Heatsinks for DIY LED lamps

[SupraSPL]

Are your findings all with heatsink temps?
What about using a thermocouple on the Tc measuring spot to know exactly?
You use kapton tape most of the time right?...the measuring point is accessible for cxa/b's that way unlike with certain holders.

When I check temp droop I monitor the power dissipation with a pair of multimeters to make sure I am accounting for any changes in driver output and I monitor the lux as the COB and heatsink warms up from ambient to stable. Then I calculate the lux/W and what percentage it has changed to find the temp droop figure.

So the process of measuring temp droop does not require measuring Tc or Th. Of course I am curious about the corresponding temps for a given temp droop so I have done many experiments to try and estimate junction temps and get familiar with what heatsink temps correspond to what temp droop figures. I also try to watch the change in Vf as the COB warms up to add data point to the junction temp estimates. I use an infrared baby food thermometer to check heatsink temps but I have tried many methods including thermocouple and shielded mercury thermometers. I have found that measuring Tc (for me) is impossible due to interference from the COBs photon/infrared output, no matter which method I try. I experimented with switching the COB off to try and get the reading without photon interference but that also proved not-repeatable or reliable. You can also try shielding the thermometer from the COB output using piece of aluminum foil which will block visible and infrared photon output except what is reflecting from behind it or from your own skin. I will continue to experiment with these methods but ultimately I find measuring the temp droop directly the best indicator of thermal performance.

So eventually I settled on measuring the hottest part of the heatsink and monitoring the Vf change to get repeatable results quickly and easily. BUT that only works if you are confident that your thermal interface is solid. If the thermal interface is poor the heatsink temp will actually be lower, the Tc would be higher but this cannot hide from a temp droop test. Anyway, for an aggressively low temp droop I try to keep my heatsink temps below 30C (<1% temp droop). That can be tricky with passive cooling, but doable if you are running soft and benefiting from the circulation fan. For commercial lamps I think 45C (<5% temp droop) is a good maximum temp for the hottest part of the heatsinks, again assuming you have a solid thermal interface (sufficient TIM with no air bubbles).

 

[SupraSPL]

So let's say the dip in the middle of the heatsink in Slitlos's picture is .75mm - right in the middle where a COB would be mounted. And let's also say that a properly applied grease layer is .25mm. So right in the hottest part of the COB, there's a 1mm thick layer of TIM. That much is OK? Has this horse already been beaten too much?

Going against convention, I would much rather use too much TIM rather than too little. So I put a nice layer on and then squeeze the excess out until I am satisfied that the entire interface is filled completely and that the layer is no thicker or thinner than necessary.

I just checked a piece of 3.5" heatsink and found the dip in the center to be .1mm. It can be quite a lot of work to get rid of that dip and I am confident your thermal performance will not change. I compared a stock Alpine 11 to a polished one as well as a stock Rosewill RCXZ1 copper plate versus a polished one. No difference in thermal performance in any of those tests, even at high dissipation levels.

If you have a smallish build or you are good with belt sanding, and you like the knowledge that your heatsinks are truly flat I say go for it. Once they are flat at 80 grit, you are only a few steps away from being polished. Polishing them will to some extent reflect light from the canopy back down to the canopy.

For the record I don't expect folks to take my word as gospel. I encourage those that are curious to give it a shot, more data points are very welcome. You can check temp droop with a kill-a-watt and a lux meter if you pre-warm up your drivers somehow (or if you can take the change in driver efficiency into account) but of course without warming up the COBs before the test. The lux meter and LEDs have to remain in the exact same position during testing but other than that pretty easy. You can check the average temp droop of your entire tent at once or any LED fixture this way.

 

[bicit]

Going against convention, I would much rather use too much TIM rather than too little. So I put a nice layer on and then squeeze the excess out until I am satisfied that the entire interface is filled completely and that the layer is no thicker or thinner than necessary.

I just checked a piece of 3.5" heatsink and found the dip in the center to be .1mm. It can be quite a lot of work to get rid of that dip and I am confident your thermal performance will not change. I compared a stock Alpine 11 to a polished one as well as a stock Rosewill RCXZ1 copper plate versus a polished one. No difference in thermal performance in any of those tests, even at high dissipation levels.

If you have a smallish build or you are good with belt sanding, and you like the knowledge that your heatsinks are truly flat I say go for it. Once they are flat at 80 grit, you are only a few steps away from being polished. Polishing them will to some extent reflect light from the canopy back down to the canopy.

For the record I don't expect folks to take my word as gospel. I encourage those that are curious to give it a shot, more data points are very welcome. You can check temp droop with a kill-a-watt and a lux meter if you pre-warm up your drivers somehow (or if you can take the change in driver efficiency into account) but of course without warming up the COBs before the test. The lux meter and LEDs have to remain in the exact same position during testing but other than that pretty easy. You can check the average temp droop of your entire tent at once or any LED fixture this way.

Have you by chance compared a non-polished and polished heatsink under passive cooling conditions?

 

[littlejacob]

Bonjour
Is anyone could tell me what size, in cm if possible, a heatsink must have to cool passively 50w.30x30x3cm, less, more?
@SupraSPL i have access to aluminium "leaves"(don't know the english for this) 3mm thick, is it possible to build a profile with like tunnels with ventilator in it or it is not possible.I can use many plates of aluminum and many machines to transform it.
Have a nice day★

 

[bicit]

Bonjour
Is anyone could tell me what size, in cm if possible, a heatsink must have to cool passively 50w.30x30x3cm, less, more?
@SupraSPL i have access to aluminium "leaves"(don't know the english for this) 3mm thick, is it possible to build a profile with like tunnels with ventilator in it or it is not possible.I can use many plates of aluminum and many machines to transform it.
Have a nice day★

Do you have a picture by chance? You should aim for ~90-110 cm^2 of surface area for adequate passive performance.

IDK if you already tried this, but try plugging it into google translate

 

[HydroDC]

Bonjour
Is anyone could tell me what size, in cm if possible, a heatsink must have to cool passively 50w.30x30x3cm, less, more?
@SupraSPL i have access to aluminium "leaves"(don't know the english for this) 3mm thick,

Perhaps in English, the word is sheets - as in pieces of flat aluminum. If so, there have been some DIY builds using aluminum square stock (which could be made from sheets - or bought as square stock) cooled by a small blower at one end. Not sure if there would be enough cooling for high powered COBs.

 

[littlejacob]

Hello guys
Yes i was thinking of building something like a tunnel with a fan who blow fresh air on a side and another on the other side to evacuate hot air.
The thing is i still don't know the size of a heatsink who would cool passively 50 w cob? 20x20x3 cm is enough without fan. I'm going to had a fan but i would like something who can preserv my cob to burn in case of fan failure.Just the minimum size and weight please.I believe it is not gonna be easy cause there's so many different H.Sink plain extruded, etc so minimum for 50 w and i will be happy.
Have a nice grow weeders!

 

[HydroDC]

Hello guys
Yes i was thinking of building something like a tunnel with a fan who blow fresh air on a side and another on the other side to evacuate hot air.
The thing is i still don't know the size of a heatsink who would cool passively 50 w cob? 20x20x3 cm is enough without fan. I'm going to had a fan but i would like something who can preserv my cob to burn in case of fan failure.Just the minimum size and weight please.I believe it is not gonna be easy cause there's so many different H.Sink plain extruded, etc so minimum for 50 w and i will be happy.
Have a nice grow weeders!

The idea of a heatsink is to put a sufficient amount of surface area at the point and immediately around where the heat is being generated. A flat piece of aluminum, with no fins, is not that. So if you end up with heatsink in your design (and this thread has many discussions of which heatsink designs work for different COB choices), there's little reason to go with a tunnel in addition to that. Just add a fan and a shroud to the heatsink. The tunnel/blower designs I've seen have used low powered (3 watt, 5 watt) LEDs, not high powered COBs. It might work as long as the airflow was high enough, but that's not the passive protection you say you want.

 

[littlejacob]

Bonjour
Ok thank you
Is a 1000x165x35cm (40x6,7x1, 3inch) 14 fins is enough for 4cxb3590 at 50w? Cut in 2 half with a 120mm fan in th middle on each or a 100mm on each cob with 2 alimentation and plug fans on each pannel on a different one in case of failure i may still have one fan who still cool.
Have a great day

 

[HydroDC]

Bonjour
Ok thank you
Is a 1000x165x35cm (40x6,7x1, 3inch) 14 fins is enough for 4cxb3590 at 50w? Cut in 2 half with a 120mm fan in th middle on each or a 100mm on each cob with 2 alimentation and plug fans on each pannel on a different one in case of failure i may still have one fan who still cool.
Have a great day

May I suggest you spend a bit of time studying this thread as well as the thread called "Heatsinks for Europe"? I think you'll find all the answers you need.

There also may be someone in the "Heatsinks for Europe" who can better help with "alimentation". I think you want to power your fans from two different power sources in case one of the power sources fails, but I'm not really clear what alimentation is. If you have a sufficient passive heatsink, and fans as well, I shouldn't think additional failsafe means are needed. Best of luck with your project.

 

[SupraSPL]

Inspired by @tightpockt fried egg I have a COB torture test in mind. A sacrificial Cree CXA1512 low bin, mounted onto an unsanded heatsink of insufficient surface area passive cooled, with cheap but decent thermal paste (Ceramique) sloppily but thoroughly applied, barely kapton taped in place, and driven hard, well above its maximum. BUT I will make certain that the thermal interface remains intact and it will be interesting to see what kind of temp droop we get and if the COB survives. If the COB dies quickly the test proves nothing because of all the abusive variables I threw at it, but if it survives awhile it will tell us a lot.

I mounted and soldered it in about 2 minutes. Pre heated the driver and begin temp droop test. The heatsink is not up for this job, it will be handling ~25W of heat but is only good for about 7W.


CXA1512 @800mA (133% of operating maximum), starting at 21C

30 seconds - rapid drop in output as the junction warms up, 22% temp droop.
60 seconds - 23% droop, heatsink is still cool 26C
10 minutes - 39% droop, heatsink is getting hot ~ 45C
60 minutes - 53.4% droop, almost stabilized, heatsink above 55C not good!
75 minutes - 55% droop, COB is taking damage Vf is rising instead of falling
80 minutes - when I click the light off and watch it fade i can see that only half is lighting up correctly.

Interesting test I can do when the heatsink is super hot, if I click the power off for 15 seconds and allow the junction temp to nearly equalize with the heatsink temp, the output has hugely recovered when I switch it back on (30% temp droop), but then falls rapidly to 55% droop as the junction heats back up. This may help us get a better idea of the actual thermal resistance of the COB package and the actual junction temps.

Its dead! Only flickers when I turn it on. Looks like something burned in there. Like we say in motorcycle racing, you don't know how much is enough until you know how much is too much. Next test I will back off a bit.

Post-mortem:

 

[SupraSPL]

To put those numbers in perspective, DIY COBs at at 25W can be less than 1% temp droop, DIY COBs on 50W on CPU coolers test about 2-3% temp droop. The new Optic Lighting Vero 29 V2.0 that I tested was about 3% temp droop and they were driven at 80W ea. The first model they offered with the Vero 18s tested at 14% and was easily reduced to 2% after the update.

 

[testiclees]

Folks, I was able to divide the heat sink easily with my mitre saw. No bucks, hangups etc,

I did though struggle with balancing my design. Since the heat sinks are cut to fit the odd foot print they do not mirror each other. One is 13 x 12 the other is 18 x 12. I need to get my rigging straightened out. The heat sinks are grimy and need to be thoroughly cleaned.

The overall length is 48". My plan is to mount 2- cxb3590 3k (72v) @.700mA on each heat sink, passively cooled.

But I did make some progress and the concept seems to be workable. I could use some suggestions on an easy balancing method that works with my single solid anchor point for hanging.

Also how much longer for the lil lady? She was 10 weeks from flip yesterday.



EDIT: "Like we say in motorcycle racing, you don't know how much is enough until you know how much is too much." I learned too much about too much. ZX10r, ZX6r
@pocono.

 

[PurpleBuz]

nice job, that's a great big fan to cool the heatsinks, that box fan

 

[HydroDC]

I could use some suggestions on an easy balancing method that works with my single solid anchor point for hanging.

Great creative re-use of an old light. Glad the cut went easy.

Riggers would say to have 4 equal length legs from your central hanging point going to four equally spaced attachment points as measured from the center of the load. Find the light's center of gravity when it's on your workbench and use that to space out 4 equi-distant attachment points - the heavy end will hang out further from those points.

You also have a very sharp sling angle in your picture. Each leg is probably supporting something like 3 times an equivalent vertical load because of the tension factor. So your quite right to use a stout material for your legs like the chain I see.

 

[littlejacob]

Hello weeders
Anyone can tell me why black heatsink are more (efficient) expensive?
I'm going to buy the 1000x165x35mm heatsink that I'm going to cut in four pieces and add 4 140mm fan at 5V to cool 4 cxb 3590 3000k @1400mA
Is it ok to cut it in 4? Should i cut it in half with 2 cob on each? I want to go that way because i want to move them independently!
Thanks and have a great day

 

[littlejacob]

And I forgot:when i am going to make a "box" with aluminum sheets (i know the word now) will i have to touch the H.S with aluminum to have a better heat spreading?
See you★

 

[salmonetin]

...maybe this helps a bit...

https://www.rollitup.org/t/active-and-passive-cooling-on-anodized-heatsinks.873356/

saludos

 

[EfficientWatt]

Bonjour petit Jacob,

Anodized (black) :
- Passively, are more efficient because they have better heat transfer with surrounding air.
- With active cooling (fans), it doesn't make any real difference.

Anodized is more expensive, because they have to be anodized !

You should know that non anodized heatsinks (grey) will get a little better at passive cooling overtime, as they slowly get oxidized.

Hello weeders
Anyone can tell me why black heatsink are more (efficient) expensive?

 

[guod]

You should know that non anodized heatsinks (grey) will get a little better at passive cooling overtime, as they slowly get oxidized.

it gets better and better