Heatsinks for DIY LED lamps

SupraSPL

Well-Known Member
So I put the 45cm²/W setup in the tent and let it stabilize. I hung it in a area that I thought was getting the least airflow. Ambient is 25.5C. Heatsink temps are 31-33C, which corresponds to about 5% temp droop, really not too bad for passive cooling.

Then I hung it in an area with good airflow and temps were 30.5-32.5C.

In the same tent there are 10.08" modules running at about 98cm²/W passive cooled. The ones getting the least airflow are at about 31C and the ones getting the best airflow are at 28C. I think that shows how much more effective the 8.46" profile is for passive cooling that the tall fins of the 10.08" in a grow space.

There is an active cooled 5.88" heatsink that is at 26-27C.

Overall, it is good news, it looks like we can use less heatsink than 90cm²/W for passive cooling if there is some circulation airflow and using the wide fins spacing with short fins. Even better if it is for side lighting and the fins are running vertically.
 
Last edited:

epicfail

Well-Known Member
The 10.08 is one of HSUSAs worst designs for passive cooling, the fins are too tall and too close together. It is however a serious chunk of aluminum and will work great with only the slightest amount of air movement. The 10.000" and the new 12" profile they have listed on their frontpage would be great for passive cooling of LEDs.

A test I am interested in seeing is how well the cheaper 3.5" or 4.85" profile work with active cooling. Specifically could they be used as light bars with a cob every 12" or so and fan on them? Maybe something similar to how mine are (and others) built but swapping out the more expensive 5.886" profile. Preferably the 4.85" would be used as it is less than $1 an inch and that would make for significant savings. I think that profile might actually be great for cooling these high efficiency COBs with a 120mm fan and a cover to force the air the full length of the fins. I was going to order a 24" piece for $24 so I could test this out but its going to be $38 to get it here so this become too expensive for just testing purposes alone, so I guess now I wait and hope someone else tries its out first.
 

Bueno Time

Well-Known Member
I was wrong on my wiring idea for the 35x Vero 10s and now that I thought that through and figured that out, rather than wiring the COBs across different heatsink bars I changed the plan from 5x 32" bars with 7x Vero 10s per bar to 7x 20" bars with 5x Vero 10s per bar. Still one 50mm fan per bar in the center with covered fins forcing air down the length of the sink. The overall length of 2.08" profile is reduced 20" and max watt/inch went up to ~1.37 from ~1.19 (probably wont even run it quite maxed out but maybe). There would be 7 fans vs 5 though in the newest revision, each cooling a 20" long bar vs 32" long so that may well increase the cooling effect of the fans enough to make up for the loss of heatsink length overall.

Im expecting an increase of about 10C above ambient in tent temperature with the active cooling. Kind of hard to tell though until you build it and actually put it into action in the grow space.

Heres an idea of the proposed plan for anyone interested. ~27.4w per 20" length of 2.08" profile with 1 50mm fan per bar.

TOP



BOTTOM

 
Last edited:

SupraSPL

Well-Known Member
The 10.08 is one of HSUSAs worst designs for passive cooling, the fins are too tall and too close together. It is however a serious chunk of aluminum and will work great with only the slightest amount of air movement. The 10.000" and the new 12" profile they have listed on their frontpage would be great for passive cooling of LEDs.

A test I am interested in seeing is how well the cheaper 3.5" or 4.85" profile work with active cooling. Specifically could they be used as light bars with a cob every 12" or so and fan on them? Maybe something similar to how mine are (and others) built but swapping out the more expensive 5.886" profile. Preferably the 4.85" would be used as it is less than $1 an inch and that would make for significant savings. I think that profile might actually be great for cooling these high efficiency COBs with a 120mm fan and a cover to force the air the full length of the fins. I was going to order a 24" piece for $24 so I could test this out but its going to be $38 to get it here so this become too expensive for just testing purposes alone, so I guess now I wait and hope someone else tries its out first.
I am with you EF. The only reason I am paying the 10.08" any attention is because that is what I have on hand. They were great in the old days when you needed a ton of cooling to pack the light in. I figured I could overwhelm the heat by using lots of surface area and relying on the circulation fan and that does seem to work, although not very cost efficient. It is nice to see how much better the 8.46" works but that profile is not the most ideal either. The way I see it, passive cooled should have the fins cut the short way and active cooled should have the fins as long as possible. So you could use the 12" profile cut short for passive cooling and the short, low, wide spaced fins would be great for convection. Same goes for the 10". But if you need them spaced further apart, I think the 3.5", 4.23" and 5.88" is a great option.

The 5.88" has the widest fin spacing. The 4.85 is very cheap but it has a much thinner base. Not sure how much that will affect its cooling ability but I will order a test piece to find out.
 

alesh

Well-Known Member
Was gonna ask if you could have an external fan hit it and check again. Off axis, or even an oscillator. Doesn't take much to pull much of that heat away.

Just to confirm the math from before. In your example above, you said 2 x 3070 @ 1.4a. So about 104 watts of dissipation @ 42% efficiency means about 60 watts of heat.

The 8.46" model has a c/w/3" of 1.10. I come up with a 12" length based on the 45cm2/w, correct?

So to plug in the numbers.

104w total x 58% loss as heat = 60w of heat to dissipate.
60w x 1.10 c/w/3" = 66 degrees per 3 inches of sink
66 / 4 = 16.5 degrees C above ambient.

Sound like you did better than that, unless ambient was rather low.
I think there's something wrong in the calculations. You have to calculate/estimate total thermal resistance and then do the math. Here's a nice table for thermal resistance estimation.
hs_correction.jpg
Rth = 1.10 °C/W/3" * 0.54 = 0.594 °C/W
Trise = P*Rth = 60W * 0.594°C/W = 35.64°C
 

alesh

Well-Known Member
So would that mean I should be seeing a max heatsink temp of 35.64+22=77.65C?
No, 35.64°C + 22°C = 57.64 °C. Guess it was a typo. If so, then the answer is yes.
At least for a device which produces 60 watts of heat placed on 12" piece of the heat sink (Rth=1.1°C/W/3") with ambient temperature 22°C.It also assumes that the fins on the HS are positioned vertically. Completely passive, but free air convection is not restricted.
It looks pretty high, but it should also mean that even completely passive, the COBs are running safely under operating limits and no damage will be done. It might even handle the third COB, especially in a grow room where some moving air is expected.
 

getsoutalive

Well-Known Member
@alesh, Can you tell us where that conversion table is sourced from? Not saying it is incorrect just yet, but since the numbers seem to agree with our real world measurements much more closely than with the conversion taken into account, I am a bit suspect. The c/w/3 for all HUSA models is listed in the FAQ and there is no mention of any conversion necessary there.

Remember that we are talking about heat rise in a dead air space, so these are worst case scenarios we are testing here. We are using these calculations to ensure that we do not destroy the chips with heat in case of fan failures. With a small amount of air moving, the temps drop quickly, as @SupraSPL demonstrated above.

My personal choice for 40-50w use is this.....with a TR of 1.02 c/w and a surface area of 1670 cm2
http://www.led-heatsink.com/seepricesandbuy.php?cateid=655

And for pushing up towards 100w one more like this....with a TR of 0.67c/w and a surface area of 3194 cm2
http://www.led-heatsink.com/seepricesandbuy.php?cateid=1277

Just noticed a large price drop on these also....
http://www.digikey.com/product-detail/en/19754-M-AB/345-1123-ND/3175829 with a TR of 0.68c/w. Not sure on the surface area
 

alesh

Well-Known Member
@alesh, Can you tell us where that conversion table is sourced from? Not saying it is incorrect just yet, but since the numbers seem to agree with our real world measurements much more closely than with the conversion taken into account, I am a bit suspect. The c/w/3 for all HUSA models is listed in the FAQ and there is no mention of any conversion necessary there.

Remember that we are talking about heat rise in a dead air space, so these are worst case scenarios we are testing here. We are using these calculations to ensure that we do not destroy the chips with heat in case of fan failures. With a small amount of air moving, the temps drop quickly, as @SupraSPL demonstrated above.

My personal choice for 40-50w use is this.....with a TR of 1.02 c/w and a surface area of 1670 cm2
http://www.led-heatsink.com/seepricesandbuy.php?cateid=655

And for pushing up towards 100w one more like this....with a TR of 0.67c/w and a surface area of 3194 cm2
http://www.led-heatsink.com/seepricesandbuy.php?cateid=1277

Just noticed a large price drop on these also....
http://www.digikey.com/product-detail/en/19754-M-AB/345-1123-ND/3175829 with a TR of 0.68c/w. Not sure on the surface area
The table is from here. It does calculate with only one source in the center so for 2 COBs, the factors should be slightly increased. But the point is, 12" piece of a HS will not have 4 times lower Rth than 3" piece. If you assume so, you're calculating senseless numbers. I've meant no offense, I am simply pointing out an error. Another info here.

And that was my another point, to calculate worst case scenario. As you can see, if all your fans die your LEDs will be OK.
 

getsoutalive

Well-Known Member
The table is from here. It does calculate with only one source in the center so for 2 COBs, the factors should be slightly increased. But the point is, 12" piece of a HS will not have 4 times lower Rth than 3" piece. If you assume so, you're calculating senseless numbers. I've meant no offense, I am simply pointing out an error. Another info here.

And that was my another point, to calculate worst case scenario. As you can see, if all your fans die your LEDs will be OK.
Thank you. Obviously trying to get this correct, no offense taken. All helpful info is appreciated.

I would note that from the page linked....

The published extrusion data shows natural convection performance for a three inch section with a centrally located point source heat load. Because the heat load is assumed to be at a point rather than uniformly distributed, thermal resistance does not change linearly with length. (The ends of a very long extrusion would be cooler than the center and therefore the transfer of heat to the surrounding air is little, if any.) It is therefore necessary to apply a correction factor to published data for extrusion lengths shorter or longer than three inches. The corrected thermal resistance for different lengths of extrusion is obtained by multiplying published °C/W/3-in data by the appropriate factor from the following table:

The table provided assumes a single point source. So this may be more complex than the simple conversion if we are talking about multiple chips spread over the face. Which is why @SupraSPL's numbers were much lower than suggested by the table. Not that with cobs, we have an even distribution of heat, but the table is correcting for temp deltas based upon a single heat source in the center. This is assuming that it was a single 104 watt chip on that 12" slab and so the temp gradient from the center to the edges would be higher than having the two 52 watts chips straddle the centerline.

Again, as I prefer single chips fixtures and star shaped sinks, I have not had to deal with this previously. Thanks again.
 

salmonetin

Well-Known Member
Passive Cooling

Passive Cooling Systems are based on the fact that Hot Air Moves Upwards, thus Creating Airflow along the Surfaces.
This is called Natural Convection.
There are many standard heat sinks available, but it is also possible to design your own heat sink.
In general, a Passive Cooling Solution requires a Larger Heat Sink Than an Active Cooling Solution.

Some Additional Design Guidelines for Passive Cooling include:

Limit the Number of Thermal Interfaces in the Thermal Path from Module to Ambient.

Thick Fins Conduct Heat Better Than Thin Fins.

• Large Spacing Between Fins is Better Than Small Spacing Between Fins.

Make Cooling Surfaces More Effective by Using Proper Conductive Materials, Appropriate Thickness and Correct Fin Orientation.

Thermal Radiation Plays a Significant Role => Anodized or Powder-Coated Surfaces are Preferable to Blank Surfaces

http://www.lighting.philips.co.uk/pwc_li/gb_en/subsites/oem/download/fortimo_led_dlm_system/gen-5/DLM-Gen5-Design-in-guide.pdf

saludos
 
Last edited:

SupraSPL

Well-Known Member
No, 35.64°C + 22°C = 57.64 °C. Guess it was a typo. If so, then the answer is yes.
At least for a device which produces 60 watts of heat placed on 12" piece of the heat sink (Rth=1.1°C/W/3") with ambient temperature 22°C.It also assumes that the fins on the HS are positioned vertically. Completely passive, but free air convection is not restricted.
It looks pretty high, but it should also mean that even completely passive, the COBs are running safely under operating limits and no damage will be done. It might even handle the third COB, especially in a grow room where some moving air is expected.
Yes quite a typo, burning the candle at both ends I think. Anyway, the highest surface temp I was able to measure was 38C and that was the metal directly behind the COB and the hottest spot I could find.

The fins were oriented horizontally. The air in the room was completely still and I allowed a very long time to reach thermal stability, with several consecutive measurements the same. I was using a very accurate and precise infrared baby food thermometer so there is a high degree of confidence in the measurment. The reason for the much lower temp must be because of the large size of the thermal interface (CXA3070 X 2) and as @getsoutalive pointed out, the heat load was spread to two points as well.
 

churchhaze

Well-Known Member
Another thing to keep in mind for passive cooling is to prefer a higher fin count rather than longer fins. (given the same total surface area)

If you want a 12" x 5" platform, consider 5 inches of the the 12" profile rather than 12 inches of the 4.90" profile.

This gives you more air intakes for convection to draw air through. Passive cooling losses effectiveness with longer fin channels.
 
Last edited:

salmonetin

Well-Known Member


thanks churchhaze ...Another thing to keep in mind for passive cooling is to prefer a higher fin count (center pic) rather than longer fins.(right pic) (given the same total surface area)...This gives you more air intakes for convection to draw air through. Passive cooling losses effectiveness with longer fin channels....;)

...for my POV ...Correct Fin Orientation for sidelight too...;)

saludos
 
Last edited:
With all the new testing is it possible to have 3 vero18 running at 700 mA on an 18 inch 4.885 heatsink from heatsink USA passively cooled. I think with all the calculations I came up with 13 c above ambient does that sound right?
 

happy75

Well-Known Member
Ya it has crossed my mind that maybe it was something else that caused it. Maybe I hooked up the COB to the HLG-180H-C1050B while the switch was on and it got damaged without my noticing. Also, the actual Tc may have been as high as 85C, but still should have been OK with that heat at 1050mA.
I had sometimes my cobs running without any active cooling. I have checked every led and I guess that only 4 were broken, and 4 were only shattered on top but still functioning. I have no clue why these broke down and the others didn't. They were all from the same batch. I did not check all drivers, but maybe that was the problem.
 

ballist

Well-Known Member
does anyone know where to buy uncut lengths of extruded aluminum in EU? From memory back when I was in au land it was much cheaper to buy uncut extruded from a wholesaler than buying cut lengths in retail. I think it was 2m standard length.
 
Top