Mountain Bike Reviews Forum banner
1 - 13 of 13 Posts

·
Registered
Joined
·
177 Posts
Discussion Starter · #1 ·
Hi all,

it is too cold in my garage so I am doing lots of thinking rather than building. This is going to get fairly heavy so stop now unless you are bored.......

A mate just happens to be a thermal simulation expert. Out of interest I got him to simulate an ally block 25x25x10mm with a 3W 5x5x0.5mm heat source in the middle of one face. Apparently steady state (25degC ambient) the whole thing goes to 130degC. If the ally is anodised it reduces to 102degC. The colour of the anodising doesn't make any significant difference.

Obviously this isn't representative of any light I might build but it might be useful if I had more info.

So far I know the following:
XRE: approx 3.7W input at 1A drive. 5x5mm solder (heat transfer) pad and a thermal conductivity between junction and pad of 8degC/W.
MCE. approx 10W input at 700mA/die. Approx 3x6 contact pad, thermal conductivity of 3degC/W (junction - pad).

What I really need to know is how much of that input power is converted to heat and how much to light. I can work out the lm/W figures (about 86 for an MCE M bin and 82 for a XRE Q5?) but that isn't quite what I need.

Somewhere I found the equation that luminous radiant power (watts) = lumens/683.
So 86 lm/W = 0.126W per W or 12.6% efficiency. Meaning that 87.4% of input power goes into heat which is a bit scarey because it implies the simulation above isn't too far from the truth......

Somewhere here I saw a rule of thumb for surface area of the heatsink per watt of input power. Does anyone know what it is? It would be good to simulate this and see if it gives the same sort of surface temperatures that you guys experience.

Toby
 

·
Registered
Joined
·
60 Posts
As I am in a process of building my first LED light (two 2xCree R2 lights) I was naturally curious as to what kind of heat sink would I need. After going through the Cree XR-E thermal design documentation I was able to calculate the maximal thermal resistance of a heat sink. With a good thermal contact between the emitter and the heat sink I got these numbers:

Vf=3.6V
If=1A
Tj_max=145C
Ta_max=40C - maximal ambient temperature

Nleds Rh_max [C/W]
1 20
2 9.6
3 6
4 4.3

The tricky part is to determine the thermal resistance of the actual heat sink. According to http://homepages.which.net/~paul.hills/Heatsinks/Heatsinks.html, when dealing with finned heat sinks (think old CPU heat sinks) the thermal resistance can be determined from a 'volume' of a heat sink (width x depth x height).

Nleds V [cm^3]
1 10?
2 25
3 50
4 90

I have a 50x50x25mm heat sink (V=60cm^3) so it should be enough for two (maybe even three) LEDs (with a matching thermal resistance of about 6C/W). This is a 'worst case' natural convection mode - standing still with no wind. With a forced convection the thermal resistance should be significantly better.

My heat sink has a surface area of about 300cm^2 and should, in theory be just about able to handle 3 LEDs so about 100cm^2 per LED would be an absolute minimum surface area.
Please note that this is all just pure theory.

I have also found the http://www.aavidthermalloy.com/products/extrusion/european_extrusions.shtml to be useful in determining what kind of thermal resistance can you expect from a heat sink. For comparison, according to the data on that site a heat sink most similar to mine should should have a thermal resistance of 5C/W which is almost exactly what I would expect from it's volume so I am pretty certain that this number is correct.

To sum it up. According to my calculations you should provide about 35cm^2 (5.6"^2) of an exposed surface area per watt.
 

·
Registered
Joined
·
177 Posts
Discussion Starter · #3 ·
well I found the 1 sq inch per watt recommendation. Interesting as the simulated design has very close to that. I doubt many people are getting their lights to 130degC so I think I need to talk to my mate again.......
 

·
Registered
Joined
·
177 Posts
Discussion Starter · #4 ·
dsvilko: interesting, I've not understood what you said properly but the conclusion - 35cm2 or 5.6" sq is 5.6x the recommendation found elsewhere. But maybe it matches better with the simulation.

I have found that the cree XRE datasheet recommends junction to ambient of between 10 deg/W to 25deg/W depending on ambient (110 to 50ish). (at 1A)
I am not sure I really understand that but I guess they are saying that a 25deg/W will give a 100degC junction temperature rise and a 10deg/W will give a 40deg junction temp rise.
This roughly adds up. 3watts x 10 deg/W = 30degC. Jn-heatsink is quoted at 8 so total of 38degC rise.

If you are using a 5degC/W heatsink and 2x3W R2 leds then I would assume your junction temps would be roughly 5*6+8=38degC. If ambient is 25 then 63deg junction temp?

It sounds like you understand this much better than me, can you confirm what I have just said is roughly right?

Toby
 

·
Registered
Joined
·
60 Posts
First off, even though I am a physicist I knew next to nothing about thermal management a few weeks ago. I'm certainly no expert. All I did is to read a Cree's thermal management document and google a bit. I admit that numbers I got are a bit on a safe side but they should not be very far off. On candlepowerforums I found mention of 3sqin/W rule of thumb which to me sounds about right. With a significant forced airflow maybe you could even get by a 1sqin/W but I would not be comfortable sitting still with that setup.
My calculations are based on the available data for the classical finned heat sinks and should be more or less correct. The translation from the actual heat sink data to an abstract surface area data is a bit of a guesswork.
I also don't understand what junction-to-ambient data means but in the XLampThermalManagement document they are quite clear as to how to calculate the necessary thermal resistance. Did you go through this document?
Also keep in mind that most people probably don't know the emitter temperature in their lights. With a less than ideal thermal path between the star and the heat sink you could easily get maybe 15C/W thermal resistance of emitter-heat sink alone. That means that with 50C light body temperature the emitter could easily be overheating.
 

·
Registered
Joined
·
177 Posts
Discussion Starter · #6 ·
well I was an electronic engineer about 15 years ago so you are better qualified than me still. This is starting to make my head hurt :)

I hadn't seen that cree document, I'll give it a read. I found the 1sqin/W here somewhere. It does sound like the 3sqin/W you found on candlepower is the better figure.
It is really that emitter (junction) temperature that I was trying to understand. In the ideal case where the emitter is perfectly connected to the headsink we could say that the emitter is pretty close to the surface temperature of the heatsink. However, as you say the thermal path from the junction to the heatsink will be far from ideal.

This is really why I wanted to go down the simulation route. The design I am working on has a number of thermal limitations (different materials etc) and no proper heatsink so I wanted to see if I was overstressing it.

I've just had a chat with the thermal guy. The trouble with the simulation is that he is used to using data I don't have and is struggling to translate what I do have to the model. .I am beginning to think the traditional approach (more like what you have done) is going to be easier.
Or maybe the suck it and see approach most here have done is sensible, I'm not trying to make a commercial product here!
 

·
Registered
Joined
·
60 Posts
The emitter is on nowhere near the same temperature as the outer heat sink. Even with an extremely good thermal pathway between the star and the heat sink, the junction (emitter) temperature is 35-40C higher than the heat sink that it's mounted on (on the point of contact)! When you add up a thermal resistance of the heat sink the temperature difference becomes even more significant. That's certainly something to keep in mind when trying to assess the adequacy of your heat sink from it's outer temperature.
Maybe I just worry too much but I have still not seen a modern high-power LED in person (waiting for my first leds to arriver from DX) and would really like not to burn them :) Even though I live in a capital city of Croatia (Zagreb) I have yet to see a single Cree emitter mounted on a bike. Here a norm for a commuting bike is a small led blinker or rarely a 2.5W halogen. I guess my 4xCree XR-E R2 (WG) will stand out :)
 

·
Registered
Joined
·
177 Posts
Discussion Starter · #8 ·
I agree the emitter is no-where near the heatsink temp. That was only in my ideal case where there is no thermal resistance from the junction to the heatsink. In practice it will be more like 10degC/W (according to cree) which matches your estimate of 35degC difference.

I think that we are both worrying more than we need to as there are plenty of marginal designs (thermally) out there that work just fine, particularly as there is normally plenty of airflow on a bike and ambient temperatures are lower when it is dark. However, it is quite possible these designs are getting less light and shortening the life of the emitter without realising it.
Whether this really matters is arguable. I think someone calculated that even running close to the max junction temperature will still give a life of a year or so and I doubt anyone will really notice a 20% drop in light output (or whatever it is) in practice.
 

·
Registered
Joined
·
60 Posts
Well I am pretty certain that my two primary lights should be ok. What I would like to know is is something like this really possible:
http://www.dealextreme.com/feedbacks/browseCustomerPhotos.dx/sku.3256~id.12920
http://www.dealextreme.com/feedbacks/browseCustomerPhotos.dx/sku.3256~id.12919
I have the same light which I would also like to modify with a spare Cree Q5 and a 16-mode driver. The light currently has, beside 2 standard leds, a 1.5W halogen bulb that draws about 0.3A.
What do more experienced people here think? Can it be done?
 

·
Registered
Joined
·
60 Posts
Thanks for the links!

According to the charts in the Cree XR-E led specification document it should be possible to determine the junction temperature from the drop in the light intensity. As the junction temperature rises there should be a linear drop in the light intensity:

Tj = -390*Imin/Imax+416

Where Imin is any kind of light intensity measurement (lumen, lux..) after a few minutes that the light was turned on and Imax the same measurement immediately after turning the light on. It would be nice if someone with a lux-meter would try it. :)
 

·
Registered
Joined
·
536 Posts
tobymack said:
Hi all,

it is too cold in my garage so I am doing lots of thinking rather than building. This is going to get fairly heavy so stop now unless you are bored.......

A mate just happens to be a thermal simulation expert. Out of interest I got him to simulate an ally block 25x25x10mm with a 3W 5x5x0.5mm heat source in the middle of one face. Apparently steady state (25degC ambient) the whole thing goes to 130degC. If the ally is anodised it reduces to 102degC. The colour of the anodising doesn't make any significant difference.

Obviously this isn't representative of any light I might build but it might be useful if I had more info.

So far I know the following:
XRE: approx 3.7W input at 1A drive. 5x5mm solder (heat transfer) pad and a thermal conductivity between junction and pad of 8degC/W.
MCE. approx 10W input at 700mA/die. Approx 3x6 contact pad, thermal conductivity of 3degC/W (junction - pad).

What I really need to know is how much of that input power is converted to heat and how much to light. I can work out the lm/W figures (about 86 for an MCE M bin and 82 for a XRE Q5?) but that isn't quite what I need.

Somewhere I found the equation that luminous radiant power (watts) = lumens/683.
So 86 lm/W = 0.126W per W or 12.6% efficiency. Meaning that 87.4% of input power goes into heat which is a bit scarey because it implies the simulation above isn't too far from the truth......

Somewhere here I saw a rule of thumb for surface area of the heatsink per watt of input power. Does anyone know what it is? It would be good to simulate this and see if it gives the same sort of surface temperatures that you guys experience.

Toby
Don't forget about the ?degC/W of the thermal compound between the heat sink and emitter or MCPCB
 

·
Registered
Joined
·
177 Posts
Discussion Starter · #13 ·
in the cree thermal management doc that dsvilko found it assumes about 1degC/W for the MCPCB and the thermal contact to the heatsink. I would have guessed it would be higher than this but I guess Cree know better... :)

I think that assumes the MCPCB is nice and flat as is the heatsink and so the thermal paste is only filling very small defects, not creating a layer between the two. If you are using thermal epoxy to stick down the MCPCB then I think it would be much worse than this.
 
1 - 13 of 13 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top