DiY LED - Cree CXA3070

SupraSPL

Well-Known Member
So you are passively cooling these?
Yessir they are providing 5450 sq cm for 50 dissipation watts so it is about 110 sq cm/watt. Although they are passively cooled, the circulation fan in the tent (ceiling fan in this case) makes a huge difference in heatsink temp. I turned the fan off for a few hours and tested heatsink temps but that was before I had the infrared thermometer so I will do another test. Currently the ambient temp is 21C and the heatsink temp is 24C. Tj is ~44C, estimated by measuring the change in vF from cool to warm. The surface of the emitter itself seems to be about 32C.

Running them this cool means there will be virtually no lumen depreciation over a decade of use and they enjoy an output increase of about 6.5% compared to running at Tj 85C
 

mtnstream

Active Member
You would need a PWM driver with a arduino ($) controlling dimmer. Then you could program to simulate sunrise/sunsets. See reefers and diy setups.

Dimming could be extremely useful. We could decrease the output during the first few weeks of flowering and increase it as we spread the plants out and the canopy grows wider and deeper. Maybe dim during the last week also?

KNNA once pointed out that when it comes to growing lamps, the best kind of dimmer is an adjustment to the actual drive current of the constant current driver. Dimmers that cycle the led on and off cause the led to suffer from current droop during the on cycle. It does achieve the goal of dimming but not as efficient of a method especially for long term use. That is as far as my knowledge goes, how would a phased dimmer operate?
 

smokey the cat

Well-Known Member
Yes I noticed they pick up dust and particles very easily. I kind of wack them with a lint free cloth rather than wipe them. That might be something we will have to do regularly.

The compressed air idea might work well but of course it is not actually air in there it is fluorocarbons. May also may leave a residue from the bitterant on the LED and the buds.

Bridgelux says to clean cob surface with isopropyl alcohol. I've done that here a few times with no issue, it'll evaporate completely leaving zero residue and won't interfere with the integrity of the emitter.

Great product to have in the house in general too.
 

mtnstream

Active Member
Agree, q-tip dipped in isopropyl worked fine on cree cxa 3050.

Bridgelux says to clean cob surface with isopropyl alcohol. I've done that here a few times with no issue, it'll evaporate completely leaving zero residue and won't interfere with the integrity of the emitter.

Great product to have in the house in general too.
 

SupraSPL

Well-Known Member
Good to know that the alcohol is safe on COBs.

Regarding PWM dimming, is the info in this link describing the same style as arduino pwm dimming? It is pointing out the same thing as KNNA, lower efficiency of PWM vs current adjustment.
 

SupraSPL

Well-Known Member
Very cool dimmable driver for a 100W COB. 600mA-3.1A. Claims 93% efficiency, I'd have to test that to believe it. This would be very good for a diy photography light box.
 

0xc0ffee

Member
I can't for the life of me figure out why you're running such huge chips at such a low amperage. If you're only throwing 700mA at that cxa3070 (which maxes near 2800mA), wouldn't a CXA1520 or CXA1820 be a better choice? Hell even a CXA2011 should be out there for half the price of these 3070's, with about the same output as your under-driven lamp. I'm curious to hear why you chose this particular combination of such low output vs high potential.

Enlighten me sensai...

Edit: While I'm asking questions, why do you polish the underside of your heatsink? To me, it seems like you would just be re-reflecting light that the plants didn't take up back to them. Since LEDs are omnidirectional, I'd think that the only light that you'd have a chance to reflect would have already bounced off your plants and thus not absorbed from the useful spectrums.
 

SupraSPL

Well-Known Member
It comes down to up front cost VS efficiency. Running big chip soft results in a much higher efficiency than running smaller chips hard. 3 reasons, current droop, temp droop and lumen depreciation. So rather than crank up the amps to get the photons, I just add more chips to get the photons. This is especially important for lights that will be used for extended periods of time like grow lights or reef lights.

So the up front cost is higher, but the increase in efficiency is very large and results directly in more buds. For example, the CXA3070 at 700mA is 25W and 42.6% efficient. The CXA1820 at 700mA is 27W and 31.2% efficient. That means the CXA3070 creates 36.5% more buds for the wattage, amazing gains.

------------------------------------------

I polished the heatsinks to speed up heat transfer because heatsink surfaces are very far from flat. Was it worth the effort maybe not. It is a lot of work to polish, but the most important part is the first few steps. There is really no need to go all the way. Stopping at 600 or 1000 grit might be a good point of diminishing returns. Every aluminum star has this problem also.
DSC06848a.jpg
 

caretak3r

Well-Known Member

0xc0ffee

Member
It comes down to up front cost VS efficiency. Running big chip soft results in a much higher efficiency than running smaller chips hard. 3 reasons, current droop, temp droop and lumen depreciation. So rather than crank up the amps to get the photons, I just add more chips to get the photons. This is especially important for lights that will be used for extended periods of time like grow lights or reef lights.

So the up front cost is higher, but the increase in efficiency is very large and results directly in more buds. For example, the CXA3070 at 700mA is 25W and 42.6% efficient. The CXA1820 at 700mA is 27W and 30.8% efficient. That means the CXA3070 creates 38.3% more buds for the wattage, amazing gains.
Thanks for the answer. I dug into to the datasheets last night and came to the same conclusion. Even running that 3070 at only 700ma, Cree's luminous flux graph puts it in the neighborhood of 40%, based on their 1925mA control testing. This nets somewhere around 3500 lm compared to the middle bin 3000k 1820's 2379 lm at 550mA. Quick napkin math says the 1820 is going to require about 85% of the power vs your 3070, while only delivering about 67% of the lm. This is slightly skewed though, as you're putting 24.5w of power towards that 3070 vs 20.9w to the 1820. If we drop the 3070 to 550mA, lm ends up around 2600 or about 110%. Still more efficient than the 1820, but not by a ton.

On the other hand though, the 1820 is substantially cheaper in single quantities, costing almost a third of the 3070. Depending on how many you wish to run, the material costs can get out of control rather quickly. If you're only buying 2, the 1820 would set you back about $30, whereas the 3070 costs about $76. Ramp this up to 10, and we're talking about $150 vs $380. This ends up being a huge cost to overcome with regard to energy costs. Since we're only talking about 20w each, power ends up in the neighborhood of $0.31/month each for the 3070 @550mA and $0.34 for the 1820 @550mA (at my rate of 0.046 kWh). This puts ROI on a single 3070 somewhere around 766 months (or almost 64 years) past the 1820, to gain around 200lm.

All that said though, I'm super stoked about this project. I've lurked around here for a while, sponging up as much information as I could and have learned a ton by following your builds.

I polished the heatsinks to speed up heat transfer because heatsink surfaces are very far from flat. Was it worth the effort maybe not. It is a lot of work to polish, but the most important part is the first few steps. There is really no need to go all the way. Stopping at 600 or 1000 grit might be a good point of diminishing returns. Every aluminum star has this problem also.
I had this problem too. I'm not sure if the stars warp when the LEDs are reflowed or if they start off that way, but they're definitely far from flush. In my case, a couple of 6-32 screws helped a ton in getting as much contact between the star and the heatsink as possible. I can see how this is a worthwhile step on your builds though, since you're attaching them via thermal glue and tape. Plus, you get the added bonus of looking fucking awesome....I have a soft spot for shiny. :D
 

SupraSPL

Well-Known Member
0xc0ffee you are asking the right questions. here is where I am coming from.

CXA1820 vs CXA3070.png

My math comes up that the 1820 will dissipate 27.3W at 700mA versus 24.7W for the 3070. The 3070 makes 3450lumens and the 1820 makes 2793 lumens. So 139lm/W vs 102lm/W = 37% less buds/watt

So if your CXA3070 input power is 600W it would cost about $34/mnth and lets say yielded .4gr/watt in 4 weeks. That is 240gr @ $12 = $2880 worth of bud/mnth.
For the CXA1820 input power is 600W it would cost $34/mnth and yield would be .25gr/watt in 4 weeks. That is 150gr @12 = 1800 worth of bud/mnth.

So lets say we built a large enough CXA1820 lamp that it would match the yield of the 3070. It would have to be 960W and would cost $54/mnth in electric. It would require 40% more space, heatsinks, drivers, wiring, and hanging gear. It would weigh 40% more. It might create enough heat that you would need to kick on the AC and certainly you would need more airflow from circulation and exhaust fans which means more noise and more carbon required and quite a bit more electric cost lets call it 70/mnth.

As far as parts cost, the 600W CXA3070 would be ($40X20 COBs) $800 + ($12X20 drivers) $240 + ($42.5X10 heatsinks) $425 + $100 wiring/chain/shrink tubing/solder/flux/paste etc = $1565
For parts cost 960W of CXA1820, 800W($15X30) $450 + ($12X30 drivers) $360 + (42.50X16 heatsinks) $680 + 140 wiring/chain/shrink tubing/solder/flux/paste etc = $1630

So using my design, it actually costs more to build an equivalent output CXA1820 lamp, would require much more work and space and would be less stealthy. It would also cost 2X/month in terms of electricity. I calculated based on 15cents/kWh because you have to add delivery and supply cost. For those who are in areas of very expensive electricity this becomes even more important (hawaii, NYC, AUS, europe etc) Hawaii cost 40cents/KWh.

Considering these lamps could yield 240gr/mnth, the parts cost for the entire build is paid in less than a month and the monthly bill is $34/mnth after that. The lamp will never diminish in brightness or burn out like HPS.

---------------------------------------------------------

I agree the CXA1820 should not be run harder than 550mA in which case it is 34.2% efficient versus 31.2% @ 700mA. This would mean an increase cost up front but it is a positive economy IMO.
 

Gaius

Active Member
Thanks for all the info fellas!

I just started a new thread to follow my progress in building 3 DIY panels with the CXA3070s (just a few left on digikey FYI).

It'd be amazing if anyone has time to assist (or mock) me throughout this crazy endeavor.
 
You have me seriously rethinking my plan on running 18 of the vero 18s at .7 A on 2 hlg-185cs in favor of 16 of the cree.

Question for you, what size of space are you going to be running the 600 W in (5 x 5?), what type of grow (SOG, ScrOG, trees?) and how did you arrive at the .4g/4 weeks, vs the .25g/ 4 weeks?

I am looking into insight into how much power to run. With delivery mine is about .25$ a KW/h. Looking to do 4 large plants, topped after second node, with 4 main colas each, in a 5 x 5. I am wondering if 400 W (100 W over each plant) will cut it.

Thanks again for the great posts, I will be following this closely!


0xc0ffee you are asking the right questions. here is where I am coming from.

View attachment 2983044

My math comes up that the 1820 will dissipate 27.3W at 700mA versus 24.7W for the 3070. The 3070 makes 3450lumens and the 1820 makes 2793 lumens. So 139lm/W vs 102lm/W = 37% less buds/watt

So if your CXA3070 input power is 600W it would cost about $34/mnth and lets say yielded .4gr/watt in 4 weeks. That is 240gr @ $12 = $2880 worth of bud/mnth.
For the CXA1820 input power is 600W it would cost $34/mnth and yield would be .25gr/watt in 4 weeks. That is 150gr @12 = 1800 worth of bud/mnth.

So lets say we built a large enough CXA1820 lamp that it would match the yield of the 3070. It would have to be 960W and would cost $54/mnth in electric. It would require 40% more space, heatsinks, drivers, wiring, and hanging gear. It would weigh 40% more. It might create enough heat that you would need to kick on the AC and certainly you would need more airflow from circulation and exhaust fans which means more noise and more carbon required and quite a bit more electric cost lets call it 70/mnth.

As far as parts cost, the 600W CXA3070 would be ($40X20 COBs) $800 + ($12X20 drivers) $240 + ($42.5X10 heatsinks) $425 + $100 wiring/chain/shrink tubing/solder/flux/paste etc = $1565
For parts cost 960W of CXA1820, 800W($15X30) $450 + ($12X30 drivers) $360 + (42.50X16 heatsinks) $680 + 140 wiring/chain/shrink tubing/solder/flux/paste etc = $1630

So using my design, it actually costs more to build an equivalent output CXA1820 lamp, would require much more work and space and would be less stealthy. It would also cost 2X/month in terms of electricity. I calculated based on 15cents/kWh because you have to add delivery and supply cost. For those who are in areas of very expensive electricity this becomes even more important (hawaii, NYC, AUS, europe etc) Hawaii cost 40cents/KWh.

Considering these lamps could yield 240gr/mnth, the parts cost for the entire build is paid in less than a month and the monthly bill is $34/mnth after that. The lamp will never diminish in brightness or burn out like HPS.

---------------------------------------------------------

I agree the CXA1820 should not be run harder than 550mA in which case it is 34.2% efficient versus 31.2% @ 700mA. This would mean an increase cost up front but it is a positive economy IMO.
 

SupraSPL

Well-Known Member
You are correct, the CXA3070 should outperform the Vero18 and Vero29 slightly in terms of performance and price.
Vero 18 29 VS CXA3070.png

The tent is about 7'wide 8'long and 7'tall. The canopy gets wider during the first 2 weeks and this time ended up about 5'X5' as you guessed. For now it is 300W of LED and 600W of HPS. Once I get the rest of the modules finished I plant to replace the HPS with another 300W of LED.

The other flower room is 8'X8'X7' and has a 600 HPS + 200W of diy LED. Same story there, once finished it will be 600W diy LED. Half RWB half 3000K COBs. The rooms are running flip flop to help control heat and noise.

The plant style is lollipopped trees 3' to 4.5' feet tall including buckets. Vertical space is limited because there is a ceiling fan above the canopy for quiet circulation (55W).

The VEG/cloning tent is 8'X8'X7' and is currently 80W of diy LED and 1W for cloning. Once I update the vegging LEDs they will be down to 60W. It also has a pair of diy carbon filters hanging in there. All 4 tents are connected by huge light traps so the pair of carbon filters ventilates the whole thing (tent 4 is dark room for trimming drying etc). The carbon filter fans are Max Fan 6" which adjustable so I have 6 choices of strength depending on the heat situation. Since I scrapped a pair of 600HPS I have been able to turn the fans down to low so they consume a total of 120W and are nice and quiet. Once I scrap the other pair of HPS I might be able to turn one fan off and run at just 60W.

I am looking into insight into how much power to run. With delivery mine is about .25$ a KW/h. Looking to do 4 large plants, topped after second node, with 4 main colas each, in a 5 x 5. I am wondering if 400 W (100 W over each plant) will cut it.
Wow that is a hefty electric rate and I expect rates to rise over time which is one of the reasons I am so obsessed with efficiency. To give you some idea, in my 5'X5' I have 8 buckets with 14L of soil in each, definitely on the low side of wise. I used to run 18L in each and will probably go back to that as soon as I get some extra soil cooked. Since I have a 4-5 week flip flop I have to veg up 8 to fill the space. If I tried to do 4 larger ones I suspect I would run out of vertical space but it may be different in your case. I agree with your plan of topping them. I top them, bend them and tie them as necessary to try and keep a somewhat even canopy. 400W should work great and if not you can always add more modules later. Maybe the 3590 will be available by then.
 

SupraSPL

Well-Known Member
how did you arrive at the .4g/4 weeks, vs the .25g/ 4 weeks?
I figured those numbers as a minimum in case someone is growing non commercial varieties (like the ones I love). .8gr/W/8 weeks should be do-able and a lot more for commercial varieties. Since we are talking about cost/month I just broke that number in half and called it .4gr/W/4 weeks. I applied the penalty to the CXA1820 and thats how I got .25gr/W/4 weeks number.

Last run I only got .56 but that was including a 600HPS with a very old bulb so I won't have conclusive data until I get and all LED crop. I do have other useful data points though, in the other tent a pair of 600HPS with very old bulbs (18 months) gave me only .3gr/W/8 weeks, very consistently using the same varieties and all else being equal. So I suspect that the LEDs more than doubled the yield of the aged HPS bulbs. At the end of the day I saw my numbers go up when I scrapped the HPS for the LED, even though it is half the wattage.
 
Top