Heatsinks for DIY LED lamps

[SupraSPL]

Wow lots of brain power in here, great discussion I am just catching up on You are correct bicit I recommend 110cm²/dissipation W for passive cooling lamps of typical DIY efficiency (~38-40%) or 90-100cm² for lamps of very high efficiency ~48-50%. Those numbers were adapted from KNNAs recommendations and we have decreased the required surface area as LED efficiency has improved over the years. My goal has been to stabilize at Tj 50C. But in retrospect I came up with that goal before we got access to high efficiency whites, which behave much better with a high Tj than the reds do. And the Vero line has just stepped that up even further with amazing thermal performance at high levels of dissipation.

I have been impressed by the cooling performance of the CPU coolers. But if we are willing to accept a slightly higher Tj for the all-white COB lamps, how little power (if any) could we get away with to cool the lamps? Higher Tj should not affect the canopy temps or ventilation speed because ultimately the lamps are creating the same amount of heat as they were before, just moving it away from the COB more slowly. It has got me wondering, would passive cooling at slightly higher Tj be more efficient overall.

 

[SupraSPL]

I am working on testing this further. The first test I did was overbuilt passive cooling, 115cm²/W with CXA3070s running at 49% efficiency. Here are the results:
Ambient temp 22C
Heatsink temp stabilized at 31C with no airflow, convection only
light loss warmed up was 1.86% versus ambient baseline

Then I added a 120mm fan running very softly at 5V (dissipating .463W) consuming .555W.
Heatsink temp stabilized at 24.5C
light loss was down to only .25% versus ambient baseline

That sounds great BUT when you work the numbers we get:
Ambient 49% efficient overall
Passive 48.1%
Active cooled gently 48.3%

So in this case, the addition of active cooling does not seem to be worth the extra complexity and cost. Next test will be at 90cm²/W, a much better test because it is close to how I actually use them and it will give us a better idea what the fan can for us.

 

[AquariusPanta]

I am working on testing this further. The first test I did was overbuilt passive cooling, 115cm²/W with CXA3070s running at 49% efficiency. Here are the results:
Ambient temp 22C
Heatsink temp stabilized at 31C with no airflow, convection only
light loss warmed up was 1.86% versus ambient baseline

Then I added a 120mm fan running very softly at 5V (dissipating .463W) consuming .555W.
Heatsink temp stabilized at 24.5C
light loss was down to only .25% versus ambient baseline

That sounds great BUT when you work the numbers we get:
Ambient 49% efficient overall
Passive 48.1%
Active cooled gently 48.3%

So in this case, the addition of active cooling does not seem to be worth the extra complexity and cost. Next test will be at 90cm²/W, a much better test because it is close to how I actually use them and it will give us a better idea what the fan can for us.

How many COBs were used, at what currents, and over what kind of HeatSink? Dimensions?

I'm struggling to grasp what your working with visually lol.

 

[churchhaze]

Okay, so what exact temperature does that start kicking in?

I think you're making a lot of bad assumptions.

Dynamic frequency scaling (also known as CPU throttling) is a technique in computer architecture whereby the frequency of a microprocessor can be automatically adjusted "on the fly," either to conserve power or to reduce the amount of heat generated by the chip. Dynamic frequency scaling is commonly used in laptops and other mobile devices, where energy comes from a battery and thus is limited. It is also used in quiet computing settings and to decrease energy and cooling costs for lightly loaded machines. Less heat output, in turn, allows the system cooling fans to be throttled down or turned off, reducing noise levels and further decreasing power consumption. It is also used for reducing heat in insufficiently cooled systems when the temperature reaches a certain threshold, such as in poorly cooled overclocked systems.

 

[SupraSPL]

AP, that test was a pair of CXA3070s mounted onto a 10.08" X 6" heatsink, running at 690mA. The next test is the same setup running at 850mA. These heatsinks have a relatively wide space between the fins and the fins are 2.5" tall. The fan is a typical 120mm. This is really not a profile that I recommend because we could probably use more spread and it is not setup of for very efficient use of active cooling, especially compared to something like the 5.88" profile becuase you want the fan to blow air into every channel and you want the channels relatively long.

 

[SupraSPL]

Increased power for this test, dissipating 60.5W, same heatsink, 90cm²/dissipation W

Ambient temp 22C
Heatsink temp stabilized at 33C with no airflow, convection only
light loss was 3.3% versus ambient baseline

Then I added a 120mm fan running softly at 5V (dissipating .463W) consuming .555W.
Heatsink temp stabilized at 25.5C
light loss was 1.32% versus ambient baseline

I increased fan power to 9.35V (dissipating 1.56W) consuming 1.87W
heatsink temp stabilized at 24C, only 2C over ambient
light loss was .88% versus ambient baseline

So when you work the numbers we get:
baseline efficiency 47.5%
Passive cooling 45.9%
Active cooled 5V = 46.5%
Active cooled 9V = 45.7%

As you can see going from 5V to 9V actually decreased the efficiency of the system. So in this case there would be no point in adding active cooling BUT if we reduce the heatsink surface area or increase the power, active cooling could become very helpful. So the next test should be double the power.

 

[AquariusPanta]

Increased power for this test, dissipating 60.5W, same heatsink, 90cm²/dissipation W

Ambient temp 22C
Heatsink temp stabilized at 33C with no airflow, convection only
light loss was 3.3% versus ambient baseline

Then I added a 120mm fan running softly at 5V (dissipating .463W) consuming .555W.
Heatsink temp stabilized at 25.5C
light loss was 1.32% versus ambient baseline

I increased fan power to 9.35V (dissipating 1.56W) consuming 1.87W
heatsink temp stabilized at 24C, only 2C over ambient
light loss was .88% versus ambient baseline

So when you work the numbers we get:
baseline efficiency 47.5%
Passive cooling 45.9%
Active cooled 5V = 46.5%
Active cooled 9V = 45.7%

As you can see going from 5V to 9V actually decreased the efficiency of the system. So in this case there would be no point in adding active cooling BUT if we reduce the heatsink surface area or increase the power, active cooling could become very helpful. So the next test should be double the power.

I'm beginning to see a trend through your test results, an interesting one at the very least. But before I ruin the rising surprise, I will wait until you finish demonstrating what I see to be a linear relationship between the inclusion or absence of the known variables.

@SupraSPL - I'm joyous that your leading the charge on this subject of experimentation. A bunch of us would have been happy slamming down some computer fans on a HeatSink thinking we were on top of things but it's evident that there is much more to it then that.

 

[zep_lover]

@SupraSPL i have a 44 inch piece of the 2.08 inch profile from heatsink usa.i figured it would have just under 105 cm/2 per watt if running 90 watts.did i figure that correctly?

 

[SupraSPL]

44" length would be 3715cm². So with 90W that would be 41.27cm²/W

 

[bicit]

Supra- Do you have any test results for the 5.886 profile?

 

[SupraSPL]

Increased power again, dissipating 108W, same heatsink, now 50cm²/dissipation W

Ambient temp 22.5C
Heatsink temp reached 40C with no airflow, convection only
light loss was 7% versus ambient baseline

Added 120mm fan running softly at 5V (dissipating .463W) consuming .555W.
Heatsink temp stabilized at 28C
light loss was 3.36% versus ambient baseline

I increased fan power to 9.35V (dissipating 1.56W) consuming 1.87W
heatsink temp stabilized at 26.5C
light loss was 2.94% versus ambient baseline

So when you work the numbers we get:
baseline efficiency 42%
Passive cooling 39%
Active cooled 5V = 40.4%
Active cooled 9V = 40.1%

Because of the white COB's surprising performance with heat, the active cooling had less benefit than expected and once again the 5V was actually better than 9V. This test is important so I intend to repeat it with the 5.88" profile next time. I have a 12" length with a 140mmm fan and a pair of CXA3070 ABs mounted.

 

[NapalmD]

(Enter captain caveman) so if I'm understanding this right, you can get away with running 2 cobs at 60.5 watts each passively without a fan, per heat sink?
If it's only 1% less efficient than mounting each one to say an arctic11, I would much rather go the heat sink route even if it's a little more in initial cost.
More streamlined and quieter.
Thanks for running these tests Supra!

 

[Positivity]

Wow..I can't believe that heat sink cooled 100w so well...amazing results!

I was going to test my CPUs coolers versus a standard heatsink for this exact reason. I had a feeling size trumps what the small CPU coolers can do.

I haven't had the same good results as you running low fan speeds. That heatsink must work especially well with cobs. Anytime I slow my fan speeds the heat builds up quickly..but my ventilation is very low cfm

 

[bicit]

Increased power again, dissipating 108W, same heatsink, now 50cm²/dissipation W

Ambient temp 22.5C
Heatsink temp reached 40C with no airflow, convection only
light loss was 7% versus ambient baseline

Added 120mm fan running softly at 5V (dissipating .463W) consuming .555W.
Heatsink temp stabilized at 28C
light loss was 3.36% versus ambient baseline

I increased fan power to 9.35V (dissipating 1.56W) consuming 1.87W
heatsink temp stabilized at 26.5C
light loss was 2.94% versus ambient baseline

So when you work the numbers we get:
baseline efficiency 42%
Passive cooling 39%
Active cooled 5V = 40.4%
Active cooled 9V = 40.1%

Because of the white COB's surprising performance with heat, the active cooling had less benefit than expected and once again the 5V was actually better than 9V. This test is important so I intend to repeat it with the 5.88" profile next time. I have a 12" length with a 140mmm fan and a pair of CXA3070 ABs mounted.

Do you have the capability of running the system on the 10.08 profile at around 180watts, or 30cm^2/w? I'm wondering what the upper limit would be with passive cooling.

Do you know if there would be a difference in efficiency if you used pwm to dim the fan, as opposed to adjusting the voltage?

Thanks for testing this stuff out for us. Saves a lot of trial and error. By the way, how did you use the superscript on this forum?

 

[alesh]

Very interesting discussion going on here. I am missing one - in my opinion - crucial point against passive cooling though. Passive heat sinks cannot be enclosed.
I agree that nowadays, with more efficient thermal designs and more efficient and temperature-resistant dies, one can get away with a little higher Tj. Meaning that passive designs can be smaller than used to be and as efficient as active designs after all.
And the trend is obvious. In the past few years, the heat sink requirements dropped rapidly. And since Cree has tested 303lm/W LED this will hopefully continue.

edit:
2bicit: There shouldn't be much difference between voltage-controlled and PWM-controlled fan. What matters is the actual RPM, not how it's accomplished. There might be some slight differences in energy consumption but I doubt it would be statistically significant.

 

[guod]

Okay, so what exact temperature does that start kicking in?

I think you're making a lot of bad assumptions.

max. core temps for some CPU

 

[churchhaze]

I have a Sandy Bridge. Does that mean it starts throttling at 69-72C? (assuming it even gets that hot)

And you're positive that's what those charts are telling you?

I think the temperatures on those charts are only max ratings, not threshold settings. So what type of bios do I have? What is my CPU frequency set to ? what is the voltage set to? Am I using a laptop? Do I even have speedstep activated?

I had bought a very large aftermarket heatsink to overclock this CPU in 2012, but got negligible frequency gains with sandy bridge, so I put it back to stock. I'm not using the stock cooler.

max. core temps for some CPU

 

[zep_lover]

44" length would be 3715cm². So with 90W that would be 41.27cm²/W
View attachment 3304283

44 inches is 111.76 cm.111.76 x 84.43=9435.89 cm 2
that is the formula right?

 

[Dloomis514]

I have a Sandy Bridge. Does that mean it starts throttling at 69-72C? (assuming it even gets that hot)

And you're positive that's what those charts are telling you?

I think the temperatures on those charts are only max ratings, not threshold settings. So what type of bios do I have? What is my CPU frequency set to ? what is the voltage set to? Am I using a laptop? Do I even have speedstep activated?

I had bought a very large aftermarket heatsink to overclock this CPU in 2012, but got negligible frequency gains with sandy bridge, so I put it back to stock. I'm not using the stock cooler.

The stock cooler will work great for almost every COB, look up the heat removal specs, expressed at TDW i think, usually up around 90 watts

I have a Sandy Bridge. Does that mean it starts throttling at 69-72C? (assuming it even gets that hot)

And you're positive that's what those charts are telling you?

I think the temperatures on those charts are only max ratings, not threshold settings. So what type of bios do I have? What is my CPU frequency set to ? what is the voltage set to? Am I using a laptop? Do I even have speedstep activated?

I had bought a very large aftermarket heatsink to overclock this CPU in 2012, but got negligible frequency gains with sandy bridge, so I put it back to stock. I'm not using the stock cooler.

CPU stock coolers are rated at TDW, usually 90 watts for power (heat) displacement is acheivable

 

[Dloomis514]

44 inches is 111.76 cm.111.76 x 84.43=9435.89 cm 2
that is the formula right?

Google '44 inches', yes you are right.