Understanding Heatsink Data sheets

Hi everyone!

we all love to keep our cobs cool, and we all love to save power, obviously passive cooling is interesting.

and we all love to save a buck when we can so we search the inter webs high and low for great heatsinks, everywhere from https://cobkits.com ( @CobKits ) https://northerngrowlights.com (@robincnn ) , cutter (@welight )

sometimes I look at data sheets of heatsinks and it makes my head swim a bit, perhaps you can help me?


thanks to @SupraSPL and others we use the simple 40cm2 per heat watt for active cooling and 110cm2 per heat watt for passive cooling in order to keep our cobs at around 50c.

let me explain my question with an example:

recently those citizen cobs have tickled my pickle, so cost effective and while they don't get crazy efficient at low currents they don't get that much worse at higher currents, if you are ok with efficiency in the 50% (like a cxb3070@1400mA) range it's a really nice option.

Im looking at running some of those sexy CLU048-1212
lil bad boys at 700mA...

@JorgeGonzales suggests efficiency of about 52%. for easy maths and to be well within the target lets call it 50%

ok so at 700mA they pull 23.7W at circa 50% efficient lets call it 12w of heat( 24x 0.5)

so I'd need a surface area of 1320cm2 (132000mm2) to passively cool each cob(12x110)

now i start looking at cutter's website as they have some nice smaller pin and star sinks, we open the data sheets for the heatsinks and suddenly feel sick....it can't be?!!?!?

the heatsink is a pin style but the data sheet suggests a surface area of only 43495mm2( target is 132000mm2) wow, not even close mmmm

how about this heatsink ? its a star style so unlike the pinned, which are better in a tilted position, should be good for a vertical application, plus its low profile and headroom will expensive in my cab.


however again the data sheet tells me it's only got a surface area of 58993mm2 (this is the 30mm high one) but the 80mm high one ( model 8680) has a surface area of 150311mm2 which meets our target.

But hang on, thats a half kilo (pound) heatsink thats like 9cm wide and 8cm tall (apron 4 inch x 4inch)... seems like a lot of sink for 12w of heat....and a cob driven at 24w... mmmm

ok lets see what some geek dude says about it



ok so I follow his fancy maths and get

If=700mA
Vf=33.9V
Pe=23.7W
Eff= 50%
Pd=11.85W
Tc=50c (we want to keep cobs cool)
Ta= 28c (thats about how hot my gardens under cob have been in the past)
dT= 22c
Rth= 1.857- 0.2 for TIM= 1.65

so anything with a thermal resistance value less than 1.65 should be sweet?

so the data sheet we looked at earlier we see that the 2nd model(8650) has a thermal resistance of 1.5, and should be good but surface area is only 95520mm2, about 2/3 of what our target was.... head starts exploding

pin fins are a lot more efficient than extruded bars, i dont think the 120 cm^2/W is what you use

maybe try this

http://www.myheatsinks.com/calculate/round-pin-heat-sink-calculator/

i do think that 50-55% is in the ballpark for 1212@ 23W

Good radials are just as good as pins, as I get to constantly point out, and it's best to rely on data sheets from reliable manufacturers star heatsinks or empirical testing from vendors or the few data sheets out there when sizing a pin heatsink. They tend to come from China without manufacturer data sheets and are rated in "watts" which is a useless number without the corresponding temperature rise, which is usually 50 or 60C above ambient.

SupraSPLs rules of thumb work for bare extruded aluminum heatsinks, but even then it might be better to look at the actual C/W/3" and do the math, because the further heatsink material gets from the heat source, the less efficient it cools.

All of which is to say it's complicated and sometimes just throwing a cob on, measuring, and going from there is best. Pin heatsinks have been extensively tested and if you stick to the same geometries as vendors have tested here you are usually ok.

Browse Mechatronix and you will see that some of their radial/star heatsinks are actually better than equivalent weight pin heatsinks, and some geometries are much, much worse.

What is the actual heat sink you are looking at?

Also, I have to point out that a common mistake I've seen around here is "all you need is the C/W number", but that number changes dramatically based on how many watts of heat is being dissipated, so it's important to know at what temperature it's being quoted at. Basically, heatsinks works better when they are hot.

Mechatronix usually has nice charts, so hopefully thats the cheap one cutter is selling, here is an example:


You can see that the 9980, which I'm using, is 1.5C/W at 10W dissipated, and 1.0C/W at 50W dissipated. If a manufacturer calls that a 1.0C/W heatsink, you might be unpleasantly surprised when your temps are much higher than expected.

@JorgeGonzales

maybe smh like this

http://www.heatsinkled.com/EtraLED-11020-Modular-Passive-LED-Star-Heat-Sink-Φ110mm-p09377407.html


I looked at the change in c/w it's 1.5 @ Pd35

1.9 @ Pd7

am i correct in assuming in this case a C/W of 1.65 or less( less is better right?) will be fine for my project?

thanks

@weed-whacker I think at 700ma you'd be around 50% efficient, according to Citizen, so 12W is probably a good estimate of dissipated heat. But remember you have to add up the cob, TIM, and heatsink thermal resistance together. Apologies if you mentioned that, I only saw you mention the TIM. The 1212's thermal resistance is about .34C/W.

So with 28C ambient, you want to keep it under ~1.83C/W total to stay below 50C, so subtract .54C, and you are looking at around ~1.3C/W.

Remember this is a bit of a feedback loop with the ambient temperature though. So if you are generating less heat your temps won't be as high which means your cobs will run a little more efficiently. So 28C might be conservative.

Either way, you are on the the right track.

Actually, I just saw the video you linked was just for 85C case not junction temps (Tc vs Tj). That difference is the thermal resistance of your COB. It's little things like that that makes comparing between manufacturers tough, and have caused me to lose it a little trying to explain it around here, but you seem like a smart cat. This is why comparing Tc vs Tc is fairest between cobs, because it takes the cob thermal resistance and cooling requirements into consideration automatically.

God I'm having flashbacks to trying to explain this in the Vero thread.

JorgeGonzales said:

@weed-whacker I think at 700ma you'd be around 50% efficient, according to Citizen, so 12W is probably a good estimate of dissipated heat. But remember you have to add up the cob, TIM, and heatsink thermal resistance together. Apologies if you mentioned that, I only saw you mention the TIM. The 1212's thermal resistance is about .34C/W.

So with 28C ambient, you want to keep it under ~1.83C/W total to stay below 50C, so subtract .54C, and you are looking at around ~1.3C/W.

Remember this is a bit of a feedback loop with the ambient temperature though. So if you are generating less heat your temps won't be as high which means your cobs will run a little more efficiently. So 28C might be conservative.

Either way, you are on the the right track.

Actually, I just saw the video you linked was just for 85C case not junction temps (Tc vs Tj). That difference is the thermal resistance of your COB. It's little things like that that makes comparing between manufacturers tough, and have caused me to lose it a little trying to explain it around here, but you seem like a smart cat. This is why comparing Tc vs Tc is fairest between cobs, because it takes the cob thermal resistance and cooling requirements into consideration automatically.

God I'm having flashbacks to trying to explain this in the Vero thread.

Click to expand...

ok cool, good point on the Tc vs Tj, i didn't catch that

ill search through the vero thread, cheers

Well I am not such a great calculating monster but at this page you can see what this company advises for the 1212. Just to get the idea maybe?
https://www.endrich.com/en/light_solutions-accessories-heatsinks?page=3

So at 1080 mA you will need a cpl-8050 passive that has a thermalpower of 1.45 K/W at 37.4 watt. Search for a cooler thats inbetween 1.3 to 1.45 K/W and you are safe to go?