I decided to upgrade from XP-G to XM-L and the cooling seemed to be my only problem. The light was too hot with XM-L, but I was not sure how hot. I needed a thermometer to test different heatsink combinations. Instead of buying a new multimeter with thermal probe, I found a cheaper solution.
Here is an idea for simple DIY digital thermometer:
It works really well. I measured and graphed my XP-G light and XM-L test rig. Ambient temperature was around 26 degrees and air was still. The XM-L is attached to L-profile aluminium stock that has around 112cm2 (17sq-in) cooling area. XP-G is inside the 1" aluminium tube with 73 cm2 (11sq-in) cooling area. XM-L is running at 2,8A and XP-G at 1,4A.
First, the picture of the test rig:
The heatsink, XM-L, Laura RS, two DealExtreme 4xAMC7135 drivers for 2.8A, thermal probe attached with thermal compound and tape, battery for powering the thermal probe and multimeter that measures the temperature (in tens of millivolts).
The temperature stabilized around 70C after 45 minutes.
The temparature stabilized around 56C after 40 minutes
Now, the tempareture of the LED die.
XP-G dissipates 4,4W and XM-L around 8,8W. Thermal resistance of the XP-G is 6C/W and XM-L has thermal resistance 2.5C/W. I searched the forum and found that the thermal resistance of the MCPCB is probably ~1C/W and the inside temperature of the case is ~12C higher than the outside. The last number is most dubious.
So, my XP-G is currently running at 56+4,4*(6+1)+12 = 98,8C and XM-L at 70+8,8*(2,5+1) = 100,8C (it is the test rig, there is no inside-to-outside transition).
If I put the XM-L inside the XP-G case, then it would run at 70*17/11 + 8,8*(2,5+1) + 12 = 150C.
This is already at the limit of the specifications, so I need to increase the size of the heatsink. I have couple of heatsinks, I will try to measure their effectivness.
Edit: I made a mistake in the last calculation, by not accounting the ambient temperature. Correct calculation would be 26 + 44*17/11 + 8,8*(2,5+1) + 12 = 136,8C.
Last edited by arnea; 03-06-2011 at 02:05 AM.
nice set up, very scientific approach.
However, I would suggest trying it with a very low airflow (from a large undervolted computer fan or a desk fan some distance away) and see what effect that has. It's fairly unrealistic to expect lights to be on full at a complete stop (for one thing, your riding buddies wouldn't appreciate being blinded) and a small amount of airflow makes an enormous difference.
Thanks for the post Arne, well done as all your posts are. I have some lights that get warm/hot, now I'll be able to see just how warm/hot. I appreciate your efforts.
You make a good point, the real world (at least in the UK) is very different than running indoors.
Originally Posted by mattthemuppet
Now I can replace teh cheap but calibrated thermon=meters I used!
It is surprising how LITTLE air movement influences the light temperature. A ceiling fan on slow half a room away, the circulation when the forced air furnace kicked on, required me to place the light on the keyboard drawer and close it under the desk to get a true still air room temperature reading. It is rare to see air sppeds lower than 5 mph/8 kph except near dawn and dusk, here. So stopped or no there is still a lot of cooling from air movement. Uphill into a wind the same speed could give you still air over the light, but you don't need full power.
Why so sesnsitive? Air is a VERY good insulator if it can only convect. Convective losses through boundary layers is slow and the air close to the light is hot. So lights protected from air movement heat up. Strip a portion to most of those hot layers away by supplying fresh lower temp air, and you not only get rid of the low conductivity issue but speed heat transfer because it is a function of temperature differentials of light to the air around it. So moving air is a conveyor belt moving heat away from the light.
Thank you for your comments. I found a mistake in my calculcations and fixed it in the first post.
I'm aware that the air movement will make the cooling much more efficient.
But building a light for everyday use that requires air movement in order do not break down or hurt someone is simply bad engineering. The body of the XM-L light would be at 94C. The user must not be punished for likely errors they will make sooner or later.
To be safe I can either reduce the power, increase the heatsink area or use smarter driver (e.g. lflex). The last option is the best, because it also protects me against situations when the even the normal cooling is not provided.
Right now I will try bigger heatsinks, but I'm tempted to convert to lflex.
I tested two heatsinks:
a) one was piece of L-profile that increased the total area of the heatsink 1,5 times. The maximum temperature of the heatsink was reduced from 70C to 58C - 1,375 times.
b) small black heatsink from old PC motherboard that increased the total area 1,72 times. The maximum temperature of the heatsink was reduced from 70C to 57C - 1,4 times.
I think it depends, if making a bike light you can get really bogged down into making it run really cool on your desk, for example, indoors one of my lights gets hot and trips the thermal protection I built in, yet I live in the UK and normally ride my bike outside where it runs cold to the touch.
I'm with the others. I think you should allow for moving air in your calculations. If you lower the input current you completely lose the advantage of using an XML. Nothing wrong with trying a bigger heat sink but I think the option you mentioned of using a temperature sensitive driver would give you the better peace of mind. Whatever you do don't wuss out and pull a *Dinotte. ( * lower the input current )
Originally Posted by arnea
It is always a good idea to use a lower output mode ( or turn the light off ) when resting. Do that you shouldn't have a problem regardless of what options you do or don't do.
You don't even need a fancy driver, something simple like a thermostat switch is a great option - http://airpax.sensata.com/site/utili.../pdfs/6700.pdf - They are the ones I use. You could use one to cut out one of the drivers if things get too hot.
Originally Posted by Cat-man-do
arnea, it looks to me like you will be fine with a little air flow provided that all of the ali to ali joints are good.
I tested my twin XP-G light in still air in the house and was quite shocked how fast it heated up and it didn't take long for the bFlex to dim the light when set to 70C.
But I used it as a walking light last night for 2 hours (on full power) with it on a head band and even at a very slow, walking through mud and over logs pace it was only around skin temp and never dimmed once.
I'm quite amazed at what a very small amount of air flow can do.
Last edited by yetibetty; 03-06-2011 at 06:35 PM.
Thank you. That was good tip. I used slightly different chip:
Originally Posted by MrLee
I hoped that I can drive the 7135 chips directly, but unfortunately that was not the case I needed two extra transistors and resistors to switch off half the 7135's when temperature reaches 65 degrees. I will draw up the schematics when I complete the light.
Here is the temparature graph: