Heatsinks for DIY LED lamps

First you calculate the perimeter:

(profile width X 2) + (fin height X 2 X # of fins) + (base height * 2)= Perimeter
Then, Perimeter * fin length = Surface Area.


You can take it a step further and add the surface area of the cut edges:
(profile width * base height * 2) = close enough

Is there any reason people aren't using something like whats in the link below?

http://www.aliexpress.com/item/2pcs-Industrial-aluminum-heatsink-heat-snk140-60-100mm-Super-power-from-cold-tablets-6063-aluminum-Engine/32225059236.html

Buying 6 of them I can get those for $36USD each landed...140x140x60mm with 12mm base on them and at around 1.4kg each, would these passive cool 1 CXB3590 72V @ 50W?

The 140mm X 100mm is 1490 cm² for $30. That works out to about 50cm²/$ which is 2X-3X the price we pay from HeatsinkUSA but may be good for those who cannot get good shipping prices from HeatsinkUSA. They are good about 12 heat W which would be good to passive cool a CXB at 25W.

SupraSPL said:

The 140mm X 100mm is 1490 cm² for $30. That works out to about 50cm²/$ which is 2X-3X the price we pay from HeatsinkUSA but may be good for those who cannot get good shipping prices from HeatsinkUSA. They are good about 12 heat W which would be good to passive cool a CXB at 25W.

Click to expand...

Actually I just checked HeatsinkUSA, and the prices are pretty good even with postage to New Zealand lol.....good to see I've got that option

littlejacob said:

Bonjour
Are you talking about the 1000X165X35mm aluminum radiator from tme? (36.50 €...43 € vat included!)
If it is this one I split this one in 4 part with cxb 3590@1400mA on each...it gets hot so if you let it like it is and don't split it you will need a 140mm fan or 4 x80mm! Plus a 50mm for the driver!
Have a great day ★

Click to expand...

you used this heat sink, LJ? Im buying these, it seems. Planning to mount the same amount of cobs. Is it not enough cm2 to be a passive heat sink? I get the surface area to be 13800cm2, which should be more than enough for these four cobs.

SupraSPL said:

First you calculate the perimeter:

(profile width X 2) + (fin height X 2 X # of fins) + (base height * 2)= Perimeter
Then, Perimeter * fin length = Surface Area.


You can take it a step further and add the surface area of the cut edges:
(profile width * base height * 2) = close enough

Click to expand...

that is great, Supra. I hope to have done it correctly. The thickness of the base doesnt matter anything? I thought otherwise, but I know next to nothing about engineering and mathmatics. I get the heat sinks I want to provide me with 13800cm2, which should be plenty, although the base on its own, is very thin.

Yes I get 13240cm², assuming a 6mm base. Is that correct base thickness? 6mm is a bit thin but I am not sure exactly how that will affect the junction temp in practice. The idea is that thicker base can better handle an uneven heat load (like 50W COBs). It is something I would like to test and compare temp droop results.

It does seem to be a good value at 165 cm²/$

SupraSPL said:

Yes I get 13240cm², assuming a 6mm base. Is that correct base thickness? 6mm is a bit thin but I am not sure exactly how that will affect the junction temp in practice. The idea is that thicker base can better handle an uneven heat load (like 50W COBs). It is something I would like to test and compare temp droop results.

It does seem to be a good value at 165 cm²/$

Click to expand...

Actually 5mm base, sounds too thin for passive cooling ..
@littlejacob : Have you measured temp behind cob and far from cob by any chance ?

Update: new lamps arrived, this time with true cxb 3070 and better drivers. Also, I asked for the lamp bodies anchor point to be slightly pushed forward, for better weight distribution. (I can post pictures if anyone is interested). So far so good: the new design works much better. Less heat, better light. There is no doubt possible about the performance increase between cxa 3070 4000k z4 and cxb 5000k db. Anyone looking for good cob lamps, I recommand this way: order chips and drivers separately to make sure the items are of good quality, and have them shipped to the lamp manufaturers, which pieces stuff together, and ships the finished lamp overseas. Only 1x international transport costs and the possibility to have custom lamps made (chips, drivers type, reflector angle and so on). Rough estimation is around $60 for cxb 3070 per lamp and $50 for cxa 3070.

Price breakup
18 lamp + manufacturing costs
32 cxb 3070
5 active pfc driver
5 shipping/taxes

Depending on where you live and what transporter you use, the last part can be significantly higher. Also, some manufacturer only ask $10 for the lamp, but I did not try any of thoses yet.

SupraSPL said:

Yes I get 13240cm², assuming a 6mm base. Is that correct base thickness? 6mm is a bit thin but I am not sure exactly how that will affect the junction temp in practice. The idea is that thicker base can better handle an uneven heat load (like 50W COBs). It is something I would like to test and compare temp droop results.

It does seem to be a good value at 165 cm²/$

Click to expand...

HI supra

I was working on a spreadsheet to quick calculate the cooling capacity of heat sinks found online but I ran into a few things that made me question some of my assumptions and would really appreciate your help(ill post this spread sheet here for others if there is interest)

so I used the formula "profile width X 2) + (fin height X 2 X # of fins) + (base height * 2)= Perimeter
Then, Perimeter * fin length = Surface Area.


You can take it a step further and add the surface area of the cut edges:
(profile width * base height * 2) = close enough"

you gave earlier to calculate the surface area

then I had the chips total watts and effeciency to calculate heatwatts-then off heatwatts to calculate how many of said chip could fit on that heatsink(using 120cm2 per watt passive/40cm2 per watt active)

But i ran some practice ones and got funny numbers(or maybe they are correct)

I tried this heat sink https://www.fasttech.com/products/1348407

surface area:910cm2
watts cooled passive: 7.5875
watts cooled active:22.7625

ok 7.5watts passive sounded about right to me
then I tried

http://www.digikey.com/product-search/en?pv624=3&FV=fff40012,fff80068&k=heatsink&mnonly=0&newproducts=0&ColumnSort=0&page=1&stock=1&quantity=0&ptm=0&fid=0&pageSize=500

surface area: 312.764
watts cooled passive: 2.606
watts cooled active: 5.787

But the data sheet says it can be used for 11w?

I know that material will play its partetc but those numbers are so far off-what would you estimate the numbers shoudl be for those heatsinks?

also is there a rough way to calculate for surated edges?

and I always thought that heasink wieght(like how thick the fins were etc) would playa part in cooling capability, is that not true?

also while ive got your attention could you give me the formula for calculating radial heatsinsk please?

sorry for all the questions but id like to make a good spreadsheet we can all use easily

alas if this has been done please point me to it

Generally I aim to run LEDs significantly cooler than Tj 85C when they are being used in a grow lamp. I picked up that habit from KNNA and back in those days we used a lot of 630nm reds LEDs that were very sensitive to temp droop.

Now most of us are running temp droop resistant LEDs like blues, whites and deep reds so there would be no significant harm running them at 85C, and many fixtures do design for that. The difference between running the LEDs at Tj 40C versus Tj 85C is a light loss of ~6-7%. Not a huge amount but with the right heatsink you could get consistently 7% more bud with no increase in electrical cost and no increase in heat. That is enough for me to continue recommending that we go big on heatsinks, but it is not required to make a functional grow light.

That data sheet you linked is suggesting a temp rise of 60C when handling 11W of heat, passively and vertically if I understand correctly. That could lead to an LED junction temp well over 100C. The LEDs can handle that but at those temperatures lumen output will depreciate somewhat over time just like HIDs and fluoros

So the word on ganja street is that COBs like to be coooool.... so the cool kids run them better!

True not all heatsink surface area is equally effective. You are correct that heatsink weight plays a part due to fin thickness and base thickness. The shape of the heatsink matters as well, wide short fins are more passive cooling friendly/surface area. For that reason I recommend these profiles as passive cooling friendly:
3.5"
4.23"
4.9"
5.88"
7.28"
8.46"
10"

Vertically oriented fins as in radial heatsinks should be even more passive friendly/surface area. I have never attempted to calculate the surface area of a radial heatsink but I would suggest:
Fin surface area = (fin length * fin width * 2 * number of fins)
+ Cut edge surface area = (2* 3.14 * mounting surface radius squared)
+ Cylinder surface area = (fin length * 2* 3.14 + mounting surface radius)

It would be possible to test the performance of a radial heatsink by monitoring temp droop and comparing the results to known heatsinks. That would allow us to come up with a surface area recommendation for passive cooling using radial heatsinks.

Regarding serrated fins, if you are referring to the heatsinkUSA 4.6" profile, due to its relatively tall fins, closely spaced fins it is active cooling friendly. Honestly I am not sure how much bonus it will get from the serrations?

@heckler73 sent me a .pdf about serrated heat sinks and their efficiency advantages. Perhaps he could drop the link here again for you?

ttystikk said:

@heckler73 sent me a .pdf about serrated heat sinks and their efficiency advantages. Perhaps he could drop the link here again for you?

Click to expand...

The serrated design is for vertical passive applications, which takes advantage of natural convection to equalize the flow. It tries to bridge the extremes represented by continuous designs (the one everyone generally uses) and pin designs (mostly found on MoBos, last I checked). It may not have the same effect in horizontal applications, except perhaps in a forced tube. It may not even work properly in that condition because of possible turbulence.

But that's also my intuition, which doesn't have a great batting average. You never know, it might work there, too. It may only need some minor modification, perhaps imbalanced fin placement or a gradient of serration is better for horizontal ?

I imagine if one needs to keep the flow laminar (i.e. not turbulent), then a lower fan speed would be required.

SupraSPL said:

Generally I aim to run LEDs significantly cooler than Tj 85C when they are being used in a grow lamp. I picked up that habit from KNNA and back in those days we used a lot of 630nm reds LEDs that were very sensitive to temp droop.

Now most of us are running temp droop resistant LEDs like blues, whites and deep reds so there would be no significant harm running them at 85C, and many fixtures do design for that. The difference between running the LEDs at Tj 40C versus Tj 85C is a light loss of ~6-7%. Not a huge amount but with the right heatsink you could get consistently 7% more bud with no increase in electrical cost and no increase in heat. That is enough for me to continue recommending that we go big on heatsinks, but it is not required to make a functional grow light.

That data sheet you linked is suggesting a temp rise of 60C when handling 11W of heat, passively and vertically if I understand correctly. That could lead to an LED junction temp well over 100C. The LEDs can handle that but at those temperatures lumen output will depreciate somewhat over time just like HIDs and fluoros

Click to expand...

ok tnx so thats prob why the number i got are significantly lower

is 120cm2 per watt passive/40cm2 per watt active correct?


does this look right?( excuse the format its a work in progress) the chip is a vero18@350mA

I have pretty well settled on a lamp module to cover a 2'x2' grow. Basically 4 CXB3070s @3000K BB or better bin, mounted on 2ea 23" pieces of the HSAmerica 5.88" profile powered by an HLG-185H C1400B. Since I'm going to use the same fixture for both veg and flower, I'm considering adding some blue stars. Is adding blue stars worth the effort? The driver is maxed out with the 4 CXBs so adding stars is a bit tricky but I believe I have a way to do it. I'll use a switch (simple double pole double throw) to select veg or flower. In flower mode the stars will be idle and the 4 cobs will be in series. In veg mode The driver will power 2 parallel strings, each string will have 2 cobs, in series along with 1/2 of the stars. Without adding any stars the "re-connection" of the cobs would cut their power in half (700ma instead of 1.4A). The stars will see a maximum of 700MA. Potentially I could add about 40 stars & the stars could be a mix of colors & whites. Should I just add cool white or a mix of blue, red and cool white?

Dang, I wish I could edit that, supposed to be:

Vertically oriented fins as in radial heatsinks should be even more passive friendly/surface area. I have never attempted to calculate the surface area of a radial heatsink but I would suggest:

Fin surface area = (fin length * fin width * 2 * number of fins)
+ Cut edge surface area = (2* 3.14 * mounting surface radius squared)
+ Cylinder surface area = (fin length * 2* 3.14 * mounting surface radius)

Jerry Liebler said:

I have pretty well settled on a lamp module to cover a 2'x2' grow. Basically 4 CXB3070s @3000K BB or better bin, mounted on 2ea 23" pieces of the HSAmerica 5.88" profile powered by an HLG-185H C1400B. Since I'm going to use the same fixture for both veg and flower, I'm considering adding some blue stars. Is adding blue stars worth the effort? The driver is maxed out with the 4 CXBs so adding stars is a bit tricky but I believe I have a way to do it. I'll use a switch (simple double pole double throw) to select veg or flower. In flower mode the stars will be idle and the 4 cobs will be in series. In veg mode The driver will power 2 parallel strings, each string will have 2 cobs, in series along with 1/2 of the stars. Without adding any stars the "re-connection" of the cobs would cut their power in half (700ma instead of 1.4A). The stars will see a maximum of 700MA. Potentially I could add about 40 stars & the stars could be a mix of colors & whites. Should I just add cool white or a mix of blue, red and cool white?

Click to expand...

Ok so you're using the same combo I'm using, only you're going to light just 2/3 of the square footage I am; this is going to blast your babies with over 1200PPfd, so there is no way in hell you're going to need or notice the addition of a few watts of 'stars'. So, save your money and brain damage and apply these resources instead to the extra cooling and ventilation your lil girls are gonna need. Just my .02

I just ordered my leds and drivers from Kingbright. I need help on heatsink to choose from heatsink usa. I want to run this passive. The only thing I know for sure is how long I want it to be. 32 inches I'm mounting 4 3590 and mw 185. If it's going to be 32 inches long what width do I need to buy. Thanks and sorry I'm not good at math.