Heatsinks for DIY LED lamps

DonPetro

Well-Known Member
DP, that gives you 4550cm² so at 110cm²/W, it would be good for about 40W of dissipation. So that would be (2) Vero 18s at 700mA.

In this case, if you added an 80mm or 92mm fan and ran it a 9V, (~2W) it should be sufficient for 115W, plenty for (4) Vero18s at 700mA.
So i could get away with 5 Vero 10 @ 300ma. Is there a driver out there that would match up?
 

SupraSPL

Well-Known Member
The 97cri chips are pretty effective at coaxing common herbs to bolt.
Cool I did not know that and I learned a new term.

So i could get away with 5 Vero 10 @ 300ma. Is there a driver out there that would match up?
Yep that should work. So you need a total of 130vF. If you split them in 2 strings, the longer one would need 78Vf. This one should work but it would probably be only 270mA. I will look around to see what else I have that would work. If you want high dimming you could use the HLG-60H-C350A
 

Mechmike

Well-Known Member
So i could get away with 5 Vero 10 @ 300ma. Is there a driver out there that would match up?
This one would do nicely.
http://www.ebay.com/itm/Constant-Current-Driver-for-36pcs-1W-LED-25-36-x-1W-Driver-Waterproof-100v-240v-/321329909926?pt=LH_DefaultDomain_0&hash=item4ad0c150a6
It does run at 300ma which I think is better than 350ma for efficiency reasons. I have seven Vero 10s running in my clone and mother tents all driven by similar drivers and I'm happy with the amount of light they emit.
 

AquariusPanta

Well-Known Member
So i could get away with 5 Vero 10 @ 300ma. Is there a driver out there that would match up?
Hey Donnie. I've found that variables for any given COB project change freely as new ideas and constraints are applied to the puzzle; dimensions for a container, number of COBs, multiple drivers vs. one driver, etc.

That being said, you've got to learn how do the numbers for yourself, otherwise your missing out on some of the headache, which later on raises the level of enjoyment and self-empowerment that arises from doing it yourself (for the most part). I'm not only underlining you, but a bunch of others here on RIU.

At first, it's tricky as all hell doing the math but eventually it becomes second nature and before you know it, your doing the math in your head with no overhanging sweat. The reason I'm stirring the pot is that while Supra and others get a kick out of assisting members, they shouldn't have do all of the guesswork. Furthermore, if your dealing with electricity, you should at least be able to know how to calculate math. It's sort of like a precursor to an activity that could bring you lots of happiness or worse, self-imposed fatality.

And I welcome any arguments and/or negative vibes from others. This is a forum site aye?

Here's my treat; I go off this anytime I'm comparing driver and COB numbers:

http://www.meanwell.com/webnet/search/seriessearch.html

 

DonPetro

Well-Known Member
Hey Donnie. I've found that variables for any given COB project change freely as new ideas and constraints are applied to the puzzle; dimensions for a container, number of COBs, multiple drivers vs. one driver, etc.

That being said, you've got to learn how do the numbers for yourself, otherwise your missing out on some of the headache, which later on raises the level of enjoyment and self-empowerment that arises from doing it yourself (for the most part). I'm not only underlining you, but a bunch of others here on RIU.

At first, it's tricky as all hell doing the math but eventually it becomes second nature and before you know it, your doing the math in your head with no overhanging sweat. The reason I'm stirring the pot is that while Supra and others get a kick out of assisting members, they shouldn't have do all of the guesswork. Furthermore, if your dealing with electricity, you should at least be able to know how to calculate math. It's sort of like a precursor to an activity that could bring you lots of happiness or worse, self-imposed fatality.

And I welcome any arguments and/or negative vibes from others. This is a forum site aye?

Here's my treat; I go off this anytime I'm comparing driver and COB numbers:

http://www.meanwell.com/webnet/search/seriessearch.html

Thanks man i understand where you're coming from. I am getting a grasp on the math part of it. Sometimes i get an idea and am looking for an answer right away so i ask a point blank question. The answers i get usually help me to take it further and modify the idea on my own.
 

bicit

Well-Known Member
Just ran a test on the 3000K CXA3070 AB bin, mounted onto an Acrtic Apline 11 Plus heatsink with 92mm fan and Prolimatech PK3 thermal paste. The surface was flattened to 1000 grit. The idea of this test is check the theory that at 700mA, there would be virtually no decrease in light output from Ambient temp to fully warmed up, thermally stabilized. Ambient temp was 21C. I did the test with 3 different fan speeds, 5V, 9V and 12V.

At 700mA (25W)
With 5V fan there was a decrease of .9% and a Tj increase of 7C (based on the Vf change).
With 9V fan there was a decrease of .6% and a Tj increase of 6C
With 12V fan there was a decrease of .44% and a Tj increase of 6C

I also did the test at 2.2A (87W) but it was a Z4 bin and mounted onto a different CPU heatsink with a copper pad that was manufactured perfectly flat. The results were interesting though:
At 5V there was a decrease of 3% and a Tj increase of 22C
At 12V there was a decrease of 2% and a Tj increase of 18.5C.

An finally, tested a Vero18 3000k 97CRi at 2.2A (69W), mounted on a CPU heatsink with copper pad, manufactured flat. This Vero is running at very high current and it is a very low efficiency COB (18%) so I was expecting much worse performance than this, a testament to the excellent thermal package of the Vero.
no fan! - 22% decrease, Tj increase of 55C
5V fan - 12.5% decrease, Tj increase of 29.3C
12V fan - 5.8% decrease, Tj increase of 11C
Supra-do you have a vero 29 and a driver capable of running at at 300-400ma? Im curious about temps and output when the large chips are driven super soft. Particularly when passively cooled.

The 5000k chips make for an attractive T5HO replacement.
 

SupraSPL

Well-Known Member
Did some testing on the Arctic Alpine 11 Plus fan:

@ 4.91V draws 74mA, dissipates .363W (silent)
@9.32V draws 143mA, dissipates 1.33W (slightly audible)
@12.24V draws 193mA, dissipates 2.36W (audible bot not loud)

So if we add 20% power adapter losses it is:
.44W
1.6W
2.8W

So if you run a CXA3070 at 1.4A, it dissipates about 52W. With the fan running at 9V it adds 1.6W of power usage compared to passive cooling, so you could say it reduces efficiency of the light from 42% to 40.7%. I guess that raises the question, how hard should we run the fan?

It also makes me wonder, how much cooler are the active cooled COBs running compared to my passive cooled lamps. I will have to test that next.
 

AquariusPanta

Well-Known Member
Did some testing on the Arctic Alpine 11 Plus fan:

@ 4.91V draws 74mA, dissipates .363W (silent)
@9.32V draws 143mA, dissipates 1.33W (slightly audible)
@12.24V draws 193mA, dissipates 2.36W (audible bot not loud)

So if we add 20% power adapter losses it is:
.44W
1.6W
2.8W

So if you run a CXA3070 at 1.4A, it dissipates about 52W. With the fan running at 9V it adds 1.6W of power usage compared to passive cooling, so you could say it reduces efficiency of the light from 42% to 40.7%. I guess that raises the question, how hard should we run the fan?

It also makes me wonder, how much cooler are the active cooled COBs running compared to my passive cooled lamps. I will have to test that next.

I'm curious about the AA11P, as I just scored a few from Newegg (couldn't resist with free shipping!).

While testing it at different potentials (voltages), did you use different power supplies in order to record these figures or was there another way of generating them?

Also, I'm a little lost on your calculation for efficiency with the COB - I'm guessing your trekking over new territory.

So you added the dissipated wattage for both products, fan and COB, and then divided the original dissipated wattage of the COB by this (52W / 53.6) and then multiplied this by 42% to get 40.75%.

Interesting claim Supra. However, I am not convinced solely on the math that the COB efficiency is directly effected by the heat shed of the fan, as you have formulated, but I do believe that fans do give off heat. Whether or not it's negligible is questionable.


"The amount of heat produced by the fan is more than just the heat you can feel from the motor. ALL of the energy used by the fan becomes heat, and this amount of heat is comparable to that produced by a light bulb or a person -- say, around 60 to 150 watts, depending on the fan. Now, whether or not this rate of heat production is consequential depends, of course, on what it is compared with.

If a fan is being used to exchange cool air from outside with warm air from inside, or cool air from near the floor with warm air from near the ceiling, then the heat produced by the fan is relatively negligible. But if the temperature difference between the "cool" and "warm" air is small, the net effect of the fan is merely to blow a thin layer of warm, moist air away from your skin and replace it with other air from the same room. In this case the fan serves no purpose when you leave the room, and if you should certainly turn it off.

I do not know how much heat a candle produces. "

Tim Mooney

http://www.newton.dep.anl.gov/askasci/phy00/phy00393.htm
 

epicfail

Well-Known Member
Yep. Not sure if I think it's a good idea though.
I haven't put that waterblock to use yet because I'm not sure its a good idea also. I do however have a couple vero18s and my pepper cabinet should be ready in a few days. I could run it for a week or so and test it out now that I have parts and space. Maybe I will try and set it up this weekend if I'm not too busy, the problem is I will have to kapton tape them on as I have no way to attach them without puncturing the block. Anyone have any ideas on how to attach them?
 

SupraSPL

Well-Known Member
@AquariusPanta Good question, what I am saying is that the energy consumed by the fan was technically unnecessary because if we use passive cooling we get to enjoy the full efficiency of the LED. But when we use active cooling there is a penalty and I wanted to see how much that would affect the overall efficiency. So I figure it like this:

CXA3070 3000K AB @ 1.4A = 52W @ 42% efficient (21.84 PAR W out of 52 dissipation W)
Fan power consumption at 9V = 1.6W minimum so now:
CXA3070 3000K AB @ 1.4 = 52W + 1.6W = 53.6W
So 21.84 PAR W out of 53.6 dissipation W = 40.7% efficiency (equivalent)

So this demonstrates that actively cooling is like getting 40.7% efficiency out of a 42% efficient COB. BUT if the COB is running a bit cooler with active cooling, it might make up some of that difference.
 

SupraSPL

Well-Known Member
Got a pair of second hand Arctic Freezer Pro 7 Rev2 130W CPU coolers w 92mm fan today to experiment with, $12 each shipped. The CXA3070 fit perfectly on the copper pad (1 3/16" or 30mm) but it would be too small for CXA3590. Not sure about Vero29 I will have to check on that. They have 45 large fins but the heat has to go through 2 extra thermal interfaces before it gets to the fins. The copper pad is not polished very well. 220 grit sand paper improved it quite a bit but it was tricky to flatten because I have to be very careful not to roll the edges too much or I will actually reduce the surface area for the COB.
DSC07700a.jpg


So I went against the grain, trying to remove these machine marks with 220 grit. But I was in a hurry and left it like this for now, smoothed with 600 grit and left the machine marks. A nice improvement from how they came, but I will go all the way flat and then smooth to 1000 grit when I get more time.
DSC07702a.jpg
 
Last edited:

SupraSPL

Well-Known Member
Luckily there is just a bit of copper pad around the CXA so the slightly rolled edges will not reduce surface area contact after all. Next thing to do is test the performance against the 100W Arctic Alpine 11. The Freezer Pro 7 has a lot more fin surface area, but the Alpine 11 has no extra thermal interfaces to deal with.

DSC07703a.jpg
 
Last edited:

SupraSPL

Well-Known Member
Interesting test results. I will post more about it tomorrow but the short version is, the Arctic Alpine 11 seems to have outperformed the Freezer Pro 7 and the round intel cooler with the copper pad and core. The difference at 1.4A was almost negligible, .85% loss in light versus 1.6%. I think the reason a 100W cooler is working better than a 130W cooler is because the Alpine 11 has the COB mounted onto a larger thermal mass and a direct path to the fins. So when cooling relatively small loads, the "capacity" of the cooler does not come into play as much as its "speed" or thermal resistance.

Another way of looking at it, the Alpine can move the heat to the fins more quickly, which really helps reduce Tj in our application (where there is almost no thermal gradient, fins are almost at ambient temp). So as long as the heat does not overwhelm the fin surface area capacity, it performs better than a heatsink with lots of surface area but a greater thermal resistance.

If we were to crank massive amounts of heat through it, the Freezer 7 would probably perform much better than the Alpine 11.

As a final thought, at 1.4A and fully warmed up, we are only losing .85% of the light output compared to Ta20C=Tj 20C! That is with the Arctic Alpine 11 and fan running at 9V. I would say we have no worries about heat and that confirms we would experience almost zero lumen depreciation over the years. Running the fan at 5V, the light loss was only 2.3%. Even with no fan at all and the heatsink too hot to touch, output loss was only about 7%. These COBs are very strong in terms of temp droop performance.

So the next test will be the heatsinkUSA 5.88" performance.
 
Last edited:
Top