Heatsinks for DIY LED lamps

@SupraSPL

Well-explained SuperSupra

I plan on laying down some PK inbetween my Arctics and Veros tomorrow. I'll try out your advice.

Great Supra thx for the hint !

I'll have to put my hands on some paste then, heatsinks are on the way and I hope I can proceed with pannel construction this coming WE ... Can't wait any longer now

Here are some carbon pads that are 20W/m.K , but they are expensive.....although 150mmx150mm
http://www.newark.com/t-global/t62-150-050-100g/graphite-pad-0-5mm/dp/54M3594

Good deal on fans, 120mm for $2 each

http://www.newegg.com/Product/Produ...MC-012315-Latest-_-CaseFans-_-35103052-S2A10B

I grabbed a four pack of Xigmatek 120mm fans for same price with free shipping. They have a brushless motor and fluid circulative bearings. Tracking says COBs will arrive tomorrow. Heatsinks and drivers shouldnt be far behind. Almost time to build.

SupraSPL said:

Good deal on fans, 120mm for $2 each

http://www.newegg.com/Product/Produ...MC-012315-Latest-_-CaseFans-_-35103052-S2A10B

Click to expand...

I just noticed that this large passive heatsink has come down in price significantly.....

http://www.digikey.com/product-detail/en/19754-M-AB/345-1123-ND/3175829

$20 ea. down from $28+

It can easily keep a 3070 or V29 cool at 1.4a and more.

@Positivity I didnt want to interfere with @stardustsailor thread. I dug out the 2.4A driver and here is the data you requested. All the tests were done at 12", pulsed at 23C.

For those who don't know, we cannot directly compare COBs with different color temps because it can significantly affect the luxmeter, but we can use it a general guide and it is good data to compare the Z2 Z4 and AB. The Vero 29 gets slightly penalized because it has more in the red/deep red and the generic COB numbers are getting a slight bonus because it has a green/yellow emphasis.

Generic 100W COB warm white
322.32 lux/W

CXA3070 3K Z2
378 lux/W

CXA3070 3K Z4
388.5 lux/W

CXA3070 3K Z4 test #2
386.4 lux/W

CXA3070 3K AB1
406.6 lux/W

CXA3070 3K AB2
411.4 lux/W

CXA3070 3K AB3
423.6 lux/W

CXA3070 3K AB3 test #2
423.4 lux/W

CXA3070 3K AB4
423.8 lux/W

CXA3070 3K AB4 test #2
428.17 lux/W

CXA3070 3K AB5
408.83 lux/W

CXA3070 3K AB6
410.3 lux/W

CXA3070 3K AB7
426.46 lux/W

Vero29 3K #1
419 lux/W

Vero29 3K #1 test #2
420.2 lux/W

Vero29 3K #2
427 lux/W

Vero29 3K #2 test #2
425.2 lux/W

And here are some temp droop figures. Same configuration, all running at about 2.45A. Clearly the Vero package flows and/or stands up to heat better, but not a large difference and not a fair comparison unless it was the CXA3590. We have more evidence that sanding and upgrading paste made no difference, maybe even decreased performance?

No significant difference in the cooling performance of the Rosewill vs the Alpine11. Fan power consumption was not taken into consideration for these figures, although if I increased fan speed any higher it would hurt efficiency overall, especially for the Rosewill.

These COB are running hard, 2.45A = 95-105W of dissipation. I was impressed by the performance of both COBs and both heatsinks.

Rosewill RCX-Z1
-----------------------

CXA3070 3K AB6 mounted on Rosewill RCX-Z1, stock
pulsed 410.3 lux/W
(5V fan = .72W) 391.3 lux/W
4.6% temp droop

Vero29 3K #1 - mounted on Rosewill RCX-Z1, stock
pulsed 420.2 lux/W
(5V fan = .72W) 411.3 lux/W
2.1% temp droop

Vero29 3K #2 - mounted on Rosewill RCX-Z1, sanded + PK3 paste
pulsed 425.2 lux/W
(5V fan = .72W) 415.2 lux/W
2.35% temp droop


Alpine 11
-------------

CXA3070 3K AB3 mounted on Alpine 11, sanded + PK3 paste
pulsed 423.4 lux/W
(7.5V fan = 1.02W) 407.5 lux/W
3.75% temp droop

CXA3070 3K AB7 mounted on Alpine 11, sanded + PK3 paste
pulsed 426.46 lux/W
(7.5V fan = 1.02W) 407.5 lux/W
4.45% temp droop

CXA3070 3K AB4 mounted on Alpine 11, stock
pulsed 428.17 lux/W
(7.5V fan = 1.02W) 414.4 lux/W
3.2% temp droop

And the CXA3070 3K Z4, running at 2.47A, mounted on Intel F09A-12B1S1. Nice heatsink, same same weight as the others, bout a pound with a thick copper core that is very flat without sanding.

pulsed 383.04 lux/W
(5V fan = 1.1W) 371.54 lux/W
3% temp droop

Is it almost time to bring in the CXA3590s for comparison and are they hard to come by?

As you already mentioned @SupraSPL , we cannot directly compare COBs of different CCTs. But we cannot directly compare different COB with a lux meter, too, as they have different LES and spatial distribution. This could affect lux meter readings.

Coolermaster stuff is hit or miss. Usually it's very good quality, especially their somewhat newer line of mechanical keyboards (they are costar boards, like filco... the BEST quality), but my recent CPU heatsink was CM and had 2 of those 120mm fans, and they had terrible bearings. They both started grinding against the plastic. Both fans failed very quickly. Their more expensive 90mm fans, however, have been great quality and the bearings seem well designed for the purpose. I think they have a few dud 120mm fans products.

I'd test those fans before designing an entire rig around them. If those are the same ones my CPU sink had, the bearings are utter shit.

Edit: This was the product http://www.newegg.com/Product/Product.aspx?Item=N82E16835103089

Second Edit: The propeller blades on the design you posted look a lot higher quality than the ones on this so I think this particular dud was just a one off failure.

SupraSPL said:

Good deal on fans, 120mm for $2 each

http://www.newegg.com/Product/Produ...MC-012315-Latest-_-CaseFans-_-35103052-S2A10B

Click to expand...

SupraSPL said:

This test is to compare the Arctic Alpine 11 ($10) versus the Rosewill RCX-Z1 ($13). The Rosewill received bad reviews, probably due to its lack of capacity (surface area 1525cm²). This means it would perform poorly under large loads of heat. But for our purposes, it has a nice flat copper pad, a thick base plate and thick fins. It has a high RPM so I didnt bother testing it at 9V. In summary, the Rosewill very slightly outperforms the Alpine 11, and that is using the stock paste with no surface prep for both.

I added the Arctic Freezer 7 Pro for a direct comparison.

Tested using CXA3070 AB bin running at 1.45A. Ambient temp was 23C

---------------------------------
Stock Alpine 11 (MX4 paste)
fan 5V - 2.9% temp droop
stable heatsink temp 28.5C
fan .555W

fan 9V - 1.87% temp droop
stable heatsink temp 26C
fan 1.87W
----------------------------------
Stock Rosewill RCX-Z1
fan 5V - 2.11% temp droop
stable heatsink temp 28C
fan .72W
-----------------------------------
Arctic Freezer 7 Pro
fan 9V - 1.33% temp droop
stable heatsink temp (25C fins, 29C copper pad)
fan 1.63W

fan 5V - 2.43% temp droop
stable heatsink temp (28.5C fins, 31C copper pad)
fan .504W
-----------------------------------

Baseline 42%
Alpine 11 @5V - 40.36%
Alpine 11 @9V - 39.8%
Rosewill @5V - 40.56%
Freezer 7 @5V - 40.6%
Freezer 7 @9V - 40.2%

Click to expand...

So in other words scraping off the stock paste and using arctic silver epoxy was a dumb thing for me to do? Lesson learned

I

tightpockt said:

So in other words scraping off the stock paste and using arctic silver epoxy was a dumb thing for me to do? Lesson learned

Click to expand...

I patented that dumb move

tightpockt said:

So in other words scraping off the stock paste and using arctic silver epoxy was a dumb thing for me to do? Lesson learned

Click to expand...

Not at all, the Arctic more than likely is double the thermal conductance of the stock paste......thermal impedance is probably also decreased....

not a terrible upgrade by any means....

The Alpine 11 uses Arctic MX4 paste, thermal conductivity 8.5 W/(mK) so no worries either way. Same goes for the Rosewill RCX-Z1. The printed on paste is large enough for the Vero 29 or CXA3590 and it performed just as well as the PK3, maybe even better although hard to say at this point with an unknown margin of error. Ultimately this is great news for us DIYers, much easier to slap a build together with ease

alesh said:

As you already mentioned @SupraSPL , we cannot directly compare COBs of different CCTs. But we cannot directly compare different COB with a lux meter, too, as they have different LES and spatial distribution. This could affect lux meter readings.

Click to expand...

I agree, we get precision but not accuracy in this case. Different CCT, LES size and different output patterns. The CXA3070 is 115° and the Vero29 is 120°. But still this data does give some general guidance, especially comparing the different bins and the temp droop data is solid.

SupraSPL said:

I agree, we get precision but not accuracy in this case. Different CCT, LES size and different output patterns. The CXA3070 is 115° and the Vero29 is 120°. But still this data does give some general guidance, especially comparing the different bins and the temp droop data is solid.

Click to expand...

It's very useful data, no doubt about it. I'd be awesome if someone there had an integrating sphere.

Eraserhead has said in the past that someone could always ship him something for them to run in his sphere. Email A51 and see if the offer still stands.

Hey Supra I have someone asking advice in my thread and was wondering if the Alpine 11s will handle (effectively cool) a Vero 29 at 2100mA per? Also what voltage should be ran for the fans at that current?

Figured you would be the guy to ask since you might have tested Vero 29 at high(er) currents on the Alpine coolers.