DIY LED Fixture - Pointers for future build

Slightly excited so I'm reviving this to show some testing-

Citizen CLU028 97CRI 4000k Neutral White (11.4 watts at 38volts, 300mA)

Luxeon Lime

Luxeon K16 Royal Blue (easy to find all over FleaBay, AliExpress, and similar sites, these are the real deal, just surplus) (for Toshiba remote phosphor fixture) these can do 45 watts of dissipation, for 23,000+mw of radiance!

Lashed up really quick on a RapidLED Premium Heatsink, white/lime driven in series at 300mA, royal blue at 700mA.




These suckers are very, very powerful, and the colors are fantastic. It has a very subtle hint of that pink/magenta look that will be easy to counteract with the other colors I will be adding to the array. The overall look is pretty close to a 12,000k halide, and much less pink than the picture shows-



Pardon my potato quality picture, it ended up very dark for some reason, and the color isn't true at all...


The plan is so far to add 2x violet, 4x cool/regular blue, 1 or 2x cyan, and build two total clusters, and it will end up over a 40breeder or similar size setup for a "display" frag setup. Even as a standalone setup it looks phenomenally better than the basic cool white/royal blue combo, and would do very well as an array on its own for a warmer desired look. This also blows away the "dream chip" types of multichips in cost and efficiency, as well as having nicer color, even unsupplemented.

The best part? Those K16 arrays are only $5 each shipped, and the Citizen whites are less than $8, although you could do similar quality with less CRI if you went with a Bridgelux Vero for $3.75 each, and end up with a very economical base set to build on with your other colors.
 
Last edited:
Slightly excited so I'm reviving this to show some testing-

Citizen CLU028 97CRI 4000k Neutral White (11.4 watts at 38volts, 300mA)

Luxeon Lime

Luxeon K16 Royal Blue...


...The plan is so far to add 2x violet, 4x cool/regular blue, 1 or 2x cyan, and build two total clusters, and it will end up over a 40breeder or similar size setup for a "display" frag setup. Even as a standalone setup it looks phenomenally better than the basic cool white/royal blue combo, and would do very well as an array on its own for a warmer desired look. This also blows away the "dream chip" types of multichips in cost and efficiency, as well as having nicer color, even unsupplemented.

The best part? Those K16 arrays are only $5 each shipped, and the Citizen whites are less than $8, although you could do similar quality with less CRI if you went with a Bridgelux Vero for $3.75 each, and end up with a very economical base set to build on with your other colors.
How is the blending on this smorgasbord when its all together?

I always thought this would make a super inexpensive build for really big tanks where you want to push out lots of light per LED grouping.
 
How is the blending on this smorgasbord when its all together?

I always thought this would make a super inexpensive build for really big tanks where you want to push out lots of light per LED grouping.

Well, 6" above the waterline, and over my 12" tall 20 long it blends well. Once the other colors are added I can comment further, but I may end up using a diffusion panel of some sort so I don't have to hang the light so high.
 
Last edited:
Another bump here with a tidbit/teaser of the build mentioned earlier-


image uploader


Just need to wire up the drivers and associated bits. Getting there slowly! Can't wait to test it out!
 
Heatsink- RapidLED 20" premium enclosure

Two clusters of the following emitters each-

1x Citizen CLU028 4000k, 97 CRI neutral white
1x Luxeon K16 royal blue
4x Luxeon Rebel ES true cool blue (470-480nm)
2x Luxeon Rebel ES cyan (one each from rapidLED, one from Steve's, not that it makes much difference)
2x SemiLEDs Hyper violet 3.0 (U70 bin, 415-425nm)
1x Luxeon Rebel lime

Total LED cost is just under $100 or so for both clusters, not factoring in shipping.

The plan so far is to drive most everything at 700mA, and the Citizens at 300mA.

The funny thing is, that I don't really have a tank in mind for it yet! I've got a 20 Long reef that I've been toying with upgrading to a 40 breeder, and I might just do it now. I'm thinking one of these rigs would be perfect over a 40 breeder or maybe even something bigger?

Maybe a 57? Those weird things with the 36x18 footprint? No idea. Either way I'll dim it down and run it on the 20 long and see how it does.
 
Last edited:
Ignoring the drivers, control, fan, etc, the LEDs themselves on both channels will pull-

Royal blue- 61.6w (700mA, 44v, x2 strings)
White- 22.8w (300mA, 38v, x2 strings)
Blue- 18.3w (700mA, 13.08v, x2 strings)
Cyan- 8.1w (700mA, 11.6v)
Violet- 5.2w (700mA, 14.8v)
Lime- 3.78w (700mA, 5.4v)

All combined is 108w total, and 84.4w will be on the white/royal base colors. White/royal will be on a 48V PSU with LDD-H drivers, everything else will be on a 19V PSU and run off of LDD-L drivers. This is of course assuming everything is running full blast, and it's not likely that I would run everything that high other than a few channels for tuning color. Each color is split out into its own channel, for six channels of adjustment. Should be pretty rad :thumbsup:

Might get a little tight for wiring, I really could use a PCB that had 8 drivers on it that are functionally compatible with the LDD series, but the only source I know (O2surplus) has gone dark, so I'm trying to shoehorn everything in the fixture as is.

I'll try to tally up a parts list with prices soon. It's surprisingly cheap so far.

Should I start a new thread? This kind of snowballed into it's own project!
 
Last edited:
5 pennies from designing a LED fixture:
1. make it is in standard way:
all drivers should be the same, like all of them 1000mA or all of them 700mA. It will save tons of time when one channel burn out.
2. About spectrum: somewhere here it was a nice article, in a short: usually 80% royal blue, cyan +red + white = 20%.
I'm actually constantly changing the spectrum because all day blue water is boring - we are enjoying the color of corals, right? So let's change the light to make it nice.
3. From datasheet they usually note the maximum current. So if it says 700mA - that means 700mA maximum if temperature of heatsink less then 60%. In other words if I have driver for 700mA I will chose LED for 1000mA just to have some buffer.
4. Heatsink should be big enough to release the heat from LED. So the heatsink should stay warm. Trying to minimize the size will require active air movement, that means noisy cooler.
For each cooler they usually note a nose in dB. Would say > 20dB is unacceptable. Also the size of cooler - the bigger the size the more air flow and less nose.
5. They recommend 5W per gallon. to reduce the light by dimming it is a 5 minute job, but to add some power is a big trouble. So it is better to have a buffer -the bigger the better. At least to make picture - nobody likes blue picture, so this 20% of red+cyan+white should be enough to make a nice daylight picture.
 
5 pennies from designing a LED fixture:
1. make it is in standard way:
all drivers should be the same, like all of them 1000mA or all of them 700mA. It will save tons of time when one channel burn out.

Sort of agree. Especially if everything is to be adjustable, it is a lot easier if they are all the same. Socketed drivers (like the LDD series) are also a godsend for troubleshooting purposes. Also makes it easy to simply swap to a lower current driver if you wanted to for better dimming resolution at lower drive currents.

2. About spectrum: somewhere here it was a nice article, in a short: usually 80% royal blue, cyan +red + white = 20%.
I'm actually constantly changing the spectrum because all day blue water is boring - we are enjoying the color of corals, right? So let's change the light to make it nice.

Sort of agree. Red can be overpowering even in small amounts with individual emitters. That's one of the driving reasons behind choosing a high CRI base white. Looking at a warmer white in the 3000-4500k range that is high CRI will naturally have better color rendition than a cooler white usually will. If we already have adequate red spectrum in our white, we don't have to "add it back" by using separate reds for most applications.

Cyan is a good thing, and in conjunction with lime and regular blue helps counteract the pinkish color many of these setups get when using so much royal blue and warmer whites or red supplementation. Cyan is at the edge of a good photosynthetic sweet spot, as well as regular blue, so definitely worth having. Added benefit is the light will appear "brighter" to the eye.

Regular blue is useful for photosynthesis, and for bringing out colors that royal blue can't. Definitely worth having along for the ride. Too much can look like Windex though. Standard cool and neutral whites are usually lacking in both cyan and regular blue (but abundant in royal blue due to the base LED "showing through") so adding them is useful.

3. From datasheet they usually note the maximum current. So if it says 700mA - that means 700mA maximum if temperature of heatsink less then 60%. In other words if I have driver for 700mA I will chose LED for 1000mA just to have some buffer.

Definitely. You encounter some diminishing returns once you push them too hard anyway. Going with 70% of the max current is usually a good starting point. For Luxeon series LEDs in particular 700mA is a great point for longevity and efficiency. Going from 700mA to 1A isn't as useful for the near 50% increase in current you only get like ~20% increase in output at much more heat.

4. Heatsink should be big enough to release the heat from LED. So the heatsink should stay warm. Trying to minimize the size will require active air movement, that means noisy cooler.
For each cooler they usually note a nose in dB. Would say > 20dB is unacceptable. Also the size of cooler - the bigger the size the more air flow and less nose.

Definitely. A big fat heatsink and a set of quality 120mm PC fans is a good way to go.

5. They recommend 5W per gallon. to reduce the light by dimming it is a 5 minute job, but to add some power is a big trouble. So it is better to have a buffer -the bigger the better. At least to make picture - nobody likes blue picture, so this 20% of red+cyan+white should be enough to make a nice daylight picture.


Watts per gallon is largely irrelevant when it comes to LED... PAR and coverage are more important. With so many different types of LED on the market the differences in efficiency can be huge.

Definitely agree on oversizing the capabilities of the fixture. You can always dim it down, or swap drivers out as needed. :thumbsup:
 
Might get a little tight for wiring, I really could use a PCB that had 8 drivers on it that are functionally compatible with the LDD series, but the only source I know (O2surplus) has gone dark, so I'm trying to shoehorn everything in the fixture as is.

I actually designed a 6-channel PCB designed to fit in the RapidLED heatsink a while ago but didn't have the funds to build my DIY light so it got shelved. My thought was to have everything inside the enclosure so only 48V input was needed. PCB has a built-in DCDC converter to step down to 6-12V (so you can adjust the max fan speed), 6 channels using the LM3414 (each with pot to adjust current) and Wemos/PCA9685 for PWM dimming. Everything is surface mount so you can mount the board directly to the heatsink to cool the drivers.

yyhtDM5.png


If you're interested i could swap the Wemos out for 2 extra channels, although you'd need an extra connector then for PWM.
 
I actually designed a 6-channel PCB designed to fit in the RapidLED heatsink a while ago but didn't have the funds to build my DIY light so it got shelved. My thought was to have everything inside the enclosure so only 48V input was needed. PCB has a built-in DCDC converter to step down to 6-12V (so you can adjust the max fan speed), 6 channels using the LM3414 (each with pot to adjust current) and Wemos/PCA9685 for PWM dimming. Everything is surface mount so you can mount the board directly to the heatsink to cool the drivers.

yyhtDM5.png


If you're interested i could swap the Wemos out for 2 extra channels, although you'd need an extra connector then for PWM.

That is rad!

I wouldn't mind one of those as is! I can always piggyback another board or two off of the PWM pins of the Wemos, since I need several identical channels anyway. One of those boards as is, and another with a couple additional drivers instead would be extremely handy. I need 8 drivers total, but only six channels of control. One of those boards with a blueacro two-channel driver board piggybacking along work be perfect.

Do you have any finished boards yet, or just the files? I'm definitely interested either way.
 
Those boards do look great, is the fan DC controllable at all? I'd love to see a driver board that had something like a "if I am drawing so much current then switch on my fan output"... Would like to not have an extra board but still have the fan controllable in the event you have to use a two wire fan...
 
That is rad!

I wouldn't mind one of those as is! I can always piggyback another board or two off of the PWM pins of the Wemos, since I need several identical channels anyway. One of those boards as is, and another with a couple additional drivers instead would be extremely handy. I need 8 drivers total, but only six channels of control. One of those boards with a blueacro two-channel driver board piggybacking along work be perfect.

Do you have any finished boards yet, or just the files? I'm definitely interested either way.

Just the files at this point, never had them made since after pricing out everything it was too much. The parts cost ~$50 and cheapest cost for PCBs is $24 although not sure if that's with shipping. How are you soldering skills? Or would you prefer to buy a populated one?

I did some code based on Blynk although haven't had a chance to really test it yet. I think the board would also be compatible with the SSLAC16 software O2 mentioned before. Also considered replacing the Wemos with a Bluefish mini which would cost more but then i wouldn't have to worry about the software side of things.

Those boards do look great, is the fan DC controllable at all? I'd love to see a driver board that had something like a "if I am drawing so much current then switch on my fan output"... Would like to not have an extra board but still have the fan controllable in the event you have to use a two wire fan...

My plan was to have two connectors for DS18B20s to monitor heatsink/LED temperatures and use that to control when the fan turns ON/OFF. Didn't plan on PWM for the fan since the fan for those premium heatsinks isn't PWM enabled.

The board is 50mm wide and designed to sit at the edge of the heatsink so the barrel plug sticks out, but it could also sit in the middle if the LEDs are right up to the edge of the heatsink.
 
I can solder, and can do surface mount components, as long as they don't require reflow type work. Depending on difficulty/cost I would definitely be interested on one or two.

Is that 24$ for a single PCB, or a set?
 
I can solder, and can do surface mount components, as long as they don't require reflow type work. Depending on difficulty/cost I would definitely be interested on one or two.

Is that 24$ for a single PCB, or a set?

The LM3414 have thermal pads which should be re-flowed, might be possible with a hot air gun. See LM3414HVMRX/NOPBCT-ND on digi-key. Would you want the wemos or Bluefish socket?
 
The $24 was for 10 pcs, looks like the minimum order is actually 5 pcs for $19 though so not as bad.
 
Wemos socket for sure.

19$ for five ain't bad at all! Might be worth ordering ten and splitting them though. I could find a use for that many for sure.

I've got a few wemos on hand to fool with as it is.

What pcb house?

Heat gun huh? I've never actually tried it that way.
 
Wemos socket for sure.

19$ for five ain't bad at all! Might be worth ordering ten and splitting them though. I could find a use for that many for sure.

I've got a few wemos on hand to fool with as it is.

What pcb house?

Heat gun huh? I've never actually tried it that way.

Ok I'll go with the Wemos socket, worst case the SSLAC16 code should work if my Blynk code doesn't. Just finishing a couple things before I order them, was busy over the weekend so didn't get a chance.
 
Back
Top