Finally put it together. Damn, it’s bright!
110 watts at the wall.
Now I’m planning another build to supplement my Roleadro cobs. I might need some advice. 
11 Likes
Awesome! Happy to help if I can, just lemme know what you’re thinking.
2 Likes
I think that is very strange. At 104 watts and only 288 square inches of surface area, that’s only about 5.5 square inches per watt (assuming 50% heat output. That’s nowhere near the 16 inches square per watt some recommend for passive. If the heat sink is only in the 90 F range, then I suspect the LED junctions are a LOT hotter. They would have to be to radiate so much heat.
I bet if you used thermal epoxy instead of thermal tape, a lot more heat would get into the heat sink.
1 Like
I don’t have a way to accurately measure the temp of my heat sink, but the top of the aluminum plate is HOT. I can touch it without burning my hand, but it’s not comfortable. I’m guessing it’s much hotter than 90 deg f.
And I suspect that is what happened with my three defective strips.
Tj tested at 50c with no heatsink at all at 350ma. Does it still seem strange?
My junction temps measured about 6-8 degrees higher than sink temps, and still well below data sheet 50c specs.
It’s not a cob, nor does it need to be treated like one.
2 Likes
You could either aim your grow chamber fan at it so air moved across it, or mount a little whisper fan on it. Electronic Goldmine has some cheap surplus fans.
Walmart online sells the Arctic Silver Alumina Thermal Adhesive about the cheapest I have found it. You have to mix two components together and then use it quickly. Mix enough for one strip at a time.
50 C is 122 F. That will burn skin, but I agree that lots of semiconductors can run at 50 C. Maybe we are wrong and 6 square inches per watt passive is okay.
1 Like
Man!!! That reminds me of mixing silver epoxy and mounting temp sensing diodes on the cold end of IR sensors for Aerospace. We even had a machine that was designed to paint silver stripes on glass coldfingers. Eventually they just started making decals.
We used to buy a 55 gallon drum of scotch tape adhesive and mix up a couple of grams of it with Xylenes and insert in syringes for the operators to glue down the detector for the M1 Abrams MBT. Yes; the thermal imager was put on with scotch tape! I kid you not!
3 Likes
Im willing to bet you have a halfway decent multimeter with your background, any chance you have one with thermocouple? There’s multiple spots to test junction temps, but need thermocouple to do it.
2 Likes
All I know is that every time someone uses anything other than what you’re using, you feel the need to suggest they do something different. I’m ok with it, as I believe the more information people have the better off they’ll be. But there’s now several of these lights in play within this small group alone. @Myfriendis410 lost a couple of diodes on a couple of strips, out of almost 30 strips. That was something like 6 weeks ago, and none since to my knowledge. That doesn’t signify an issue with junction temps, as it likely raises current to the other parallel connected diodes on the strip, and they’re still running.
We could be wrong about what’s needed for passive cooling. Or, these diodes could just be more efficient than you’re use to dealing with, and pcb they’re mounted to pull enough heat. Or all of these lights could be waiting to implode at any second. Hlg’s 550 puts 515 watts to a 2’x3’ sheet of aluminum, and it’s probably the best selling true 1000 hid replacement to date. Their qb 120 and 134 boards blow just about any other light period out of the water on par per watt and use no heatsink at all. The 288 boards mounted on tiny heatsink with no tim at all. That’s up to 150 watts. The 50c, sure it’s hot. But it’s nothing in the led world. That’s where they’re binned, so you get minimum performance specs at that temp. Not anywhere near the 85c diode rating.
Just out of curiosity, what is tj running on your cobs?
1 Like
I never acquired one to go with the multimeter. At one time I had a J type thermocouple and readout but it’s long gone.
And no other failures. I got my replacement strips waiting for after harvest.
1 Like
I’ve got 2 words @Myfriendis410. Harbor Freight. I know it’s cheap Chinese junk but they have a multimeter for $24 that has a temp probe. I had to buy it when I was testing driverless Cobs because the one that goes on my Fluke meter is outrageously expensive.
2 Likes
I don’t think the EB series is any more efficient than their COBs. They run about 175-180 lumens per watt according to their own datasheet, which is pretty comparable to the COBs. Also according to their datasheet, they lose about 9% of their light output as they go from 25 C to 85 C, and the loss is very linear. So the cooler you can get them, the more light you get on your plants.
I have no attachment to “my way”. If people want to use strips, or individual LEDs, that’s fine with me. I even am perfectly happy with HID (and I suggest ways to get rid of heat from them), or CFLs for seedlings and clones. @dbrn32, you were the one who suggested running at low currents and cool to get higher efficiency. I just agree with you. You also said 16 square inches per watt as a rule of thumb, and I have found several other sources including Bridgelux application notes that agree with that.
My COBs have excellent junction-to-heat sink heat transfer, but I don’t have a thermocouple to measure the junction temperatures. But they are mounted on 1/4" copper that runs under 35 C. You can run them at 85 C, but their mean time before failure goes down pretty dramatically. Having a poor thermal transfer to the heat sink can make the junction 10 or even 15 C higher than the heat sink temperature. Likewise, mounting the driver in thermal contact with the heat sink increases the speed of electrolytic capacitor failure by 2x for every 10 C of temperature.
I just want everyone here to have reliable lights that have high efficiency and run for years. And I want them to have enough light and good spectrum so they can grow an excellent product.
3 Likes
This is efficacy, not efficiency. It doesn’t really take into account a lot of things, like series and parallel strings of diodes within the package. Or the fact that the mounting base of cob is about the same size light emitting surface. As to where mounting base of the strip is probably 200-300 times the size of the light emitting surface.
The current/temp/output relationship there is very good. Show me someone that is running a junction temp of 25c lol, maybe very low powered inside of a refrigerator? Looking at 25c data is useless in horticulture application except for maybe comparing one led to another. Ambient temp will have junction temps there with the system off. Knowing that, you can run the junction temp all the way up to max only losing 9% output, that’s pretty damn impressive! On a cost per watt metric, it certainly suggests some trade off there. I mean you can run 5% more efficient running two cxb3590 at 50 watts each from an hlg series driver. That’s gonna triple the price and make this more difficult than most will choose to do.
I do it all the time. That’s why I built this exact fixture and measured the temps prior to putting up a tutorial and listing all the part numbers to do it. It was not to create the most efficient grow light available. It was for the members to grab a relatively simple project, with good par per watt ratio, without needing to spend a lot of money. You and I both know that if you can complete a light like this, you can essentially build any light you could dream up. It just looks less intimidating, and requires no special tools or particular skills.
If your heatsink temp is 35c your junction temp is going to be a couple degrees higher too. Probably just about where I measured junction temps on these strips. I didn’t, nor should anyone else be seeing junction temps of 85c.
I noted this potential for the others, even though my driver is technically not in thermal contact with the heatsink. I’m sure it runs a couple degrees warmer grabbing some of the heat expelled from aluminum. But data sheet shows no loss of component life at or below 50c ambient.
3 Likes
My driver is not mounted to the heat sink. It’s actually outside the grow tent. I guess we’ll see if one fails before the other.
4 Likes
If you put 100 watts in and get 17500 lumens out all the rest of the power has to be heat. Some of that heat goes out the front of the strip and some goes into the heat sink. The temperature of the heat sink determines how much heat is transferred into it. Likewise, the temperature of the strip and the air determines how much heat goes out the front. I’m sure the junction-to-strip transfer is excellent, and the strip-to-heat sink transfer is even better. So the one thing a builder can control is the temperature of the heat sink. 25 C is nearly impossible. You would have to use some kind of active refrigeration. All I’m saying is 35 C is better than 50 C, in terms of LED life and how much heat you want to come out the front of the LED. Maybe 50 C doesn’t matter because the LEDs will last far longer than the electrolytic capacitors in the driver. Electrolytic cap death is a very real problem in all sorts of consumer electronics, and I confess I use electrolytics too because there is no economical alternative.
By the way, I think your fixture construction is quite beautiful. I just would have tried for some fins on that aluminum, but otherwise it shows what DIYers can do at their best. Maybe some of that 104 watts is being burned up in the driver and maybe you fan is blowing enough air around in the grow chamber to work like active cooling.
1 Like
Thanks! The point I was making is that efficacy is light output per watt. And efficiency is light watt vs heat watt. Two leds that are both 175 lumens per watt don’t necessarily have the same amount heat watts present at same points. Take a look at cxb3590 vs vero 29d. Both are each companies flagship offering in 36v package. Par watt vs heat watt below 1400ma is in favor of Cree. 2100ma and above in favor of bridgelux.
I didn’t see 50c on the strip build, just stating the binning data. So if you’re at 175 lumens per watt at 50c, you will be there or better at lower temps, absolutely! Using the data you provided at 9% from 25c to 85c. That’s about 1.5% per 10c at junction temp. Since we agree 25c is unrealistic. We’ll use 35c as base. If my setup is running 40-45c tj (should be accurate for most based on my testing) how much is reasonable to spend to get that extra 1-1.5% gain? How much extra work?
Anyone following can certainly grab some small aluminum cannel at home improvement store and cut to 12” lengths and mount using thermal grease. Or use some of the transfer tape that’s not quite as good. Again for a 1.5% ish gain. Or install a fan, or probably a number of other things. But whole idea of this particular project was get something good at low cost that wasn’t terribly difficult to do. And I think we were pretty successful. There’s no indications that anyone should have an issue outside of maybe an unusually high ambient temp. If and when we change the scope to an Uber efficient build, then different steps and components would be justified.
2 Likes
I wonder if anybody has ever made DIY pin heat sinks by drilling holes in thick aluminum plate and then pounding in aluminum nails? It would be important not to use steel nails because steel is lousy at heat transfer.
4 Likes
Not sure, I’ve never seen anything of the sort. But doesn’t mean it hasn’t been done.
I would think for the trouble, I’d probably be tempted to try a radial heatsink extrusion before that. But someone else may see value a little differently. I’ve seen the radial sinks for as little as $8-$10 rated for 30-50 heat watts. I’ve also seen major players rip on their flatness quality. Not sure if companies like mechatronix have addressed those potential issues or not. If I was pinching pennies I’d probably give them a whirl. More than likely would try in small quantity with inexpensive led.
Perhaps the next budget build.
I would be willing to bet some diy redneck has tried it. I live in a nuckle dragger state. I might just try that @1BigFella. Didn’t mean to jump in, but I couldn’t help myself
1 Like