Heatsinks for DIY LED lamps

SupraSPL

Well-Known Member
I was very suspect of radiant heat altering my reading since the Tj notch is so close to the LES! Thank you. I will still check the thermal interface regardless. Maybe I didn't use enough TIM or there is an air bubble or something.

I will also try to insulate the probe so it's not heated from radiation.
Maybe if you block the radiation with aluminum foil, monitor the temp and then click off the light to see how that affects the Tc.
 

churchhaze

Well-Known Member
I'm not sure how I would get a good picture showing how much is falling off. It's a very hard angle to get a picture of.

It's a very little piece of corner on each side sticking off that I doubt makes an appreciable difference in the overall thermal interface. That being said, I still don't like the vero 29 combined with the arctic apline 11. I think the cheap coolermaster 90mm fans are a lot higher quality than the fan built into the arctic alpine 11. The wires in the arctic alpine 11 rip out easily, and the molex connectors tend to get pushed out of the plastic housing easily. I actually bought 4 and only use 2. The other 2 were destroyed by abuse.

Wasn't there another cheap cpu heat sink that performs similarly that supra tested? rosewill something? The corners sticking off is reason alone to look for another option in my opinion, even though I feel it doesn't make a big difference, I'm very OCD.

It'd be awesome. A friend asked me to recommend him some parts for a DIY LED light. I wanted to make sure if I can suggest him V29+Arctic 11.
 
So I am trying to make sure I have the math down when sizing the heat sink, If I have 24" of the 4.85 sink I could run 48 watts passively at about 100 cmsq/w? And if my math is right and I have this figured out then the Arctic 11 2250cmsq/24.245=92.8cmsq/w, assuming I am running at 700mA? Is 92.8cmsq/w to low for passive?
 

churchhaze

Well-Known Member
So I am trying to make sure I have the math down when sizing the heat sink, If I have 24" of the 4.85 sink I could run 48 watts passively at about 100 cmsq/w? And if my math is right and I have this figured out then the Arctic 11 2250cmsq/24.245=92.8cmsq/w, assuming I am running at 700mA? Is 92.8cmsq/w to low for passive?
If really want to do passive, I'd consider 2x bars of 12" profile at 5" length rather than 1x 24"bar of 4.85" profile.

That way you have a better airflow path for natural convection.

(see a few pages ago for a picture of what I mean)
 
If really want to do passive, I'd consider 2x bars of 12" profile at 5" length rather than 1x 24"bar of 4.85" profile.

That way you have a better airflow path for natural convection.

(see a few pages ago for a picture of what I mean)
I will not have that kind of width between the 6 3070 Cobs, even the smaller 4.85" profile doesn't leave me to much room in the middle of the lamp. I am wanting it to passively run about 44 w of supplemental lighting. On second thought it might work if I change the design a bit but is my math correct? And is 100cmsq/w the goal to passively cool?
 
If really want to do passive, I'd consider 2x bars of 12" profile at 5" length rather than 1x 24"bar of 4.85" profile.

That way you have a better airflow path for natural convection.

(see a few pages ago for a picture of what I mean)
I see what you are saying, I went back and read again. I don't see the specifications on the 1st page for that 12" profile.
 

alesh

Well-Known Member
I'm not sure how I would get a good picture showing how much is falling off. It's a very hard angle to get a picture of.

It's a very little piece of corner on each side sticking off that I doubt makes an appreciable difference in the overall thermal interface. That being said, I still don't like the vero 29 combined with the arctic apline 11. I think the cheap coolermaster 90mm fans are a lot higher quality than the fan built into the arctic alpine 11. The wires in the arctic alpine 11 rip out easily, and the molex connectors tend to get pushed out of the plastic housing easily. I actually bought 4 and only use 2. The other 2 were destroyed by abuse.

Wasn't there another cheap cpu heat sink that performs similarly that supra tested? rosewill something? The corners sticking off is reason alone to look for another option in my opinion, even though I feel it doesn't make a big difference, I'm very OCD.
Thanks again. I'll find another cooler. Rosewill is unfortunately unavailable there.
 

SupraSPL

Well-Known Member
I will not have that kind of width between the 6 3070 Cobs, even the smaller 4.85" profile doesn't leave me to much room in the middle of the lamp. I am wanting it to passively run about 44 w of supplemental lighting. On second thought it might work if I change the design a bit but is my math correct? And is 100cmsq/w the goal to passively cool?
Using heatsink profiles with thick base, wide fin spacing and short fins, along with higher efficiency LEDs generally, I am recommending 75W/cm² for passive cooling. The 4.85" profile has relatively close fin spacing, thin base and is relatively expensive /cm² so I would recommend the 4.9". A 13" length would give you 75cm² at 44W. If you need more spread, the 3.5" profile, and for less spread, 5.88".

The 12" profile was not available at the time of the original post and unfortunately I cannot edit it. Maybe RIU can help me with that because it would be nice if we could keep the most relevant up to date info on page 1. Anyway the 12" profile is relatively expensive/cm² and is limited to 12" lengths, but I agree with church that it would be very good at passive cooling

HUSA 846 12.png
 
Last edited:
Man I thought that I understood this, what am I figuring wrong? I calculate 13" of the 4.9" sink (234.78 cm² x 13) = 2817.36 cm²/44 w = 64.03 cm²/w you are saying it is 75 cm² what am I doing wrong? Second question would be if any sink is greater than 75 cm²/w (for passive use) why does the design matter? Is it that one design removes heat faster than another? And thanks for the help, I can't believe how much I have learned in this forum but still feel like I know so little.
 

sforza

Well-Known Member
Has anyone tried attaching multiple COBs to one side of an aluminum or copper plate and attaching a Alpine or Rosewill CPU coolers to the other side of the plate? I would use thermal paste between both the COB base and the CPU cooler. I would line up the CUP cooler so that it would be directly over the COB. So the COB heats up the plate and the CPU cooler cools the plate, not the COB directly? Is there any reason why doing it this way would be a bad idea?

I was thinking of putting four CXA3590s on a 3/8" thick aluminum plate and running four or five CPU coolers. If four, one cooler for each COB, if five one cooler for each COB and an extra one in the middle.

It seems like you would still be putting the same amount of heat into the system and taking the same amount of heat out of the system but plate would absorb and dissipate a bit of the heat and if polished would also act as a reflector.

Thanks for any insight anyone can offer on this plan.
 

alesh

Well-Known Member
Man I thought that I understood this, what am I figuring wrong? I calculate 13" of the 4.9" sink (234.78 cm² x 13) = 2817.36 cm²/44 w = 64.03 cm²/w you are saying it is 75 cm² what am I doing wrong? Second question would be if any sink is greater than 75 cm²/w (for passive use) why does the design matter? Is it that one design removes heat faster than another? And thanks for the help, I can't believe how much I have learned in this forum but still feel like I know so little.
4.9" extrusion has outer perimeter 89.6cm so 13" piece has outer area 2958.1cm^2 making it 67.2cm^2/W.
And yes, there are significant differences in HS passive performance. Best suited for passive cooling are extrusions with thick base and lesser count of thick fins with wide spaces between them. On the other hand, CPU coolers (lots of dense thin fins) would be of terrible performance w/o fan.
 

bicit

Well-Known Member
Supra- Have you had a chance to play around with the 4.85" profile for active cooling? I'm wondering if I can get away with a 30" length and a single 120mm fan to cool four, vero 29's@1.05a.
 

sforza

Well-Known Member
The thermal resistance would be significantly higher than just putting the cobs directly on the heat sinks.
So does higher thermal resistance mean that the cobs would run considerably hotter if attached to a plate between the actively cooled heat sinks rather than directly to the heat sinks? I thought that was the case, since everyone it seems does attach the COBs directly to the heat sinks, but I guess I don't understand why. Could you elaborate a bit to help me understand better?
Thanks for your help.
 

SupraSPL

Well-Known Member
@sforza If you used a thick copper plate, it would quickly spread the heat away from the COB, but it needs surface area to dissipate that heat. So the heat has to pass through the thermal interface between your plate and the aluminum heatsink surface to get to the fins. The quality of your thermal interface would affect how much resistance was created. The resistance could cause a slight thermal gradient, which would lead to a slightly higher junction temp in the COB and therefore more temp droop (light loss)

So in practice if your thermal interface was very good, you might reduce your temp droop from 2% to 1.5% or something along those lines (with a hard running CXA3590). But if the thermal interface was poor for any reason, it could increase your temp droop from 2% to 5% for example. If you are running your CXA3590s soft (~25-50W), you are probably getting less than 1% temp droop, maybe even less than .5%. In that case there are no significant gains to be had by investing in more cooling.

The heatsinks we use are normally a nice 1 pound chunk of aluminum/copper. I have tested temp droop of CXA3070s of different heatsinks at high dissipation levels (100W) to see which perform best. So far the winner has been the large heat pipe style OCZ Vanquisher. Second place the blow through style with copper core, like this, although the mounting surface is too small for CXA3590 or Vero29. The Rosewill RCX-Z1 and RCX-Z200 use copper plates mounted to aluminum and the performance is similar to the Alpine11 which is solid aluminum.

All at 2.47A:
OCZ Vanquisher (large heat pipe style cooler) with CXA3070 3K AB = 2.6%
blow through style Intel with CXA3070 3K Z4 = 3% temp droop
Rosewill RCX-Z1 with CXA3070 3K AB = 4.5%
Alpine 11 with CXA3070 3K AB = 3.5-4.5%

Rosewill RCX-Z1 with Vero 29 3K = 2% temp droop

CXA3590 so far untested
 
4.9" extrusion has outer perimeter 89.6cm so 13" piece has outer area 2958.1cm^2 making it 67.2cm^2/W.
And yes, there are significant differences in HS passive performance. Best suited for passive cooling are extrusions with thick base and lesser count of thick fins with wide spaces between them. On the other hand, CPU coolers (lots of dense thin fins) would be of terrible performance w/o fan.
So you are saying that it is the outer perimeter x length divided by the dissipated watts. I don't understand how you are getting outer perimeter, page 1 shows the "perim" for the 4.9" as 36.40. How are you coming up with the 89.6cm per inch?
 

alesh

Well-Known Member
So you are saying that it is the outer perimeter x length divided by the dissipated watts. I don't understand how you are getting outer perimeter, page 1 shows the "perim" for the 4.9" as 36.40. How are you coming up with the 89.6cm per inch?
I'm not sure which "page 1" you mean but according to this PDF the perimeter of 4.9" is 35.27" which is 89.6cm (1 inch = 2.54 cm). When calculating area you gotta use same units (I'm in Europe and use metric units). You can either convert it to cm or calculate the area in inches and convert to cm^2 afterwards (1 inch^2 = 6.4516 cm^2).
 

Dloomis514

Well-Known Member
I'm not sure which "page 1" you mean but according to this PDF the perimeter of 4.9" is 35.27" which is 89.6cm (1 inch = 2.54 cm). When calculating area you gotta use same units (I'm in Europe and use metric units). You can either convert it to cm or calculate the area in inches and convert to cm^2 afterwards (1 inch^2 = 6.4516 cm^2).
There has been a slight discrepancy between the data on Page 1 of this thread and the 'official' HS-USA data, but the discrepancy is small, in this case 36.4 in^2 vs 35.27. I think the difference was discovered to be a minor math error on Page 1, but don't sue me if i am wrong.
 

alesh

Well-Known Member
There has been a slight discrepancy between the data on Page 1 of this thread and the 'official' HS-USA data, but the discrepancy is small, in this case 36.4 in^2 vs 35.27. I think the difference was discovered to be a minor math error on Page 1, but don't sue me if i am wrong.
I finally understand that you're referring to page 1 of this thread lol. You know, stoners... That's where it came from (and the discrepancy). Poor @SupraSPL had to calculate all the perimeters while they were in that pdf the whole time.
Just a small detail - perimeter is in inches, area in inch^2 (since we multiply inch*inch).
 
Top