DIY LED: Spectrum and 3W or Not?

Alright, well first let me say hello. This is my first post.
Second, these questions have been asked before, but I want to see what the current consensus on them are.

I want to build a large number of small wattage LED grow lamps. I have determined that the lowest cost per watt is my only real constraint. I decided that I do not care as much about the "quality" of the LEDs, but the "quantity." Everyone has their own opinion of this, but if I can build twice as many units at the same cost, I will recoup the difference in output.

In order to do this I think I want to use 3 colors, which is my current design using 455nm, 620nm, and 660nm lights. Would it be better for me supplementing these 3 waves lengths with Cool White, Neutral White, Orange, UV, etc?

The end objective being cost, I currently believe my best option is to purchase the LEDs in bulk at a low cost. 3W LEDs appear to be the easiest to source in bulk (can be found on ebay), plus are more efficient cost per watt then smaller LEDs. My question is there any benefit to going to larger LEDs when considering their lower cost per watt may be negated by the increased cost in driving the 5 or 10W models?

Those questions aside, I have a basic design already. I am using the cheapest 3W LEDs I can find and an AC input, 60W driver. A simple wind tunnel heat sink and a couple small fans.

Here is my basic design, I would build many of these units:

10 - 3W Red LED 620nm, 2.0-2.5V, .8A, 60-80lm, 120deg
10 - 3W Red LED 650nm, 2.0-2.5V, .8A, 70-80lm, 120deg
6 - 3W Blue LED 455nm, 3.2-3.6V, .75A, 40-50lm, 120deg
1- .68A LED Driver, 60W, AC in 120V
2 - VGA fans, 12VDC, 22.8cfm, 4 cm x 4cm

Very simply, I will make a tunnel out of two pieces of Aluminum channel. I will then mount LEDs on the bottom of the tunnel. A fan on both ends in a push-pull configuration will blow air down the tunnel for cooling.

By my calculations, pushing .68A through the will make 45W. I believe I can make this entire unit for around 30 dollars. Those are real watts, and a price of around 67 cents a watt seems to be beatable only by purchasing made panels from china at high wattage. I should be able to place these throughout my grow instead, mount them in any direction and basically bombard my plants from all angles.

So what do you think?

Thanks ahead of time guys, I have really enjoyed reading your threads and forums.

In your situation you would probably be better off going china direct. Just make sure the fixtures are PROVEN to work. I remember seeing some good grows with eshine. Last I remember they were around a $1/w. Buying in bulk would probably get it even cheaper.

Research the gpw sticky and retailer sticky.

If you do still want to build it you would have to prototype it and test. You wouldn't want to go through all that trouble to grow schwagg. I don't think your design would need two fans either. You could cut the fans in half..or put two fans pushing. I think you would be wasting a lot of air pulling.

I'm sure someone else will chime in and give their 2 cents

Looks pretty cool, have you seen Rives, "Volksled - Diy LED for the masses"? Bet you could find something in there.

My only problem is no 500-600nm, I like a little fuller spectrum even in veg. [i.e. Emerson effect]

How much will the spectrum change with cheap diodes. I have been reading recently about spectrum changes due to heating and it quickly changes even in "higher" quality diodes.Spectrum changes due to heat haven't been documented very well, mainly because, who has $50k for the photospectrometer.

One thing that stands out is the lumens. I know that it isn't the greatest factor because it's what the human eye can see, but 1750 lm [+-50] @45w, doesn't seem like a very efficient light even at real watts. [Or are you adding in your fan too, etc?]

I have been eyeing Bridgelux Vero18's.

You could go even cheaper with other multichips. Why not run a couple of 10w multi's and one or two oddball 3w to even out the spectrum? or even 2 20w multi's. One CW and one WW. Still could add some 3w NW's, to round out the spectrum too.

5pcs 10w WW multichips Ebay Cheap @1100lm each.

I want to run light bars with accent lights. The accent lights could be built for cheap!

How about 30w [actually 25.4watts @ 800ma x 32V ] Accent lights? Probably about $35-40 to make each. Fan - heatsink - driver - Diode.

30W Multichip CW and WW varieties/ $18 Ebay


Arctic PSU Heatsink/Fan $7 Newegg

Driver and Paste for $10-15.

Although, the spectrums are quite as "robust" yet in the multichips, but the cheaper ones, seem to have a good variety.

Cheaper Yet.

50w WW Multichip w/ Driver $16

+ a cheap cpu heatsink/fan....$25 dollar light.

I think i definitely need to look into the higher powered leds. I guess then I am still concerned about the spectrum I would choose to cover.

My understanding is that the main benefit of LEDs was the ability to focus on the specific spectrum which activates photosynthesis.
What are the benefits of switching to full spectrum white lights? I was planing on only supplementing the wavelengths produced by the red and blue LEDs, not produce then entire visible spectrum.

Now Relax said:

My understanding is that the main benefit of LEDs was the ability to focus on the specific spectrum which activates photosynthesis.

Click to expand...

The best way I can explain it is like this....YES! .....with buts'...


Starting light graph shows a bunch of different lights. Metal halides are known for vegging because of their blue spectrum. Most people have to buy special bulbs to flower or supplement with MH. HPS isn't on here, but has a curve like tungsten, which actually probably tops out around 3200K, not quite as high as on this graph.
So then you have chlorophyll which wants both red and blue, not just red and not just blue per say and maybe even a few things in between, some aquamarines, some greens, some yellows, some invisible light.

[Cannabis is light hungry, even in veg, it will consume a "fuller" wavelength if offered]

LED's can be built however you want. Spectrums are like opinions, everybody's is usually different. To an extent.
You just have to pick according to your budget and when All parameters line up, current, voltage, etc.


Here is photosynthesis and chlorophyll a&b peaks:
this shows the usual R&B peaks that are sought out...there is a "dead" trough starting <500nm and ramping up just after 600nm...


....but here is PAR...where there is not a trough, more a gentle curve [<500 - >600] + other peaks in blue and red with invisible light as well in uv and ir.


Here is one more, that can only be called experimental @ this point. I believe it is the spectral irradiance of the sun, but at what parameters. It gives an interesting breakdown on the % of spectrum each occupies. Very interesting! I can't track down the label.....so until further......

What are the benefits of switching to full spectrum white lights? I was planing on only supplementing the wavelengths produced by the red and blue LEDs, not produce then entire visible spectrum.

Click to expand...

I did put up the second graph on chlorophyll which is represented by Kelvin temperatures, this may give an idea on white light spectrums.

I have been reading that less bleaching seems to be occuring with more white, but WW that have reds in the 652nm [660nm] are hard to find, most seem to peak around 640 or so.

I think in theory, using a fuller spectrum with whites could provide more harvestable PAR, according to the PAR charts compared to just chlorophyll levels.

Quality post right there, Oregonism!

It's too bad that there isn't much scientific data on the topic BUT in theory the white LEDs are added to provide a small dosage of full spectrum (350-750nm) apart from the red/blue peaks that are heavily supplemented.

CW comes in handy if you have small LED lights and want to supplement just a tad of blue (peaking at 450nm that is), while WW provides more orange (580-600nm) and far red (700+). IMHO in almost all cases it's better to for CW because WW peaks at ~600-620 and you're using 630/660nm LEDS anyways.
(The CW vs WW debate in non all-white lights shouldn't play such a big role, this is probably negligable.)

I think the diodes you get in one of those big 10-100 watt multi-diode LED things are generally lower quality than single diodes, but a 100-watt LED has 100 actual diodes built into it. If I had to build a light from scratch I would get a powerful driver for some 50-100 watt white leds and some less high powered drivers for some red or deep red and royal blue LEDs and put them all on the same big heatsink. The 50-watt LEDs I've seen run at 25V, 1500mA each, I think the 100-watt ones are like 20V at 3000mA each. So you need a different LED driver for those. In my experience, generic Chinese LEDs are getting pretty close to the quality of CREE and Philips, Does CREE or Philips make 50-100 watt multi-diode packages like you can get from China?

In terms of efficiency the chinese stuff is nowhere near the big brands (Osram, Cree, Philips), this I can assure you
You have to build multi diode packages youself if you want Cree/Philips etc.

All UFOs I've seen yet are low quality, the diodes are lower quality and run at high currents (negating possible efficiency), on top of that heatsinking is just a sheet of metal.

I'm talking about these things

http://www.aliexpress.com/item/Free...White-4500LM-50-Watt-LED-Lamp/1154606646.html

4500 lumens is a lot. This particular one is rated as 32-34V 1500mA, so a 65-70V 1500mA constant current power supply would power 2 of these (hard to find maybe, but you can find 50 Watt LEDs with a lower forward voltage, like 25V ones that are also about 4000 lumens) Cool whites are rated at about 5000 lumens generally. you could probably power 22 CREE XM-L's with a power supply like that, The best possible cool white XM-L is rated at 590 lumens at 1500mA, 433 lumens is the highest output of the best warm white XM-L at 1500mA,

12980 Lumens from 22 cool white CREE XM-L's vs 10000 lumens from 2 50-watt cool white Leds

9526 Lumens from 22 CREE warm white LEDs vs 8000-9000 lumens from 2 50 watt warm white LEDs

I think the 100watt ones might even be a little better.

Assuming those numbers on aliexpress are correct, those are almost as good for less money, but yeah, not as good. CREE XM-L's are at least $5 each, so at least $110 for 22, 2 50-watt leds are like $6-12 each, so like $12-24 for almost the same output. The light is less distributed that way though.

tebos said:

...on top of that heatsinking is just a sheet of metal.

Click to expand...

not always the case

http://www.lightput.com/forum/viewtopic.php?f=4&t=837#p955

Lumens aren't the correct unit for measuring their photosynthetic value BUT if you want to compare them this way, here you go: the china LED from your link is at 90lm/W (assuming 4500lm are correct), XM-L2 are easily outputting 150lm/W and that's a major difference in quality. Though Cree is offering high voltage LEDs which may be a better partner for comparisons like this.

But yeah, they are more expensive, you get what you pay for.

For some reason I just don't see the point in going all white. I think a better design will focus on the peaks of PAR, yet add a few whites to cover the dip per se.

Back to the lab! Talk to y'all again a few days.

That's what I'm thinking, too.

Oregonism said:

Here is one more, that can only be called experimental @ this point. I believe it is the spectral irradiance of the sun, but at what parameters. It gives an interesting breakdown on the % of spectrum each occupies. Very interesting! I can't track down the label.....so until further......

[URL=https://www.icmag.com/ic/picture.php?albumid=48319&pictureid=1127626&thumb=1]View Image[/URL]

Click to expand...

I was thinking about this earlier. Our Sun is a main sequence star, which means that it derives its energy from the fusion of hydrogen basically.

This graph looks similar to the spectrum analysis of hydrogen.

Perhaps the peaks in PAR, are due to peaks in the suns output due to the abundance of hydrogen burning?

It would make sense that any and all plants would arrive to fruition if their cells most efficiently processed the light available to them. Meaning slight peaks in the sun's output would provide a slight advantage in evolving.

Now Relax said:

For some reason I just don't see the point in going all white. I think a better design will focus on the peaks of PAR, yet add a few whites to cover the dip per se.

Back to the lab! Talk to y'all again a few days.

Click to expand...

I look at it a little differently

If an Area 51 with mostly white produces similarly high gpw as uber efficient Hans panel. Why not use high powered whites which have better penetration and lumens per watt. Ever see Hans panel grow a big plant? I haven't..

And in a reverse from conventional thought, use the reds to enhance the quality, in a correct balance. It wouldn't take much testing to figure out how much red it takes to improve quality without going overboard.

Two of the main players Area 51 and apache are going along this line of thought. I'm sure they have done similar testing with reds to come to the conclusions they have.

If I were looking to build the cheapest, best panel I could. It would use 2 vero 18 neutral white, which I think runs at 33w. And in the middle a 30w China multi chip red. Three leds per fixture. I think any more watts than 30 in one spot is a waste of light. Should be lots of light for a 2 x 2 area. To use any lesser quality diodes would probably result in low quality

Just my opinion. I have been doing a lot of led research lately and these strike me as one of the better bangs for your buck.

Now Relax said:

For some reason I just don't see the point in going all white. I think a better design will focus on the peaks of PAR, yet add a few whites to cover the dip per se.

Back to the lab! Talk to y'all again a few days.

Click to expand...

This is my take away to this point.

Whites seem to provide good spectrum without really hitting the "peaks" we deem crucial, but certainly create more than adequate light throughout the spectrum [visible/invisible too]. While specific colors create specific adjustability of sorts within those peaks.

One thing that stands out about cheaper color diodes is some run-ins with non flowering lights like I believe Postivity has experienced.
I just read a KNNA archive where he was discussing blue and reds as they relate to current and voltage. What is interesting is is that blue stay fairly reliable with adjustment to either, while reds just drop off with heat and voltage spikes. Seems like maybe the problem is more cheap chinese drivers not holding their heat or voltage well. But my theory is that these cheap fixtures pump out the blue, but when they get hot the red plummets. That might not mean anything about diodes as I get into my theory, lol. But the gist was.....


How about using cheaper Cree XPE bins or something along those lines. Both specific colors and whites are offered.

How about just running cheaper bins of XPE 5w or 10w star arrays and building your own drivers? I think price wise you could compete within a few $5 dollars of your original goal + have a fixture that would compete with most current offerings in that size and kept cool will last 50 months or so at least.

positivity said:

I look at it a little differently

If an Area 51 with mostly white produces similarly high gpw as uber efficient Hans panel. Why not use high powered whites which have better penetration and lumens per watt. Ever see Hans panel grow a big plant? I haven't..

And in a reverse from conventional thought, use the reds to enhance the quality, in a correct balance. It wouldn't take much testing to figure out how much red it takes to improve quality without going overboard.

Two of the main players Area 51 and apache are going along this line of thought. I'm sure they have done similar testing with reds to come to the conclusions they have.

If I were looking to build the cheapest, best panel I could. It would use 2 vero 18 neutral white, which I think runs at 33w. And in the middle a 30w China multi chip red. Three leds per fixture. I think any more watts than 30 in one spot is a waste of light. Should be lots of light for a 2 x 2 area. To use any lesser quality diodes would probably result in low quality

Just my opinion. I have been doing a lot of led research lately and these strike me as one of the better bangs for your buck.

Click to expand...

You almost describe my ideal light bar. Really digging bars over fixtures for coverage and multiple points of source. I think they might be a tad shallow for 2x2 flowering. That's less than 25w/sqft. I think you would want maybe 40-50. but veg would undoubtedly knock it out. Channel dim on the WW and you are rocking.

Now Relax said:

I was thinking about this earlier. Our Sun is a main sequence star, which means that it derives its energy from the fusion of hydrogen basically.

This graph looks similar to the spectrum analysis of hydrogen.

Perhaps the peaks in PAR, are due to peaks in the suns output due to the abundance of hydrogen burning?

It would make sense that any and all plants would arrive to fruition if their cells most efficiently processed the light available to them. Meaning slight peaks in the sun's output would provide a slight advantage in evolving.

Click to expand...

That is an interesting insight NR. Who says it doesn't change either as in which peaks are they seeking? Plant evolution almost certainly didn't have static levels of PAR harvesting throughout history.

That said, I really posted it, to point out, how much of the spectrum was actually invisible @ >50%!

I really enjoy seeing people experimenting with the invisible spectrums on both sides, UV and IR. I think it is frequently left out of the conversation. Main reason: our puny brains can't see it, so we easily dismiss it!

Definitely on the low end of w/sqft, but if your making a whole bunch of them you could put them a little closer together and take advantage of the overlap.

I'm going the other direction. 300w of xml, Philip deep red, semiled uv. $1000. Big cost but should be a once and done kinda thing. Quality, power, and efficiency all wrapped up on one large heatsink. Then hopefully, I can take a nice long break and enjoy the finer parts of this hobby...

Oregonism said:

I just read a KNNA archive where he was discussing blue and reds as they relate to current and voltage. What is interesting is is that blue stay fairly reliable with adjustment to either, while reds just drop off with heat and voltage spikes. Seems like maybe the problem is more cheap chinese drivers not holding their heat or voltage well.

Click to expand...

This has nothing to do with the drivers actually, blues are more efficient because the phosphor layer ontop of the LED DIE is thinner (=letting more light through it) and red having a thicker layer (=less light output), also reds are more sensitive to ambient temperature rise while blues are fairly stable and higher current means also more heat.
Those two factors contribute to the (in)efficiency at higher currents of red/blue LEDs.