal glue (1.5K/W). I then added a temperature to my LED Driver (80C) and some convection to all the other parts but that didn't seem to work.
  (Ignore the heat power load in the picture, capture was taken whilst the next simulation was calculating)
  
  I then changed the temperature load for a Heat Power (8W - Led consumtion) which gave me slightly better results.
  Surely the 8w power doesn't bring the temperature up to the 80C I'm reading on a my real life thermostat?
  
  Unfortunalty due to the computer I'm using each simulations is taking over half an hour, leaving very little room for trial and error so any advice as to how I can get a more accurate simulation would be much appreciated.
  
  Thanks in Advance.
  Like • Show 0 Likes0
  Actions
  - Jared Conway @ Charles Fried on Jun 24, 2014 12:29 AM
    Mark Correct
    Correct Answer
    Half hour isn't bad.
    
    If you want to cut down time, do a hand calc first and build your analysis manually. That will give you a ballpark to work with and also help you identify what methods of heat exchange need to be taken into account.
    
    For example , radiation always exists but if convection is more powerful, sometimes it can be ignored. There is no right answer. The results in flow will be accurate to the assumptions you have. Reduce assumptions, the closer it will be to your physical results. Same thing goes for materials used and how you add the power. Some trial and error will be needed.
    Like • Show 0 Likes0
    Actions
    - Charles Fried @ Jared Conway on Jun 26, 2014 5:01 AM
      Mark Correct
      Correct Answer
      I should of mentioned half hour was for a very simple simulation (little power and convection), once I added radiation and selected all the faces it took 7:30 hours to complete, only to find out it had crashed when I returned.
      
      Ok I guess that makes it a little clearer, my LED module is "trapped" inside the lamp shade so that cancels all radiation to the heatsink.
      
      As far as I'm aware it shouldn't matter wether the temp values differ to real life measurements as long as I consistently use the same input values throughout all heatsinks.
      Like • Show 1 Like1
      Actions
Charles Fried Jun 24, 2014 11:38 AM
Mark Correct
Correct Answer
Ok, so after many failed attempts this seems like the most realistic one although it's still far from anything I'd expect.
Surely the part sitting against my LED Module (turquoise) should be pretty much the same temp/colors?!
It also indicates that the top of the heatsink is at room temp, which is not possible, any ideas?
Like • Show 0 Likes0
Actions
- Jared Conway @ Charles Fried on Jun 25, 2014 12:08 AM
  Mark Correct
  Correct Answer
  can you elaborate on your:
  1. setup
  2. expectations
  3. what approximations/assumptions you think your analysis is making?
  Like • Show 0 Likes0
  Actions
  - Jared Conway @ Jared Conway on Jun 25, 2014 12:08 AM
    Mark Correct
    Correct Answer
    it would also be good to know what kinds of test cases you've run to make yourself confident that the software is working properly
    Like • Show 0 Likes0
    Actions
  - Charles Fried @ Jared Conway on Jun 25, 2014 5:27 AM
    Mark Correct
    Correct Answer
    Sure, so I've got a 6063 Aluminium heatsink, this includes the board on which the LED's and driver sit on (12v - 8W).
    All surfaces are then joined together with some thermal glue 1.5 K/W.
    The driver which produces the most heat has been tested in real life to reach around 80c with the current heatsink.
    Although when setting up the analysis 8W of heat power showed very little heat so I increased it to 50w.
    
    SW set up: Contact areas: 1.5 k/w
    
                     Convection: Coefficent 250 W/(m²K)
                                       Ambient 300 K
    
                     Heat Power: 50 W
    
    I'd expect the heat to transfer into the outer shell of the heatsink just like it would it real life.
    
    The analysis shows a very abrupt heat transfer, perhaps dissipating before it gets a chance to reach the heatsink.
    
    Once I am happy with the simulation I was then going to modify the geometry to try and increase heatflux into the outside thus reducing total temp of the driver and increasing product life time.
    
    As for the software test, I haven't done any. I never though I should consider it not working properly, this is a pretty basic simulation.
    It's also done pretty well with previous stress simlations, results which were tested on a physical model afterwards.
    
    If it helps here's a picture of the product, as mentioned in a previous post the outerpart is shielded from radiations coming from the module;
    Like • Show 0 Likes0
    Actions
    - Charles Fried @ Charles Fried on Jun 25, 2014 8:10 AM
      Mark Correct
      Correct Answer
      Here's a new simulation which seems to be more accurate but still not perfect as it indicates that the top is at room temp.
      
      The issue was mainly the Contact Set, although the numbers were accurate there was very little transfer.
      I've now set the whole model to"bonded".
      
      The convection also seems to be very strong, in the picture below I've only assigned it to the top lip of the heatsink.
      
      Like • Show 0 Likes0
      Actions
      - Jared Conway @ Charles Fried on Jun 25, 2014 11:55 PM
        Mark Correct
        Correct Answer
        250W/mK, is this thing oil cooled? not surprised that you have room temp there, you have great convective cooling.
        
        Note, this is solidworks simulation thermal analysis, not flow simulation. lots of assumptions/approximations that you should probably start tracking. for example for convection, you need to apply to ALL faces to have accurate convection. otherwise the faces are assumed to be insulated.
        
        is this an education project or commerical (job) project? getting some direct help would probably be really helpful.
        Like • Show 0 Likes0
        Actions
        Charles Fried @ Jared Conway on Jun 26, 2014 7:25 AM
        Mark Correct
        Correct Answer
        I see, guess that's also where I went wrong.
        After a google search is seems like the value should be more between 10 and 50 W/mK.
        
        It's a commercial project but the product is working as it is.
        There's always room for improvement especially as the product hasn't gone under any simulation before.
        As i'm sure you can tell I'm not an engineer but a product designer.
        I'm not after any accurate results but an approximative visualisation of the heat transfer and how changes in geometry would improve this.
        If it had a greater importance I would of seeked external help.
        
        I'm getting there though, determined to make use of this.
        Like • Show 0 Likes0
        Actions
        Jared Conway @ Charles Fried on Jun 26, 2014 11:42 PM
        Mark Correct
        Correct Answer
        be careful about not looking for an accurate solution
        you could choose boundary conditions to give you any result that you want
        and that would mean the heat flow wouldn't be representative of what is actually happening and so the design changes wouldn't necessarily be valid
        
        if you're looking for some help on setting it up so you can do some true design to design comparison, we could certainly help you at hawk ridge, jared@hawkridgesys.com
        Like • Show 0 Likes0
        Actions

Improving Heatsink for LED module

Outcomes

Related Content

Recommended Content