Cree DIY COB LED? Why do people use white lights?

[Atulip]

Typically watercooling is only nominally better,take for instance the closed loop cooling craze that went on with pcs, for the same price one can get a Noctua NH-D15 that would outpreform most clcs. A good heatsink plus a high static pressure fan almost always does better for cheaper.
Now back to the topic, I have been looking at the crees and it seems like the cxb2530 performs quite well for the price, keeping in mind i have 9 drivers on the mars ii 1200 meaning that would be 18 lamps. Going with either the 3050 or the 3590 would be almost prohibitively costly in the most expensive case costing nearly $900.

Consider the 72v 3590, 1 per driver. For $40-45 per, you get 56% efficiency. 2 x 3070 are 59%, 2 x 3590 36v are 64% efficient at .7A. All top bin.

And consider the driver waste as well. The mars drivers are only ~85% efficient, compared to say a 95% mean well. So a 56% efficient cxb3590 at the wall turns into 48% on a Mars driver, 53% on a meanwell.

 

[Atulip]

I don't know if anyone mentioned spectrum under these white cobs.

Pretty sweet for our purposes.

 

[TheEEngineer]

I had not really talked about efficiency in my post however that is one of the driving factors for not wanting to spend the much higher cost for the 3590s because I in in the works of designing my own light in CAD currently which will utilize much more efficient drivers and a better overall layout for use with COBs. On another note I am looking at something that seems like for the minimal investment would greatly increase the power of the stock light. That is a CXB2530 3000k 90CRI, 18 of those would fill out the panel quite well.

 

[Atulip]

Yeah that's my go to recommendation for most mars upgrades. At 24 bucks per driver it's hard to beat. Should still be ~1.8umol/j at the wall after upgrade. A nice jump up from maybe 1.2umol/j from a stock Mars.

 

[TheEEngineer]

The reality is however that the chips that ship with the board are quite cheap, any cree will almost beat it in umol/j, not to meantion that the 3000k 90cri are a much better spectrum for flowering. The results are quite tangible.

 

[jaymu86]

The reality is however that the chips that ship with the board are quite cheap, any cree will almost beat it in umol/j, not to meantion that the 3000k 90cri are a much better spectrum for flowering. The results are quite tangible.

Yea I agree that's why I'll mix in 90cri and the 80cri good mix

Sent from my HTC Desire Eye using 420

 

[TheEEngineer]

Also thinking about potentially putting in 5x 72v 3590s for now making a box with a dot in the center essentially. If calculations are correct it should output about 50k lumens.

 

[TheEEngineer]

Actually about to pull the trigger on 7x CXB3590 72v at 3000k 80cri at $41 per. I dont mind not having the holders as soldering does not bother me at all, which saves me about 5 per light. However I am not sure if I should get reflectors or not.

Edit: I was checking voltage output on the drivers and zero load is 94v, obviously once there is a load present the voltage will drop but I did not figure they would be that high without a load.

2nd Edit: I was wondering, would 18x cxb2530 3000k 90cri be better than 6x 72v cxb3590 3000k or 3500k at 80cri. From what I understand the 72v cxb3590s are ~2.2umol/j while the cxb2530 is ~1.8umol/j so on paper it seems that the cxb2530 should be better however I cant find the efficiency of the cxb2530 thus it leaves that out. Adding that in I would think that the cxb3950s would be the overall winner.

 

[415friendly]

This seems to be the right place to ask. I have been experimenting with cheap LED's and drivers before I spend a bunch on acquiring some quality parts.

I picked up a cheap 100W power supply, a 150W boost converter, and a couple of cheap 100W LED's. I boosted the voltage to 33V, wired the LED's in parallel, but was surprised that I could only get .8W draw out of each of them; I had thought that they would pull the maximum current that they could from the circuit. I was expecting readings of about 1.5A on my ammeter.

I'm guessing now that it is because the power supply isn't really kicking out 100W. When you guys talk about power supply efficiency, assuming that it is 80% efficient, then it is only capable of supplying 80W? Am I losing additional power because the LED's are only able to draw a certain percentage of the power they are rated for, or should I be looking at the boost converter or wiring?

It's been a long, long time since I studied electrical circuits in Physics 12.

 

[Atulip]

You're limited by your voltage. Download the spec sheet for the chips you're using.

Cxb3590 for example

 

[415friendly]

Oh... These LEDs were rated for 35v Max, but I think I see what you mean... As the voltage begins to max out there was a voltage drop in the power supply as well. Would there be a voltage drop from the boost converter too?

 

[Atulip]

They pull different voltages at different currents. You need a Constant current driver with enough volts to power the chips at that current. You're limited to whatever the current would be at 33v on that chip, which is probably .8A

I'm not too familiar with boost/buck/more advanced electronics, just leds.

 

[TheEEngineer]

So 6x 72v cxb3590 3500k @80cri or 18x 36v cxb2530 3000k @90cri, keeping in mind I would likely add in three more cxb3590 at some point in time to fill the panel. Also the mars ii have a very different internal structure than the earlier ones did, heatsinks seem to be better.

 

[jaymu86]

Actually about to pull the trigger on 7x CXB3590 72v at 3000k 80cri at $41 per. I dont mind not having the holders as soldering does not bother me at all, which saves me about 5 per light. However I am not sure if I should get reflectors or not.

Edit: I was checking voltage output on the drivers and zero load is 94v, obviously once there is a load present the voltage will drop but I did not figure they would be that high without a load.

2nd Edit: I was wondering, would 18x cxb2530 3000k 90cri be better than 6x 72v cxb3590 3000k or 3500k at 80cri. From what I understand the 72v cxb3590s are ~2.2umol/j while the cxb2530 is ~1.8umol/j so on paper it seems that the cxb2530 should be better however I cant find the efficiency of the cxb2530 thus it leaves that out. Adding that in I would think that the cxb3950s would be the overall winner.

I believe so just because of the spread of lite and will fill space better in turn more lite over each plant

Sent from my HTC Desire Eye using 420

 

[jaymu86]

So 6x 72v cxb3590 3500k @80cri or 18x 36v cxb2530 3000k @90cri, keeping in mind I would likely add in three more cxb3590 at some point in time to fill the panel. Also the mars ii have a very different internal structure than the earlier ones did, heatsinks seem to be better.

Are you putting these in a mars unit if so go with the 3590's cause 18 of those in one of them is probably alil much

Sent from my HTC Desire Eye using 420

 

[TheEEngineer]

The mars ii 1200 has three heat sinks each 16 inches long, so it would be essentially one every 2.5 inches. I think either way I will be going with the 3590s just because its more efficient and provides more light.

 

[TheEEngineer]

Now the current problem, removing the glass from the case. Anyone have any ideas other than breaking it?

 

[Atulip]

Silicone stuff all around it? I used a razor blade slid in from the front to free it.

 

[TheEEngineer]

I tried that and no luck, mine feels like it has glue on it as well. Ill try again here shortly.

 

[OlderGrower]

While this has been an interesting read, I really did not learn much about what the thread was posted as. "Why use white light" I see a benefit of being able to do plant maintenance without having to turn your lights off or to take photos, but other then that I can not really see any other reason to grow with them. I keep hearing about what great penetration they have but in all honesty, I look at the grow journals of people using them and the plants look no better or worse then one growing under a Mars or any other LED. I must be missing something.