Bridgelux EB Gen 2 Thermals

noodle-led

Well-Known Member
In 2016 I saw a Hackaday post about someone building their own grow light and I got excited about the prospect. Literally 2 google searches later I ended up here and was given more information about every aspect of LED grow lights than I ever thought I needed to know. Thanks for steering me away from grabbing some blurple LEDs and having a go at it! I've got a small tent with a couple Gen 7 Vero 29 cobs that does a pretty good job for the past year+, but the light is pretty uneven in there so I've been looking at alternatives.

The EB Gen 2s look good and their ~2ft size would make a pretty good fit for my 32x32" space. The questions I had were
  • How many should I use? My current setup is 2x COBs on 120mm pin heatsinks on a HLG-185H-C1050 running at just over 70W. They generate ~24800 theoretical lumens at ~176lm/watt with Ledil Angela 90 degree reflectors. I'd obviously want better than this so it was a good starting point.
  • How hard can I drive the strips? U-channel or heatsinks can be used to keep the temps down, but 2ft for each strip x 10 strips, the costs gets out of hand, so the goal would be to maximize the output without causing failures and minimizing cost.
  • How hot do they get? This is sort of related to the second point, but heat is largely related to reliability.
  • How many strips on a driver? The datasheet says to model driver selection with a 22.1V/560mm strip Vf but that creates some sub-optimal breakpoints given the forward voltage for Mean Well drivers. HLG-240H-CXXXX can get 8x1.4A for only 217W or 10x1.05A for 204W. HLG-185H-CXXXX are 6x1.4A (162W) and 8x1.05A (162W). I wanted to see some real world numbers before committing to getting a new driver and strips.

Testing
One 3500K 560mm Bridgelux EB Gen 2 strip was purchased from Arrow with one of my recent parts buys. The strip was powered by soldering 2 inch 20AWG wires to the two terminals at the end of the strip, and fed with 18AWG wires from a bench power supply. The strip was elevated on 1.5" wooden blocks at either end and one in the center with the LEDs facing up, unless otherwise noted. The strips were left running and checked every 5 minutes until the forward voltage was the same across two checks. Ambient temperature was 19.1C, and as an extra data point, the strip at 1400mA fully cold tops out at 20.50Vf but drops very fast as it heats up. I checked each LED's thermals for each test, but every time I ended up with the same group of LEDs being the hottest on the whole strip.

First, the numbers in free air, no ambient airflow or heatsinking.
700mA 19.26Vf 43.4C

1050mA 19.68Vf 56.4C

1400mA 20.08Vf 71.9C


Note that these are not the best at thermal transfer to the back of the strips. The back of the 71.9C strip was more than 5C below the front of the strip


It is clear that at 700mA, Bridgelux's claims are true: no heatsinking required. Next I tried a variety of materials for heatsinking at 1400mA, all attached with 20mm thermal tape. 4" wide strips of flashing material were chosen as a common spacing for 24" installations with 6 strips.
Code:
1" wide x 1/8" aluminum bar -- 20.12Vf 67.0C
1/2"x1/2"x1/16" aluminum L bracket attached to the back center of the strip -- 20.13Vf 66.3C
1"x2"x1/16" aluminum tube, attached on the 1" side -- 20.22Vf 55.5C
4" wide strip of heavy duty Reynolds aluminum foil -- 20.11Vf 67.1C
4" wide strip of galvanized steel flashing -- 20.20Vf 56.4C
4" wide strip of aluminum flashing -- 20.21Vf 55.7C
Finally I ran a set of tests with the bare strips again at different currents to note the response of power to heat. What was very interesting was that there is a knee on the power vs temperature rise curve. I didn't believe it so I ran it twice and got the same results:


As a side note, I also turned on a fan in the room for one of the tests against the bare strip at 1400mA. The Vf climbed from 20.08V to 20.16V which indicates somewhere around a 5-10C drop in temperature. I did not record the temperature in my notebook for some reason though. Was alcohol involved?

Conclusions
  • It seems the best efficiency considering output to power is somewhere around 1A, so a 1050mA driver is probably the sweet spot if you want to get the most out of these.
  • If you're planning on driving at 700mA, don't waste any money on heatsinking.
  • The lower than spec forward voltage is also a boon, meaning a HLG-185H-C1050 can drive 9 strips at 1050mA for 185W and a theoretical 32,400 lumens (175lm/W).
  • Don't waste your money putting 1" wide 1/8" aluminum bar on on each strip, aluminum foil works just as well once the temperature reaches equilibrium. This is no surprise, you haven't increased the radiant surface area by adding a slab the size of the strip to the back.
  • Galvanized steel is surprisingly good and cheap compared to the expensive tubular aluminum (which I will admit I heated with a heat gun to bring it to temperature and then let it cool to equilibrium, it was just taking too long to heat on its own). Aluminum flashing is probably the most economical solution to provide rigidity and cooling.
  • Figure at worst a 20.5V forward voltage at 1400mA when considering drivers, or a 20V forward voltage for 1050mA

In short, I am looking forward to Digikey and Arrow stocking more of these in different color temps. I'd prefer doing 5x 4000K strips and 4X 3000K strips with a switch to bypass and disable the 3000K while running them off the same driver for veg. Although maybe I am tipping my hand too much about my next build plans...
 

nfhiggs

Well-Known Member
Conclusions
  • It seems the best efficiency considering output to power is somewhere around 1A, so a 1050mA driver is probably the sweet spot if you want to get the most out of these.
  • If you're planning on driving at 700mA, don't waste any money on heatsinking.
  • The lower than spec forward voltage is also a boon, meaning a HLG-185H-C1050 can drive 9 strips at 1050mA for 185W and a theoretical 32,400 lumens (175lm/W).
  • Don't waste your money putting 1" wide 1/8" aluminum bar on on each strip, aluminum foil works just as well once the temperature reaches equilibrium. This is no surprise, you haven't increased the radiant surface area by adding a slab the size of the strip to the back.
  • Galvanized steel is surprisingly good and cheap compared to the expensive tubular aluminum (which I will admit I heated with a heat gun to bring it to temperature and then let it cool to equilibrium, it was just taking too long to heat on its own). Aluminum flashing is probably the most economical solution to provide rigidity and cooling.
  • Figure at worst a 20.5V forward voltage at 1400mA when considering drivers, or a 20V forward voltage for 1050mA

In short, I am looking forward to Digikey and Arrow stocking more of these in different color temps. I'd prefer doing 5x 4000K strips and 4X 3000K strips with a switch to bypass and disable the 3000K while running them off the same driver for veg. Although maybe I am tipping my hand too much about my next build plans...
Good info.

As to the galvanized steel - nah. Steel holds a lot of heat, but it does not transfer heat the way aluminum does. Aluminum absorbs and dissipates heat much faster than steel. The only thing better is copper. I used 3/4 aluminum U channel for my strips.
 

Hydro2112

Member
Nice work. For the 72C@1400mA, what does Tc read? Pretty sure the individual LEDs will get pretty hot, but wondering how this relates to Tc as the remainder of Bridgelux's data is wrt this.

Also, curious about the OSH park PCB in your profile.. the LED pads and the ESP make tell me its some kind of wifi controlled luminaire, what cool stuff is this :)
 

Chip Green

Well-Known Member
Props to you, for quantifying what I had concluded with my Gen 1 EBs....
The flashing material sheets I run the 1050ma panels at certainly kept the voltage droop to an acceptable level, but its nice to see some additional visual evidence...
The 700ma "racks" without any backing material, are working as advertised...
 

Randomblame

Well-Known Member
In 2016 I saw a Hackaday post about someone building their own grow light and I got excited about the prospect. Literally 2 google searches later I ended up here and was given more information about every aspect of LED grow lights than I ever thought I needed to know. Thanks for steering me away from grabbing some blurple LEDs and having a go at it! I've got a small tent with a couple Gen 7 Vero 29 cobs that does a pretty good job for the past year+, but the light is pretty uneven in there so I've been looking at alternatives.

The EB Gen 2s look good and their ~2ft size would make a pretty good fit for my 32x32" space. The questions I had were
  • How many should I use? My current setup is 2x COBs on 120mm pin heatsinks on a HLG-185H-C1050 running at just over 70W. They generate ~24800 theoretical lumens at ~176lm/watt with Ledil Angela 90 degree reflectors. I'd obviously want better than this so it was a good starting point.
  • How hard can I drive the strips? U-channel or heatsinks can be used to keep the temps down, but 2ft for each strip x 10 strips, the costs gets out of hand, so the goal would be to maximize the output without causing failures and minimizing cost.
  • How hot do they get? This is sort of related to the second point, but heat is largely related to reliability.
  • How many strips on a driver? The datasheet says to model driver selection with a 22.1V/560mm strip Vf but that creates some sub-optimal breakpoints given the forward voltage for Mean Well drivers. HLG-240H-CXXXX can get 8x1.4A for only 217W or 10x1.05A for 204W. HLG-185H-CXXXX are 6x1.4A (162W) and 8x1.05A (162W). I wanted to see some real world numbers before committing to getting a new driver and strips.

Testing
One 3500K 560mm Bridgelux EB Gen 2 strip was purchased from Arrow with one of my recent parts buys. The strip was powered by soldering 2 inch 20AWG wires to the two terminals at the end of the strip, and fed with 18AWG wires from a bench power supply. The strip was elevated on 1.5" wooden blocks at either end and one in the center with the LEDs facing up, unless otherwise noted. The strips were left running and checked every 5 minutes until the forward voltage was the same across two checks. Ambient temperature was 19.1C, and as an extra data point, the strip at 1400mA fully cold tops out at 20.50Vf but drops very fast as it heats up. I checked each LED's thermals for each test, but every time I ended up with the same group of LEDs being the hottest on the whole strip.

First, the numbers in free air, no ambient airflow or heatsinking.
700mA 19.26Vf 43.4C

1050mA 19.68Vf 56.4C

1400mA 20.08Vf 71.9C


Note that these are not the best at thermal transfer to the back of the strips. The back of the 71.9C strip was more than 5C below the front of the strip


It is clear that at 700mA, Bridgelux's claims are true: no heatsinking required. Next I tried a variety of materials for heatsinking at 1400mA, all attached with 20mm thermal tape. 4" wide strips of flashing material were chosen as a common spacing for 24" installations with 6 strips.
Code:
1" wide x 1/8" aluminum bar -- 20.12Vf 67.0C
1/2"x1/2"x1/16" aluminum L bracket attached to the back center of the strip -- 20.13Vf 66.3C
1"x2"x1/16" aluminum tube, attached on the 1" side -- 20.22Vf 55.5C
4" wide strip of heavy duty Reynolds aluminum foil -- 20.11Vf 67.1C
4" wide strip of galvanized steel flashing -- 20.20Vf 56.4C
4" wide strip of aluminum flashing -- 20.21Vf 55.7C
Finally I ran a set of tests with the bare strips again at different currents to note the response of power to heat. What was very interesting was that there is a knee on the power vs temperature rise curve. I didn't believe it so I ran it twice and got the same results:


As a side note, I also turned on a fan in the room for one of the tests against the bare strip at 1400mA. The Vf climbed from 20.08V to 20.16V which indicates somewhere around a 5-10C drop in temperature. I did not record the temperature in my notebook for some reason though. Was alcohol involved?

Conclusions
  • It seems the best efficiency considering output to power is somewhere around 1A, so a 1050mA driver is probably the sweet spot if you want to get the most out of these.
  • If you're planning on driving at 700mA, don't waste any money on heatsinking.
  • The lower than spec forward voltage is also a boon, meaning a HLG-185H-C1050 can drive 9 strips at 1050mA for 185W and a theoretical 32,400 lumens (175lm/W).
  • Don't waste your money putting 1" wide 1/8" aluminum bar on on each strip, aluminum foil works just as well once the temperature reaches equilibrium. This is no surprise, you haven't increased the radiant surface area by adding a slab the size of the strip to the back.
  • Galvanized steel is surprisingly good and cheap compared to the expensive tubular aluminum (which I will admit I heated with a heat gun to bring it to temperature and then let it cool to equilibrium, it was just taking too long to heat on its own). Aluminum flashing is probably the most economical solution to provide rigidity and cooling.
  • Figure at worst a 20.5V forward voltage at 1400mA when considering drivers, or a 20V forward voltage for 1050mA

In short, I am looking forward to Digikey and Arrow stocking more of these in different color temps. I'd prefer doing 5x 4000K strips and 4X 3000K strips with a switch to bypass and disable the 3000K while running them off the same driver for veg. Although maybe I am tipping my hand too much about my next build plans...

Well done, mate!
Confirms the results which have been published here.
EB-strip temps(passive) with and without c-channel.jpg
 

noodle-led

Well-Known Member
@nfhiggs yup, the steel soaks up the heat and spreads it out, but it's no substitute for aluminum when it comes to cooling. Still, it is more effective than just leaving them bare if a little cooling is needed and galvanized steel is already on hand. It is more conductive than the FR4 material the strips are made from. 3/4" U channel would be great but expensive for me if I'm looking at 9x 2ft strips. The light parts only cost $117 but 22ft of 3/4" aluminum L would run me over $30 (only have a big box hardware store local). Actually now that I do the math that's actually not terrible.

@Hydro2112 Good question. I used some kapton tape and heatsink compound with a bare metal thermocouple on Tc1, conveniently the location that all the FLIR images were taken. My ambient temperature was about 2C higher this morning than it was when I did the other testing, but the results are roughly identical to the FLIR. The results are a bit noisy, the bare thermocouple attached right to the metal of strip introduces a little ground loop noise because my recording device is not fully isolated or possibly the strip itself is acting like an antenna and I am picking up a TV station. This is 1400mA, 1050mA and 700mA. The numbers pretty much match those recorded above. Full data (oh still not enough posts to post links!)


Good eye though, my avatar is a prototype light controller PCB! I was trying to find the pictures of my Vero light during assembly and couldn't find them so I just went with what I am working on now. It is based on Growmau5's Intiator Puck, but I didn't like the idea of having yet another device on a timer outlet. It already has 12V running to the puck, why not build the timer into the puck itself? What resulted was a wifi puck with
  • 4x onboard XP-E 730nm LEDs, full PWM dimming on those with the wifi controller running the scheduling. Or a manual button onboard for quick on/off.
  • It also has an TI OPT3002 lux sensor, which can be used to automatically trigger the far red LEDs at lights out with no scheduling required.
  • Feature creep then took over and I added a DHT22 temperature and humidity sensor.
  • Not done yet, there's also 0-10V PWM dimming for controlling the main lighting power supply.
  • If we're also controlling the main lighting supply, there's also a knob for manual override of the main lighting supply (because who wants to break out their phone to quickly turn down the lights while working in the tent). The knob can be set to manually override into any range, say if you want manual control from 50%-100% like a model A, you got it. 10%-75%? Sure. Really trying to dial that output in, you can set it to go 80%-90% only. It can also be configured to automatically return to programmed brightness after a certain amount of time if, like me, you turn down the lights then forget to turn them back up when you're done.
  • Of course it is wifi too so it integrates with MQTT for full automation and reporting, or Alexa control I guess but who needs that?

@Chip Green Yeah! Actually your really clean build was what I was going off of for the aluminum flashing. I actually had this in my garage from another project oddly enough and I love the idea of the whole thing being covered, radiating heat, and reflecting some light back down. I was planning on using screws to attach the strips to the aluminum but holy cow trying to remove the LED strips from the thermal tape between tests was really really difficult so I don't think any screws will be needed.

@Randomblame I love graphs, I can't get enough graphs. Your postings definitely inspired me to undertake the testing and let me know that 1400mA wasn't Bridgelux saying "Sure you can run them at 1400ma... if you chill them in liquid nitrogen!" My numbers are a lot higher than yours, I'm not sure why. Aren't your tests done with DS18B20 ICs strapped to the boards? I would think they would conduct the heat better.
 

noodle-led

Well-Known Member
I thought this was funny too. After I powered down the strip, the LEDs were actually being powered through the thermocouple datalogger. I'm really scratching my head as to how I didn't fry it if the Tc1 pad is actually connected to the circuit and I just pumped 1.4A @20V across it. Next time I'll use a strip of thermal tape instead of heatsink compound.
 

Randomblame

Well-Known Member
@noodle-led
This was not my test. I cherry picked the pic from another EB thread and I don't remember if a thermo probe or an IR thermometer was used. I think also the number are from gen.1 EB strips.
And yes, there are differences but in the end it shows the same; no heatsink required at 700mA.
 

Randomblame

Well-Known Member
I thought this was funny too. After I powered down the strip, the LEDs were actually being powered through the thermocouple datalogger. I'm really scratching my head as to how I didn't fry it if the Tc1 pad is actually connected to the circuit and I just pumped 1.4A @20V across it. Next time I'll use a strip of thermal tape instead of heatsink compound.
A quick google search gave me a conclusion.

"A thermocouple is a thermoelectric device for measuring temperature, consisting of two wires of different metals connected at two points, a voltage being developed between the two junctions in proportion to the temperature difference."

Maybe enough voltage to drive a few diodes very low.
 

noodle-led

Well-Known Member
Yeah I designed the thermocouple circuit myself so I knew there was up to 3.3V on the wires, but I didn't think that the Tc points on the strips were in-circuit. I assumed they were just copper pads on the substrate isolated from the LEDs themselves. It does make sense because they would better sense the temperature this way but I wonder how you're supposed to monitor the temperature in a production setup if the Tc points are at a potential above earth ground as would be the case in a mains constant current supply.
 
Thanks noodle-led. I needed confirming that the gen.2 don't need heasinking at 700ma. I have a hard time getting hold of U channels around here and want to make a strip build.
What about head clearance? Mine would be very close to the tent ceiling, 6 inches at most. Would that affect things drastically?
 

Randomblame

Well-Known Member
You must remember that you always have some air movement inside a grow tent, which further lowers the Tc. temperatures of the strips. So it should work very well...
 

noodle-led

Well-Known Member
Exactly. I didn't include it in the post because it was already pretty information-dense, but I also did a test with a ceiling fan in my lab on low speed about 6-8ft away diagonally. Just a minor almost imperceptible airflow raised the Vf on the strips at 1400mA from 20.08V to 20.16V which we can interpolate as over a 5C decrease in temperature (to about 66C). I think as long as you've got air exhausting out of the tent you'd see similar benefit and negate the fact that they are in close proximity to the top.
 
Exactly. I didn't include it in the post because it was already pretty information-dense, but I also did a test with a ceiling fan in my lab on low speed about 6-8ft away diagonally. Just a minor almost imperceptible airflow raised the Vf on the strips at 1400mA from 20.08V to 20.16V which we can interpolate as over a 5C decrease in temperature (to about 66C). I think as long as you've got air exhausting out of the tent you'd see similar benefit and negate the fact that they are in close proximity to the top.
I will obviously have a fan going across the strips and fans across the plants as well. I have air circulating too.
I also run 2xQB304 with slate 2 heatsinks in a similar space only 4-5 inches from the deck above and it's only gets slightly toasty when I crank it to max.
Enough of these strips at low current and I don't need heatsinks. With arrow's free shipping, it might even be cheaper then driving 45mins to buy them U channels @ $25 a piece.. :blsmoke:
RUI has got the worst icons :spew:
 

nfhiggs

Well-Known Member
but I wonder how you're supposed to monitor the temperature in a production setup if the Tc points are at a potential above earth ground as would be the case in a mains constant current supply.
The power supplies driving them have ground isolated outputs.
 

noodle-led

Well-Known Member
You probably could use them if you put strips along the wide dimension. A full sheet pan is usually just under 26x18" external dimensions so along the 18" (457mm) dimension you'd have too much sticking off the edge to keep those LEDs cool. Using the 26" dimension would work great, but then you'd still need a solid method of connecting multiple sheets. Probably even 1/2"x1/2"x1/8" angle would be sturdy enough. I think the problem is price though, because I can get 24"x10ft aluminum flashing for $18 shipped so for making a 2x2 or a 2x4 wouldn't that be more economical even with the aluminum angle for the perimeters? Or is there somewhere to buy 2-3 full sheet pans for like $10 total/shipped?

Also @nfhigs you're right I didn't think about that. Of course they'd have isolated supplies.
 

ANC

Well-Known Member
IMG_20180129_080942s - Copy.jpg
I will obviously have a fan going across the strips and fans across the plants as well. I have air circulating too.
I also run 2xQB304 with slate 2 heatsinks in a similar space only 4-5 inches from the deck above and it's only gets slightly toasty when I crank it to max.
Enough of these strips at low current and I don't need heatsinks. With arrow's free shipping, it might even be cheaper then driving 45mins to buy them U channels @ $25 a piece.. :blsmoke:
RUI has got the worst icons :spew:
It is still cheaper than buying heatsinks for COBs, two 8-foot pieces goes a long way.

I used a different type of channel, a bit thinner gauge than U-channel, but much lighter and more surface area.
 
Last edited:

Chip Green

Well-Known Member
That roll flashing, is working fantastic for all of my builds.
Super easy to cut to size/shape, with ordinary heavy duty shop scissors....
I've considered those pans, but so far never actually tried one.
 
Top