I am aiming for 60ºC which is about what a noctigon can bring it to
Thats’s a high set goal which would require cca 1.2*C/W heatsink, not sure if it is achievable.
If we would ascertain a definite max temp for the LED, that would give us more headroom (size, price) to find a suitable heatsink.
This led can be run at 100 degC easily, the maximum spec-ed led junction temperature is 150degC !!!. High operating temperature is what these XHP-leds are all about!
And I suspect 100degC operating temperature is reached in many of the commonly used household led-bulbs that use leds with lower temperature specs.
A finned heatsink requires airflow (fan) or at least space for air convection (esp. above, hot air rises), i.e. a finned HS in a tube, under a closed cap will be inferior to the same space filled with solid copper/alu.
Thermal resistance (°C/W i.e. how much it heats up per W flowing through it) depends on volume. A larger HS will have lower t.r. The standard formula is (m²K)/W (K being interchangeable to °C/F). In other words it can be scaled up arbitrarily by using a larger HS - if it wasn’t for thermal conduction not being instant. The larger the HS, the more difficult it becomes to dissipate heat evenly over its whole mass.
Thermal resistance of an XHP70 is 0.9°C/W => 27°C at 30W that results in 48-62°C (aka junction temp, TJ) at 21-35°C room temperature.
The data sheet also states “Unlimited floor life at ≤ 30 ºC” so we can assume 30°C as the ideal TJ.
Wrt drivers, 1) a zener moded atiny driver will work with a 12 V supply. (Spoke with RMM about a 3S, 3 series LED system previously and this will work). 2) If a lower current of 3A is acceptable, a lm2596 type buck driver will work with.13.5V or higher input. A knob attached to the variable resistor would be used for dimming. ATM, I’m using a lm2596 driver with a 24V apple laptop powersupply and a cheap voltage display to drive 6 XTE LEDs for the room’s main light source. At 19.6V, each LED gets about 900mA. at 20.6V, about 1400mA and 14.9V, maybe 3 lumens. Going good for a few month now.
1) At 2.4A, a XHP70 needs about 12,7V according to the Cree datasheet, so 12V won’t work with a buck driver.
2) In most cases you don’t want to use two drivers - one mains powered constant voltage, and one low voltage constant current. You lose efficiency, and easy dimmability. But it makes sense if you want to re-use a power supply you already have (one not meant for LED lighting). When you dim by reducing voltage, the brightness for one knob setting is different for each LED chain because of the randomly different Vf of each LED for a given current.
All the suggested power supplies are officially certified for LED lighting, deliver constant current, and are linearly dimmable by available 1-10V dimmers - though they are not cheap (both power supplies and dimmers).
I was reading about this Peltier cooler , I don’t think it quite suites the need because it creates another heat removal issue but, then I noticed the 12 volt 6 amp power supply. Now if this power supply is safe and we can source it, that might provide the power.
how about two separate drivers, one at about .5a (~750lumens), and the other at about 2a and just switch each one on/off separately?
There are many 12v supplies available easily without having to get a kit
http://www.fasttech.com/products/0/10003749/1270701-12v-5a-regulated-switching-power-supply
that would be pretty easy, 2 7135s then a mosfet, or just run goups of 7135s 1,2,4,8. would make for a max of 5.25A if using 350ma 7135s
an 8pin attiny13 has 6 io pins which is enough not using pwm 4 for the 7135s and 2 for some sort of input.
It does not work like this, LED temperature is determined by how much power is going into it, LED efficiency, thermal mass of the heatsink and heat shedding characteristics of the fixture. There is no such thing as 30ºC junction temp, even the 25 is determined experimentally today and was determined by lighting the LED for hundredths or thousands of a second.
I thought of the meanwell ones as well but didn’t get around to checking them out, since the LED would require about 13V under load i think if they have a constant current at 15V and 2.4A that would be perfect! The LPV35-15 in your link would be amazing if it was constant current instead of constant voltage. Do they make such a model, 2.4A at voltage that covers 12.6-13V? This would eliminate the need for a driver because its already “driven” from the power supply which is great. However driver research is continuing for future iterations (modes, dimming etc.) and for other power supply use (at your own risk). An over temp shutdown at the LED would be nice if no separate driver circuitry is used, i’ll look into it.
I don’t think we need air pipes and whatnot, we just need a heatsink that can dissipate 30W and stay under 60ºC. We can run the LED at 105ºC as mentioned, but i prefer to aim for 60ºC if its attainable, and will give more LED output. If tests show this is not feasible then the target temperature will be raised. I will define 105ºC as the max LED temperature but i would like a lower temp for safety. In addition LED light bulbs are able to manage to not burn out regular LEDs and they are smaller and not finned so until proven otherwise this is a reasonable goal.
Its interesting that we can pick some cpu heatsinks with fan, all we have to determine is one that works then any rated for the same processor should work fine. The goal is for this lamp to not require a fanned heatsink, testing will need to be done on some heatsinks so there is data to work with.
I am glad to see non chinese website heatsinks, name brand products are a great idea, of course since this is a DIY set of instructions any builder can substitute as they see fit (please don’t use chinese power supplies, your life is not worth destroying to save a few bucks)
At around 100c the heatsink is to hot for safety, the “safe” temp is 60c (in Australia at least).
Hot drinks should not be served over this temp (remember when Macca’s was sued because a customer burned themselves with a coffee, they lost because they couldn’t prove it was at or under 60c).
Having a special needs daughter I MUST have a cold water mixer in the hot water line to keep the temp at or below 60c for when she turns off the cold only, ever seen a human lobster :bigsmile: at least she doesn’t do any damage to herself.
Don’t think it does not apply here, neighbours kids come in with their parents, knock over the light and make a grab for it, and grab the heatsink, if it is at 80~90c it going to do damage 
Still following with interest 
Cheers David
Any incadescent floor lamp readily available today does not meet this requirement. Sure, we can aim for 60*C, but just saying….
also the floor lamp is inherently unsafe with smaller kids, it can be easily knocked down, had it happen a few times, for toddlers it’s like candy 
So no sharp edges, and has to have sufficent weight mounted to it’s floor stand, so it doesn’t tip over too easy.
exactly, an incandescent or halogen will destroy the LED in heat production
in the end i think the host will be the hardest part
An incandescent or halogen lamp generally have insolative material between the globe and the stand, also, again usually, the hot parts are tucked inside a shade of some sort, so the damaging heat is confined to parts not generally touched, even when grabbing it in a hurry, this design spreads the heat down to an exposed area, unless the heatsink is enclosed in a tube.
One more point, they are commercially made so the manufacturer is the one to be sued if the worst happens, with this you (we) are the manufacturers, but suing aside, do you really want to be the one responsible for a serious burn to some ones hand, especially a childs :weary:
There is no way I will have any surface over 60c in my house that can be touched by anyone.
Cheers David
A table lamp holding a CFL or incandescent is easily reachable, so i hope you have neither of these. I can guarantee millions if not hundreds of millions of people do.
Again i want 60C for good output, and reducing burn risk is an excellent side benefit so lets hope it is achieved.
2 lamps in the lounge, both at 6’ high and fastened to the wall, one in the dining room which can be touched easily, its at 55c, hot but not burning, no incans or halogen in the house.
60c is a good temp to aim for, if the sink is 60c wouldn’t that be about 75~80c at the led?
Cheers David (who did not mean to go off topic, sorry)
Edit rather than new post.
A bit of info on active cooling v passive cooling.
Yes, I want to be responsable for that. Just three weeks ago my 4year old touched the hot iron that was standing on its side on the table. I told him beforehand that it was hot and he should not touch it. He did anyway and he burnt his hand (small blister), he felt very miserable the next hours (pain was over when he went to sleep). The burn has healed now and this saved me from ever have tell him more than once from now on when something is hot :evil:
(to be clear, I'm not going to let him mess with the chip pan until he is 18yrsold)
So, how does it work, is there a formula; and what do they mean by the 30°C figure then if not junction temp? I’m pretty sure TJ<30°C is possible in general, see this thread about an overclocked 95W CPU being watercooled below 15°C http://www.tomshardware.co.uk/forum/275185-29-exploring-ambient-water-cooling.
Since we’re trying to implement passive cooling, how about heatpipes? They transmit heat 10-10000x faster than copper, can be bought in lengths beginning at 10 cm (from what I’ve seen on AE), in round or flat shape. A flat pipe would be pretty easy to mount/glue on top of a flat, lengthy HS; although that would mess with the LED-at-end-of-tube design, unless you want to bend the pipe, which is tricky.
Edit: Regarding experimentation, there are various thermal calc programs (LISA, Energy2D, Sauna, etc.) out there, some of them free. None of the ones I tested make design of round heat sinks easy.
I’m pretty sure under 30º is the storage temperature of unpacked chips in the reel, you can buy a bare uninstalled LED and it should still work years or decades from now if stored below 30C
Since your sure its possible please post some data and also explain the calculated and tested for minute fractions of a second that Cree has stated for the numbers in the datasheets 
I know nothing about heatpipes so i am way out of my element on that one, if you can link some with compatible specs it would be worth testing
Since we don’t have the LED to test with its worth figuring out some theoretical numbers in software so i appreciate your efforts