Heatsinks for DIY LED lamps

I've only ever owned two COB fans (Intel) before and they've been operating for around 90 days, with no signs of stopping anytime soon.

I've noticed those who favor the passive method, whether bias is passive or not, occasionally bring up the possibility of a fan or wire failing. Assuming one knows their wiring and checks their work for weak points, how can they monitor or prevent a given fan from failing?

I like running fans, adds complexity and variety to the main equation but GOA has brought up a decent argument against using fans. I think someone earlier wrote that by using two fans, you double the odds of a fan failing. How do we eliminate that looming ordeal from taking place in our sacred gardens?

AquariusPanta said:

I've only ever owned two COB fans (Intel) before and they've been operating for around 90 days, with no signs of stopping anytime soon.

I've noticed those who favor the passive method, whether bias is passive or not, occasionally bring up the possibility of a fan or wire failing. Assuming one knows their wiring and checks their work for weak points, how can they monitor or prevent a given fan from failing?

I like running fans, adds complexity and variety to the main equation but GOA has brought up a decent argument against using fans. I think someone earlier wrote that by using two fans, you double the odds of a fan failing. How do we eliminate that looming ordeal from taking place in our sacred gardens?

Click to expand...

Use a bi-metal switch, thermistor or thermocouple? Obviously none of these will prevent a fan from stopping, but it can help the rest of the fixture if the fans implode.

Just recently we had a 9V OEM fan power adapter fail on a CXA3070 build with 5.88" heatsinks and 140mm fans, so all four fans quit. That is a bummer because it is hard to get any better quality than OEM. The adapter was not being over driven. It would have been a real bummer if they were mounted on CPU coolers, possibly destroying the CXAs. I am not saying that I think we should rule out CPU coolers or active cooling, but that is an interesting data point to keep in mind.

Greengenes707 said:

But in the situations most grow in/under...there are other ambients and temps pulling heat out of the room, like AP just touched on. Specially in a tent situation...basically non-insulated. The concept of equilibrium comes to mind...
In an active situation, EQ(equilibrium) should be achieved sooner than a passive system.

The heat will flow to the colder areas till EQ is established. The active should be closer to equilibrium faster because of the rate it is dispersed, then it will flow out the tent to the true outside ambient.

Click to expand...

The rate at which equilibrium is established has almost nothing to do with the final temperature of a given space. So it's kinda irrelevant for the point I am trying to make.

To be clear, I'm addressing the insinuation that a passively cooled lamp would add more thermal energy into a given space than an actively cooled light would at the point of equilibrium. Thus raise the final ambient temperatures higher in the passively cooled space than the actively cooled space. Final ambient temperature in this instance would be defined as the ambient temperature in a given space at the end of a given light cycle(IE after 18hours) the moment before the light is shut down. Assuming all other variables are constant.

I am NOT addressing the practicality or cost of a passively cooled vs an actively cooled luminaire. Practicality and budget are user defined values with respect to a given situation. IE Everyone has a different definition of 'affordable' and 'practical' and there are no scientific definitions for either term without additional information from the user.

getsoutalive said:

If you wish to run the 3070 @ 1a, you could most likely, though I have yet to hear anyone who has, run that passively on this $13 sink, @ 1.4 lbs each. Using your number at 18 chips that brings the cost to $240 and the weight down to 25.2 lbs.

If you wish to drive harder, this $27 heatsink, will work just fine at 2+a with a 3070 3k AB. 12 of those will get you into the 1k driven range. Still only $324 of metal and I believe that they weigh on the order of 4 lbs each.

The numbers are not as bad as you make them appear. Obviously, if you can wire up a cob, you can wire and power a fan. But with nothing more than a driver and chip that could fail in a passive setup, you are adding additional possible points of failure and a moving part that will fry your device if/when it fails.

Click to expand...

Your keep missing me...stop thinking 1a, it means nothing to me other than that seems to be the border in peoples mind for "soft" driving. I drive mine at 1.4...where cost meets efficiency.
And if you sourced form the same suppliers I would bet active is still cheaper. apples to apples.
I'm not here to argue one or the other. I driver at a current where active is better imo and specially if cost is factored in...and imo should be. I love the theories and over extravagant efficiency, but I want to save some money while running my lights and beat hps.

How does mass come into the equation for heatsinks?? vs suface area?

You can't have surface area without mass. I think it is that simple.

Organic or chem....both
Car or truck....both
Big heatsink or forced airflow....both
Big boobs or big butt...both....

I know lots side to one or the other. But usually a mix of techniques rule..

Airflow with lots of surface area will be my choice.

@AquariusPanta , I love the king slime

From the testing so far, it seems that thick base plate can reduce the Tj more effectively than increasing surface area. That is supported by the observation that the Alpine 11 100W outperformed the Freezer 7 130W. The freezer has a lot more surface area but the Alpine 11 has a larger base plate and thicker fins. In practice they both performed excellently, but the phenomenon is worth mentioning so we know which way to go.



Next up is the OCZ vanquisher, an eBay bargain. It has a larger copper pad and even more surface area. If the Alpine 11 outperforms this, we are in luck because it is nice when the cheapest option is the also the best option.


And another heatsink that deserves consideration the Rosewill RCX-Z1. If the thermal interface between the copper and aluminum is decent, it might outperform the Alpine 11. It does has a higher RPM fan, so driving it a 5V might be the most efficient approach.

Abiqua said:

Use a bi-metal switch, thermistor or thermocouple? Obviously none of these will prevent a fan from stopping, but it can help the rest of the fixture if the fans implode.

Click to expand...

I'd like to expand on this one a bit as it's something that's been gnawing at the brain and has interesting implications. Hypothetically it would be possible to use a micro controller such as an arduino to monitor heatsink temps for fan failure. If one were to add a relay into the equation it would be possible to shut the lamp down in the event of a fan failure, as well as control the lighting cycle. The arduino also has the ability to generate PWM signals, so this provides the ability to control the speed of a fan(s) with respect to the temperature of the heatsink. For those with compatible drivers, this also means that lamps could be dimmed with respect to the temp of the heatsink, or on a schedule. AKA a whole new level of optimization and efficiency.

Not to mention the host of other automation possibilities... I digress though.

SupraSPL said:

Higher mass heatsinks usually have a thicker base plate and thicker fins. That allows the heat to spread faster to the fins, lowering thermal resistance and reducing the temp gradient. Especially useful for high powered COBs and this can benefit both passive and active cooled of course.

From the testing so far, it seems that thick base plate can reduce the Tj more effectively than increasing surface area. That is supported by the observation that the Alpine 11 100W outperformed the Freezer 7 130W. The freezer has a lot more surface area but the Alpine 11 has a larger base plate and thicker fins. In practice they both performed excellently, but the phenomenon is worth mentioning so we know which way to go.

View attachment 3305636

Click to expand...

I think the problem with most cpu coolers is you end up with a bottleneck. Only so much heat can travel through the small bases. Now if you enlarge the base and heat capture area...along with the surface area...

But thats all kinda unneccessary. Just run them midpower and you can go passive or air cooled depending on how cool or light you want your setup

I wonder if any carbon nanotube heatsinks have hit the market yet.

Greengenes707 said:

How does mass come into the equation for heatsinks?? vs suface area?

Click to expand...

There is nothing wrong with going active, not saying you made a mistake, your solution works absolutely fine and I think you are very much on the right track with your "drone." There are many valid reasons to choose an active solution. I have this thing, as simple as possible, can't help myself.

I like the idea of spreading the COBS evenly across the canopy at as low a height as is necessary to evenly light the canopy. I started a new thread titled Inverse Square Law to talk to talk about cob placement. I very much like the idea of individual heatsinks that can be raised or lowered on a ratchet type system for those canopies that are somewhat less than Kansas flat. Active or passive, your choice. Keep them right down there, like 6 inches from the tops. The chips are bright as the midday Sun at 6 inches away. So get them there. If you notice a bit of bleaching back off some, but not much, just enough. Pack the chips in, one in each sq ft, run them at 1.05a.

More mass and more surface area are both good things but you absolutely can have more surface area with little mass. Think of all those cpu sinks with 50 thin fins, tons of surface area, no mass. Of course you could also have a cube, all mass, no surface area. Found that large sink here. If Cree is satisfied taking a 3050 to 2.5a on it, that is good enough for me.

I cant help but think lately after browsing cpu coolers for hours..

The heatsink usa 3.945" profile is basically a narrower version of the 10.08. It has the thick baseplate and tall fins. It would probably work better cut to cpu cooler size than a cpu cooler itself.

bicit said:

I wonder if any carbon nanotube heatsinks have hit the market yet.

Click to expand...

You mean the Darpa sponsored ones?
Scary? be warned
http://webcache.googleusercontent.com/search?q=cache:OjaRz_hsS3YJ:nano-microworkshop.com/proceedings/slides/Kaiser_Matin.pdf+&cd=8&hl=en&ct=clnk&gl=us&client=firefox-a

bicit said:

I'd like to expand on this one a bit as it's something that's been gnawing at the brain and has interesting implications. Hypothetically it would be possible to use a micro controller such as an arduino to monitor heatsink temps for fan failure. If one were to add a relay into the equation it would be possible to shut the lamp down in the event of a fan failure, as well as control the lighting cycle. The arduino also has the ability to generate PWM signals, so this provides the ability to control the speed of a fan(s) with respect to the temperature of the heatsink. For those with compatible drivers, this also means that lamps could be dimmed with respect to the temp of the heatsink, or on a schedule. AKA a whole new level of optimization and efficiency.

Not to mention the host of other automation possibilities... I digress though.

Click to expand...

Absolutely, it allows essentially whatever you could imagine to build in certain ways. Want to mimic moonlight, write a program, want to mimic the noon sun as your hottest point in your grow period, write a program for it.

Although right now in prototyping, most of these features, including MOV too, could be added without Arduino/Raspberry Pi......

Positivity said:

I cant help but think lately after browsing cpu coolers for hours..

The heatsink usa 3.945" profile is basically a narrower version of the 10.08. It has the thick baseplate and tall fins. It would probably work better cut to cpu cooler size than a cpu cooler itself.

Click to expand...

You could be on to something with that. Would fit a 92 or 100mm fan. If we need more surface area you could just get a slightly longer chunk.

SupraSPL said:

You could be on to something with that. Would fit a 92 or 100mm fan. If we need more surface area you could just get a slightly longer chunk.

Click to expand...

a 3.945x4 square is good for 50w of displacement assuming 110cm²/w. Should run about $13.08/e.

SupraSPL said:

You could be on to something with that. Would fit a 92 or 100mm fan. If we need more surface area you could just get a slightly longer chunk.

Click to expand...

Positivity said:

I cant help but think lately after browsing cpu coolers for hours..

The heatsink usa 3.945" profile is basically a narrower version of the 10.08. It has the thick baseplate and tall fins. It would probably work better cut to cpu cooler size than a cpu cooler itself.

Click to expand...

How does everyone feel about HSUSA's shipping and cutting costs?

If I had the tools, I'd be willing to do a comparison between the Arctic Alpine 11 Plus ($10 - Newegg) and Poz's chosen 3.945 profile (let's say 3.945" x 4" for a nice square ($13.08 w/o extra charges)).

While the 3.945" may end up being more expensive ($5 cutting/handling charge, plus ~$5 shipping per cut), one could simplify his or her COB fixture by avoiding the need for a fan and it's PSU ($~5). Furthermore, it would be possible to place more than one COB on the 3.945" x 4" HS profile. I think I also ran some numbers for power costs of operating a small fan at ~2W over a year (on an earlier page of this thread, see 3)

Running a fan at ~2W every day for ~15 hours over a 90-period would amount to ~$0.30 @$0.12 kWh for a single grow. A single year of use, with 4x grows, would be around $1.20 in electrical costs (keeping in mind that the cost of electricity rises over time). Multiply that by additional fans and the cost in electricity increases even more so.

[(15 hours * 2 Watts) * (1kWatt / 1000Watt) * ($0.12/kWh) * (90 Days / Season)] = $0.324 to power 1x fan over a single grow; $1.296 to power 1x fan over four grows.

I'm pretty new in the game, so I'd like to think that I'd get a year of seamless operation out of my fan, although I don't know their typical life expectancy.

@SupraSPL

Are you a fan of the DWM series??

^just depends....some will go for years, even cheapo sleeve bearings. I would say with regular maintenance you would get years out of a Noctua for instance. Hell the arctics are Suisse made, so they might last for years as well, just started using that particular brand, so I can't quite speak to their longevity.

There are a handful of reputable fans that will last a couple years at least with good maintenance.....