The Planted Tank Forum banner

Designing and Building a LED Fixture

138K views 399 replies 65 participants last post by  AquaDwell 
#1 · (Edited)
This is getting to be an obsession with me! Hours spent on the computer, searching for parts and data, sketching out design layouts, doing calculations, etc. So, I decided to share what I'm doing, hoping that cooler, more experienced heads will speak up where appropriate.

I have a 45 gallon tank, with about a 30 inch by 15 inch footprint, about 20 inches tall, with a PC light fixture mounted about 24 inches above the substrate level. My goal is to replace the PC guts with a LED array that won't break my bank account wide open, and will give me around 50-100 micromols per sq m per sec of PAR at the substrate level.

One approach is to use relatively low output LED's and many of them to get that intensity. Another is to use the fewest number of LED's that will do the job, but using high output units. The first discovery I made is that lower power LED's don't cost 1/10th of what higher power ones cost, so every arrangement I looked at, using many LED's ended up with a higher material cost than using just 3 high output LED's. So, here is the layout I decided to pursue further:


You can get this LED, http://www.luxeonstar.com/endor-reb...ma-p-179.php?zenid=mq7qbu6j6941aoe1im80dot2b2 with a lens to concentrate the light, for about $24 each.

Next I figured out the approximate angle of the cone of light I would need to get good coverage of the substrate with these. It turns out to be about 45 degrees. (The LED emits light in about a 160 degree cone, but the intensity drops pretty fast as you move from the centerline - the lens will partially correct that.)

To verify that this would give me enough light I found this chart to use to convert the LED lumens output to a PAR intensity at the substrate:
From this it looks like a good number to use is 70 lux per micromol per sq m per sec, where one lux is one lumen per meter squared. Since I will have most of the substrate getting light from 2 LED's I used 25 PAR units to calculate from: 25 x 70 = 1750 lumens per sq. meter. And, the area each LED will primarily illuminate is about a 20 inch diameter circle, or .3 meters squared. 0.3 x 1750 = 525 lumens. The LED I chose gives about 435 lumens output at 700 milliamps of current, but can be driven to 1000 mA, so this comes close to what I want.
 
See less See more
2
#2 ·
I realized that some of the spillover light can be caught by reflection off the inside surface of the glass, and by reflecting from the light fixture reflective surfaces. So, I sketched this to see how that might work:


It looks promising, to me, at least. I then priced the heat sink required to keep the LED's cool. That would be a one foot long piece of 5 3/8 wide heat sink, cut along the centerline to make two 2 5/8 wide pieces, placed end to end. A single 120 mm cooling fan mounted above this should keep the heat sink cooled. That adds another $35 for the heat sink and about $20 for the fan.

To power the LED's, drive them, I found a 40 volt DC unit - it takes about 29 volts for the LED's in series and 700+ mAmps of current. This is the unit I found: http://www.cdiweb.com/PortalProductDetail.aspx?ProdId=23479 for about $20.

So, the total cost of parts would be about $150. So far, so good.
 
#4 ·
One of the attributes of MH lights is the shimmering effect the surface ripples give you, from the shadowing effect of the small light source. I figure I will get something similar to that effect this way. But, I'm still a bit nervous about whether or not this will be an adequate amount of light, as well as the less than perfect distribution of light.
 
#5 ·
I don't have personal experience, so take my comments for what they are worth...

I agree with Epic, 3 LEDs will lead to very focused areas of light. I understand the ripple thing, but keep in mind you are comparing 150W HQI bulbs to tiny 3W LED emitters.

When you balance between the output and number of LEDs, lots of low power LEDs are no good, but too few spotlight might not be good either. Rather than 3 triple LEDs, I'd do 9 (or 12) single emitters. When you go from 1 Watt to 3 Watt you can see the brightness triple, but going from 3 Watt to 3x3 Watt doesn't do that.

Check out Cree/Seoul Semiconductor LEDs. LED technology is advancing rapidly, and what was the best yesterday could be replaced by less expensive, more efficient things today.

Adding a lens to a LED to bundle the light sounds weird to me. Add a reflector, sure, but LEDs have a directional light output already, and adding a lens IMO just reduces light output further.

If you add distance between the LEDs and the water surface, keep in mind that the inverse square law darkens things rather quickly.

And finally, for $150 you can build yourself the awesomest, brightest T5 setup you can imagine. But of course that isn't as exciting.

(Just subscribing to your thread).
 
#6 ·
Hoppy,

this might be a subtle effect, but in the schematic in your second post you neglected to take refraction on the water surface into account.

Refraction happens at the interface of two mediums with different optical indices (sp?). This should slightly reduce the amount of "spill over"

If you are interested in correcting for refraction just search on wikipedia if you don't already know how to do that.
 
#7 ·
My 4 gallon build has cost me about $300. I am going with 12 3 watt Crees. The buckpucks that are running the LEDs will be able to be dimmed. I am going with 12 so they don't give off a spot light effect and will blend well. I think the rule to good blending is 2'' or less between leds, anything more and they spotlight. I know 12 leds is overkill but i will dim them down to a good level of light.
 
#8 ·
Merk, there is no way I would even waste time on this if I felt it would cost me $300 - my wife would kill me! So, while I agree with what you are saying, I'm still looking for a way to get down to around $150.

Regloh, I agree about the refraction, but most of the light will enter the water close enough to vertical for refraction to make little difference, especially given the overall inaccuracy involved with this.

Wasserpest, I agree about the T5's too, but if I go ahead with this, it will be in part to avoid having to buy new bulbs, and it is an opportunity to study up on a new subject for me. As usual, the more I learn the less I find I know! The problem with using more LED's with lower power is that the cost doesn't drop as fast as the power, so I end up again with a $300 fixture. Tonight I started studying the Cree website, and I'm still at the utterly confused stage. A couple of good resources I found were http://www.highend.com/support/training/lightingfaq.asp and http://www.theledlight.com/technical.html A couple more weeks and I may understand the subject a little better.
 
#10 ·
The problem with using more LED's with lower power is that the cost doesn't drop as fast as the power, so I end up again with a $300 fixture.
Nowadays LEDs are available for all sorts of prices.

http://www.dealextreme.com/details.dx/sku.1445

These guys also have the SSC P7 for $19. Might be good in single bulb config over a 3gal cube. :)

Not saying this would be the best in terms of light color and such, just keep looking around and you find things cheaper and cheaper. Same applies to the other parts of the system, like fans, heatsink and such.

Keep in mind that LEDs don't have a very high efficiency - most of the wattage is converted into heat. As you go up in wattage (like 12W for something like the P7) you will have to deal with lotsa heat in a very small spot.

edit - ninja'd big time. Need to type faster. :icon_roll
 
#9 ·
After studying the CREE LED data, and the reference materials I found, it looks like an array of 12 of the 100 lumen CREE LEDs will do what I want, without lenses, with the fixture lowered back down to the top of the tank. Here is the layout:



I think it takes about 4 overlapping LED light cones to get to the 50 micromol PAR that I want, but that is a very crude guesstimate. Twelve LEDs may be overkill, but at about $3.50 each, the price is right. This whole set up should get close to my $150 goal too.
 
#13 ·
I suppose it would be reasonably easy to determine what percentage of the output is in the middle 20 degrees of the cone of light, but I didn't even try that. So, I'm not at all sure if this will give the intensity I want. The rest of the cone also emits light, but at a lesser intensity, and that adds to the intensity where it overlaps too. (Just random thoughts as I have been thinking about this.) If nothing else this is an interesting subject to study.
 
#14 ·
Here is what this last setup would cost me: approximately
12 LED's @$3.50 each----------$42 + $10 shipping (guess)
Heat Sink-------------------------$40 + $10 shipping (guess)
48 VDC Power Supply------------$30 + $10 shipping (guess)
Cooling Fan and 12VDC Power---$20 + $10 shipping (guess)
Miscellaneous----------------------$20 (guess)
---------------------Total---------$192

My current light, running 10 hours a day, is a single 55 watt PC
This LED fixture would use about 33 watts, saving me about 220 watt hours per day, or about 80.3 KW hours per year, at $0.17 per KW hour, saving me $14 per year. My bulb needs replacing about once a year at about $30 each time, with the shipping, for a total annual savings of $44 per year.

My enjoyment from making this is worth??? Priceless? Now, will that sell the wife?
 
#15 ·
My enjoyment from making this is worth??? Priceless? Now, will that sell the wife?
That may depend on whether or not that savings of $44 dollars a year is going into her pocket or another DIY project.:icon_wink :hihi:

Let us know if it does convince her so I can start breaking it down like that.:)
 
#20 ·
Sacramento uses a multiple tier electricity rate, so, since we are always in the second tier, I used that rate - that number isn't fudged at all. And, I actually have replaced my bulbs in the PC fixture at about one year, because I was disappointed with the bulbs I was using, so I'm not sure that my once a year number was fudged that much. Some folks do it more often. Then there is the intangible "cost benefit" from using less electricity and helping to stop global warming. Plus, as I said, this is a hobby, something to do for fun, and much of the fun I get from the hobby is making new things for my setup, trying them out, and trying to quantify the performance of them. I don't do this as a business.
 
#26 ·
I have the price of this project down to about $150 now, so I am starting to order parts. I plan to use 24 Cree LEDs, ordered here: http://www.dealextreme.com/details.dx/sku.2395 in an array of 3 rows of 8 each, on a grid of about 24 inches by about 6 to 9 inches. Here is how I arrived at this:

I spent a lot of time on http://www.nano-reef.com/forums/index.php?showtopic=188085&st=20 and a couple of others. This particular thread gave me some good data to use to calculate what I would need for my tank and to get the PAR I want. So, I plotted the data on log log paper so I could extrapolate it:


This data is from 11 Cree LED's in an array that is 42.5 square inches, and it would give 240 PAR micromols at the 17 inch distance I will use. I will be illuminating about 450 square inches, so, assuming that the illuminated area that the data is from is the same size as the array, to be conservative, I will need 450/42.5 times that many LEDs, or 116, to get the same PAR. But, I only want about 50 PAR, so I need 50/240 times 116, or 24 LEDs similar to those used in the reference. I chose somewhat less efficient Q2 Cree LEDs for cost savings, and because of the conservative calculation I did above.

More to follow.
 
#27 · (Edited)
The layout of the 24 LEDs will be:


This should give me more than 50 PAR in the center of the tank, and pretty close to it around the edges. The yellow circles are the coverage from the center 10 degree half angle of the light - no lenses used. But, the next 10 degrees should theoretically be even higher intensity, due to the focusing effect of the air-water interface.

To power these LEDs I needed to decide how to connect them. If they are all in series, the total voltage required would be about 96 volts, too much for cheap DC converters. So, I decided to split them into two strings of 12 in series, with the strings in parallel. That requires a 48 volt power source with up to 1.4 amps of current. I lucked out because ebay now has just such a DC converter on sale, http://tinyurl.com/cpk3a8 I offered $12 for one, and the offer was accepted. With tax and shipping, that comes to only $23.82, much less than if I were to use "buckpucks", where I would still need a DC supply. But, this also means I need to use current limiting resistors. Here is how I decided what to use:


The two sets of resistors plus the toggle switch lets me toggle between about 700 and 350 mA of current. The numerous resistors are used first to handle the power that has to be dissipated, and because of ebay again! I found these: http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&ssPageName=ADME:B:EOIBSAA:US:11&Item=180334106900 which are 3 watt 1% 15 ohm resistors, perfect for this use. I plan to put the resistors and switch in a small metal box.
 
#28 ·
Next is the heat sink. Tomorrow I will visit a salvage metals store that is aimed at DIY folks like me. They have a big stock of aluminum extrusions at very low prices. I hope to find 2 or 3 inch channel extrusions, so I can put 8 LEDs on each of three of the channels, then join them, side by side, with aluminum washers separating them for air flow all around each channel. These will give me at least 24 square inches of surface per LED, with enough thickness to also store and easily conduct the heat to the air. At worst, I can buy the channels on ebay, again, for about $30, much less than the cost of a finned heatsink extrusion.


I have spent about $28 so far, not counting the heatsink or LEDs.
 
#29 ·
This morning I visited Blue Collar Supply, with its vast collection of aluminum extrusions. I found they have 8 foot aluminum 3" x 1" x 1/8" thick channel extrusions, for $24.85 each, or $26.78 with tax included. So, now I have spent $55 total. The fun begins!
 
#30 ·
One problem my current limiting scheme was going to give me was if one LED failed open circuit, the increased current through the remaining circuit would possibly desroy those LEDs. So, I redesigned it a bit:



This will let me toggle between 350 mA and 750 mA, and if one LED fails open circuit, the current through the other leg will be limited to less than 1000mA, which shouldn't burn them out.

Today I got the power supply, and it looks very good. And, I got the "heat sink" assembled, but my wife has my camera so I can't get a pic of it yet. I still need to flatten the mounting surfaces a bit and polish it some.

While working on it I decided that "heat sink" isn't the right name for this. It is a heat exchanger, not a heat sink. A heat sink stores heat, but this is supposed to transfer the heat to the air flow going by it. Thinking of it that way makes the design make a lot more sense to me. Spent another $2.24 on nuts and bolts today. Cost is now $30.
 
#31 ·
Wow! I finally found your thread. I dealextreme link did not work. What Cree LEDs are you using? color? wattage?

BTW...I am getting my LEDs over this weekend. So hopefully I will have my fixtures ready for use by Monday and can help you with some practical data.

I could not find in your designs, how high you are planning to mount the LEDs, above the tank. Since you have a deep tank(20"), I think you should calculate the lux on the substrate surface based on the distance. I reflector might also be a good idea for restricting light to be wasted outside the tank, unless you have a hood to mount it in.

Anyways, looks good so far. I am subscribed.
 
#32 ·
http://www.dealextreme.com/details.dx/sku.2395 is the LED I plan to use, a Q2 bin, less efficient version of the Cree 3 watt LED. I plan to mount this so it is 17" above the substrate, and I extrapolated the data from the nanoreef thread where the poster actually measured PAR intensity at two depths in the water - heights above the sensor. I'm not yet fully committed to this, but having spent as much as I have I am almost certain to go all the way with it. Any PAR data I can get will be extremely useful.

I won't need any secondary lenses since the distribution is pretty good as it is, at that depth. And, yes, it will be in my existing fixture, which will reflect side light back towards the tank.

This has become an all-consuming obsession for me!
 
#33 ·
I just ordered a couple of fans, http://www.directron.com/sil.html with shipping, they are $25 total, bringing my cost up to $55. If you want to see some testing of fans for noise, check out http://www.directron.com/noise.html. I was shopping for a cheap 12 volt adapter to run the fans when I thought about all of the battery chargers I have collected from cell phones, etc. Great idea! I have a couple that are just the right size and rating. Zero cost for that one.

Today I also got the surface of my "heat sink" sanded and polished as much as I think is needed. I started with 80 grit dry wall sanding paper, which is silicon carbide paper, and it worked very well on aluminum. Then I used 150 and 320 grit to finish sand it. I should have my camera back tonight, so I will plan to take a picture and post it.
 
#34 ·
I just ordered a couple of fans, http://www.directron.com/sil.html with shipping, they are $25 total, bringing my cost up to $55. If you want to see some testing of fans for noise, check out http://www.directron.com/noise.html. I was shopping for a cheap 12 volt adapter to run the fans when I thought about all of the battery chargers I have collected from cell phones, etc. Great idea! I have a couple that are just the right size and rating. Zero cost for that one.
I got the same fans too but 60mm. 5.99/piece from Fry's and no shipping. For 4 of them for around $26. 2 for each fixture

My drivers can handle the load of the fans too along with the LEDs so I will add them in parallel in the same circuit, so that I don't have to use separate timer for them. And I will have a switch in that leg so that I can switch them off if required without switching off the whole circuit.
 
#37 ·
nice! If the dimensions match your tanks length & width you have more than sufficient surface area. While I do mine it is good to see a similar idea taking shape in parallel.

How do you plan to mount it against the reflector?
 
#38 ·
My light fixture is a DIY wooden one, with an 2 x 55 watt AH Supply bright kit in it. So, I will remove all of that, and use stand-offs to screw this to the top of it. The fans will be above that blowing in from the top, each about a quarter of the way from the end of the heatsink.

I'm currently planning to put the current limiting resistors in a small box mounted on the back of the fixture, with the 48 and 12 volt power adapters in the stand mounted on the wall of the stand, and the DC wires twisted together and running up to the light.
 
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