Tropical Fish Keeping banner

LED Lighting - DIY Style

11K views 46 replies 10 participants last post by  Geomancer 
#1 ·
So I'm an Electrical Engineer by day, novice aquariest by night.

I've been thinking for weeks of something practical I could make at home that wouldn't cost exponentially more than just buying it pre-made off the shelf.

Then I had an epiphany! LED lights to grow plants! Genius!

So I went to my favorite parts supplier and lo and behold they did list kelvin ratings, awesome. So I searched for everything between 6000K and 7000K. Got lots of results. Perfect!

The question then becomes of intensity, and how on earth do you compare LEDs to incandescent (or even fluorescent) lights.

The brightest option was a 850 lumen LED at 6300K, 11.8W. Not bad at all. Now, if I look at a table of lumen's versus watts for incandescent lights that makes it equal to a 60W bulb (or 14W fluorescent).

Does that make sense in my logic? That's one heck of a lot of light coming from a single 12mm (half inch) diameter LED!

The issue, of course, is cost. These suckers cost $26 each! However, I was only looking at the top of the barrel. Lower intensity LEDs are of course less expensive and I'm sure I could find a more cost effective $/lumen. Still, a couple of those and it isn't going to be more expensive than buying a pre-made one (A 5 LED 19" @ 6500K marketed for growing plants is $136 and is only 12W).
 
See less See more
#5 ·
Very little red in the cool white.


I would argue that lumen would not be the best way to compare light output unless the spectral output is the same. Irrandiance (watts persquare meter) would be the best way to compare, but not sure it if that info is available.


The 11.8 watt number for the LED is interesting. Is that power consumption? Any links to the sites where you are getting this info?
 
#6 ·
Yes, that's the maximum power rating.

I do my shopping at Digikey, although Newark is another big one. I find Digikey easier to find things though, both will sell electronics is small quantities.

I know lumens on their own isn't much use, but coupled with the kelvin rating should give some basis for comparison. I would hope that a 5500K LED for example would emit nearly the same 'colors' as a 5500K fluorescent? Lumens, power, and wavelength (or temperature) are all that's given typically in LED specs (well, that maters in terms of lighting). The lower power ones give their intensity in candle power (mcd specifically).

An alternative is to get wavelength specific LEDs and build an array of them, but I have a feeling plants aren't that specific, more of a range of wavelengths than just two very specific ones. That's why I went with ones rated in kelvin rather than specific wavelength.
 
#8 ·
I hope it didn't seem like I was questioning you about the numbers you had listed, I just wanted to be able to go to the sites and look around rather than ask a bunch of questions.

I would hope that a 5500K LED for example would emit nearly the same 'colors' as a 5500K fluorescent
This may be the case, but since the spectral output of fluorescent lights can be very limited, only emitting certain wavelengths and Kelvin temp technically only refers to incandescent (e.g. full spectrum) sources, the K rating of fluorescent lights can be somewhat inaccurate. As a result similar K ratings don't necessarily mean similar spectral output and trying to compare lumens of sources of dissimilar spectral output could be misleading regarding light energy output - it would be ballpark at best.

It would be easy if the radiant power numbers (in watts, which is why I asked about the 11.8 number) of the various light sources were availble - I mentioned irradiance earlier to try to avoid confusion between electrical power usage watts and light energy output watts, but since we are talking light source, radiant power is the better measure.

Regarding wavelength specific LEDs, the association between chlorophyll (a) and photosynthesis is very specific regarding wavelenth, but that is not to say that flooding plants with light of those wavelengths produces better growth. Full spectrum light (roughly equal in blue, green, and red) is generally recommended.
 
#9 ·
I hope it didn't seem like I was questioning you about the numbers you had listed, I just wanted to be able to go to the sites and look around rather than ask a bunch of questions.
Not at all =) I'm confident in the electrical aspect, but the light aspect, well, I have no experience with that.

This may be the case, but since the spectral output of fluorescent lights can be very limited, only emitting certain wavelengths and Kelvin temp technically only refers to incandescent (e.g. full spectrum) sources, the K rating of fluorescent lights can be somewhat inaccurate. As a result similar K ratings don't necessarily mean similar spectral output and trying to compare lumens of sources of dissimilar spectral output could be misleading regarding light energy output - it would be ballpark at best.

It would be easy if the radiant power numbers (in watts, which is why I asked about the 11.8 number) of the various light sources were availble - I mentioned irradiance earlier to try to avoid confusion between electrical power usage watts and light energy output watts, but since we are talking light source, radiant power is the better measure.

Regarding wavelength specific LEDs, the association between chlorophyll (a) and photosynthesis is very specific regarding wavelenth, but that is not to say that flooding plants with light of those wavelengths produces better growth. Full spectrum light (roughly equal in blue, green, and red) is generally recommended.
Yeah, that information is not provided, just the electrical power, a lot of which I'm sure is heat.

The datasheet for the small 6300K bright one that's fairly expensive at $26 each is:
http://media.digikey.com/pdf/Data Sheets/Seoul Semiconductor/W724C0.pdf
(also has a spectral graph, looks to be worse than the 5600K one to be honest.

The datasheet for the less expensive 5600K one is:
http://www.bridgelux.com/assets/files/DS23 Bridgelux ES Star Array Data Sheet DS23 120311.pdf

I suppose the only way to know for sure is to try it and see what happens :lol: The 5600K ones are only $5.26 each.
 
#12 ·
Yeah, that information is not provided, just the electrical power, a lot of which I'm sure is heat.
This is the problem. Without knowing the efficiencies of the different types of lights it would be difficult to compare them since the light output of each isn't readily available. I assume the LEDs are more efficient in converting electrical power to light power, so my guess is that an 11.8 watt LED would put out a lot more light than a 14 watt fluorescent – much more than just the 2.2 watt difference.

if you can, test different depths with a PAR meter
Yes, directly measuring the amount of light energy would be the way to go and a PAR meter would give the amount of light energy, at least the amount that reaches a surface. But you would first need to measure the output of a fluorescent fixture that is known to produce good results in order to match that with the LEDs.
 
#13 ·
Yes, directly measuring the amount of light energy would be the way to go and a PAR meter would give the amount of light energy, at least the amount that reaches a surface. But you would first need to measure the output of a fluorescent fixture that is known to produce good results in order to match that with the LEDs.
to compare, you can find par readings online, esp on lighting for saltwater tanks. it prob. wouldnt be practical to build lights then pay to buy a PAR meter but if you could borrow someones that would be the best way to test what youve built.
 
#14 ·
Yeah, not sure where I would find a PAR meeter to borrow ^_^

And I'm most certainly not going to pay $300 for one =o

If I do it I will experiment on my 20 gallon, probably get three of the LEDs. If I can do it for $50 that isn't bad compared to buying a different fixture and tubes. I already have a single 14W T8 8000K fixture, but that's certainly 'low light' and I'd rather have 'moderate'. And if the plants grow well under it... so much the better =)

If not ... well, I'll have a really bright light strip I can use to light a room hehe.
 
#16 ·
yep a DIY LED rig would certainly work. At least you are down the right path using high intensity LEDs. They are still too rich for my blood lol, even DIY. Prefer to stick with cheaper DIY fluorescent fixtures. Those LEDs put out some serious heat though, they NEED to be heat sinked or they will burn themselves out.

I've used my camera in manual mode as a cheap way to gauge the intensity of light, but you can really only compare two fixtures against each other.
 
#17 ·
Yep, I've already run through all the calculations for a heat sink and have parts selected :lol:

Trying to decide how I want to handle the large tolerances in the forward voltage, power supply, and current limiting resistor. I'd like to order everything at once to save shipping costs, but may be easier to buy the resistors after the others are known.

I'd kill for a local electronics store, or even a radio shack of a couple decades ago.
 
#18 ·
Bleh...

The more I think about this, the more it turns into a thermal nightmare :X The LED was simple, I knew a heatsink was needed for it and found a simple solution.

The rest of it though...ugh. Sometimes writing it out helps the brainstorming process.

There are pretty much two ways to drive an LED. Either you put a resistor in series to limit the current, or you use a constant current source.

A resistor is the time tested ultra simplistic and cheap method, after all the only part is a single resistor which are really easy to find and cost pennies. The issue with them is the current will vary widely depending on a number of variables. Current is what determines how 'bright' the LED is. Too small and it will be dim, too high and it will burn the device out.

A constant current source, as it implies, ensures that a constant current goes through the LED. This means its 'brightness' is a constant, and known. This requires a few more parts and there are a couple methods of achieving this cheaply.

Both methods have a power dissipation problem though. In both cases you have a resistor that must take the full current load of the LED. For the particular one I'm looking at it is 350mA which while sounds small is actually quite high. In the resistor only option it must dissipate ~2W of power (your typical surface mount resistors that almost everything uses this day can only do 1/8 W, the older through hole ones are 1/4 W). 4W+ resistors do exist (you want to be about 2 times higher than your power for reliability reasons), but they're large-ish and get very hot and of course cost more (but still not unreasonable).

For the constant current option, it is a transistor that takes the brunt of that power instead of the resistor. Heatsinks for transistors exist, but still you're looking at the thing getting ~50 deg C (122 F) higher than ambient, so you're looking at around 200 deg F. Ouch!

I'm not concerned about the parts themselves, they can take the heat, more having things that hot exposed where fingers/arms may brush against. Building an enclosure, and while your at it putting a small fan in, would make everything better but that adds yet more cost.

Debating if it is worth just buying a pre-made constant current LED driver. They are more sophisticated in how they work to get higher efficiency (thus lower power dissipation), but cost about $10 each (would need three for my purpose). They too could be custom built, but would not see any cost savings. They would make the whole project cost upwards of $100 at which point they are not much savings over commercial products.
 
#20 ·
Well, I decided to take the plunge and go for it. Why not? In any case, it should be a good learning experience.

Parts were a little more than expected, but I ordered some extras on a lot of it 'just in case'.

I went with an LED that puts out 685 lumens, which is just shy of what a GE 18" T8 Daylight flourescent bulb. The color spectrum will be a little different (the GE is 6500K or 6700K I forget) while the LED is a little warmer at 5665K.

I plan on using three of them, but under driving them. My goal is to be roughly equivalent to a dual T8 fixture. They have a viewing angle of 120 degrees, so if my trigonometry is still good (questionable...) a single one of these should light the entire bottom surface of the tank. But, of course, that would be like a 'spotlight' which can be a cool effect on its own, but I want it to be a little more evenly spread.

So, what is the cost? $70.26, including tax and shipping. $20 more than I wanted, but I will have enough left over to almost have enough to make 5 LEDs. The most expensive part was actually the sheet of plastic I'm going to use to assemble it to. If you already have a wood canopy (or old fixture to cannibalize) that's $12.50 saved right there. Without that, and the spare parts, the cost would have been around the estimated $50.

And if it dosen't work well for growing plants ... Well, I can have a very bright light strip installed in the basement laundry room ;)
 
#21 · (Edited)
From my limited experience, it seemed that the 1-5W LEDs were the most efficient compromise between intensity (lumens/watt) and price. Prehaps the ones from CREE?

When I was thinking about doing it before, I was going to use a CPPU power supply as a constant current source... Not sure if it would work, but they're cheap at the thrift store.

As for the spectrums, I was considering an array... Pretty similiar to one of these-


Those fixtures have only 3 spectrums, run about 90 watts, and put out light comparable to a 600 watt metal halide. Of course, purple fish would look weird, so I was going to add some cool white to balance. (Those UFOs appear quite dim, since our eyes aren't sensitive to photosynthetic light)
 
#22 ·
The ones I got are a spectrum (posted the graph previously) so it should be similar to regular bulbs. The 'single' LED is really an array packaged into a single bit of silicone. They are only $5 each.

I'm not sure how a computer power supply could be used as a constant current source by itself, do you have any information on that? I could see it easily being used as a DC power supply though, I'm personally using a laptop power supply. It's universal so has a switch on it to select between several different output voltages. Was $10, so not bad, I wanted the flexibility. In the future I could go with a fixed voltage one that would be a couple dollars less (or free if you have something lying around).
 
#23 ·
Well, I have a prototype set up on a breadboard. Holy crap are these suckers BRIGHT... even being under driven they are bright. Just a second of looking at it is enough to see spots for ~10 minutes after. Makes me somewhat concerned about the fish going blind actually. The lumens are about the same as an 18" T8 bulb, but all that light is coming from an area the size of a dime instead of an 18" long tube so it looks brighter when looking directly at it.
 
#25 ·
Well, okay, if you insist ;)

Here is a before shot:


And after:



The original is an 18" T8 8000K fluorescent tube. The new and improved is a single 5665K LED. You can clearly see the change in color spectrum between the two. Brightness wise, they look fairly similar although there are more shadows with the LED since it is a single point source. My plan was originally to put three of these up, but maybe only 2 will be sufficient to get my 'moderate' light goal. I don't want to deal with CO2 so I can't go too bright.

Here is a picture of the LED itself, the yellow center is about the size of a dime, that is what contains the LED array.


As you can see, I've had a bit of trouble with screw heads snapping off ;) I haven't been able to locate any #4 self threading screws, so I attempted to use regular #4 sheet metal screws with ... less than ideal results. Oh well, it's a prototype and you are only suppose to use two screws anyways ;)

This is a picture of the LED driver. As you can see there isn't much to it, just a couple transistors (with one having a heatsink), a couple resistors, and a zener diode.


The final product will be mounted to a piece of 1/4" thick plastic and the circuit will be soldered to a prototype PCB. I'm not going to special order a custom PCB as that has a minimum cost over $100 which just isn't worth it for such a simple circuit.
 
#27 ·
thats fantastic, though its all greek to me :( I guess I need to do my research! that will give me something to do for the next few days :)

on the other hand, I can wait till im rich and famous and have you kidnapped by my henchmen and forced into LED building labor. eeeeexcellent. nah, ill learn for myself :)
 
#28 ·
Well, I've finally put the time in to assemble everything together.

I'm really glad I didn't go with three of these suckers, and that would be insanely bright. Even as is I'm scared it will be too bright and I'll have an algae factory in my near future. It's an experiment, so that's okay.

I'm happy with how it turned out, but the Serpae Tetra most certainly look better under the 8000K light (more red in color) while under the 5665K they are more pale looking. Which is kind of odd because 8000K has more blue in it, not red...

I'm curious how the plants will respond.



 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top