Designing and Making a LED Light Fixture for a Planted Tank

[VaughnH]

The following is cross posted from The Planted Tank Forum:

I now have my LED light fixture operating, and working very well.

This is a 24 LED fixture, using Cree XR-E LEDs, from DealExtreme: $4.13 Cree XR-E Q2 Emitter with Star operated at 400 mA current, at about 20 inches from the substrate. Based on my adventures learning about LEDs, learning to design a fixture with them, the following is a summary of how to design and build your own for whatever light intensity you want.

 

[VaughnH]

Designing a LED light fixture
to produce 100 micromols per square meter per second PAR.

Using my data plus the data given by Cree for their LEDs, I made these charts:

Notice that as you get closer than about 20 inches from the LEDs, the inverse square relationship between intensity and distance switches to much closer to a linear relationship (intensity = 1/ distance instead of intensity = 1/ distance squared). The pink line is for the first setup of the LEDs when they were running at some much higher current than 400 mA, and quickly burned out the power supply. The black line is for the current 400 mA LED setup, plus some data from my AHSupply fixture.

To use this for designing a fixture to give 100 micromols PAR, first, decide how far from the substrate you want the LEDs to be? Let’s assume you want that to be 22 inches.

How much LED current do you want to use? At 700 mA or less, the LEDs should last for at least a few years. Higher current will reduce the life of the LEDs. Let’s assume you want to use 700 mA current.

My LED array gives about 55 micromols of PAR at about 22 inches, with 700 mA current. (Determined by studying those charts) You will need 100/55 or 1.8 times more light. To get that you need to space the LEDs closer together to get more of them illuminating each square inch of the substrate.

My LEDs are spaced at 3 inches on centers, or .11 LEDs per square inch of the fixture. Increasing that to 1.8 x .11 or .20 LEDs per square inch, will give the needed PAR. With my tank, that would require 4 rows of 11 LEDs instead of 3 rows of 8 LEDs. And the spacing would be 2.25 inches.

A good heat sink, fully capable of handling this array of LEDs can be made from 2 inch width aluminum channel, 6061 T6 Aluminum Channel 2" x .170" x 1-1/4" x 60"-Long - eBay (item 200355522940 end time Jul-22-09 11:31:12 PDT), or equivalent purchased locally. Four 24 inch lengths would be needed, which from the ebay source, would cost about $50. These could be bolted together, side by side, separated with 1/8” thick spacers, to make about a 8 3/4 inch wide heat sink.

For different foot print tanks you can adjust the number of LEDs to cover the foot print. For a 55 gallon tank, for example, 48” x 12.5”, the heat sink could be the same 8 3/4 inches wide, but 42 inches long, requiring 4 rows of 18 LEDs, 72 total.

You can also use brighter LEDs. For example you could use the Q5 version of that Cree LED and get 22% more light per LED, or use the best Luxeon Star and get about double the light per LED. Using these brighter LEDs lets you use fewer of them, but I suggest not getting them farther apart than 3 inches, to avoid any spotlighting effects.

 

[VaughnH]

Powering LED Light Fixtures

LEDs require a DC power supply, and an electronic device to provide a constant current to the LEDs from that power supply. Automobiles use 12 volt DC power for everything, with a socket or sockets available for plugging in 12 VDC appliances, from radios and cell phones to camping equipment like mini-refrigerators. To make those appliances usable on standard 110 VAC circuits there are many AC to DC adapters available, including this one, Amazon.com: Wagan 5 Amp AC to 12V DC Power Adapter: Automotive , which costs only about $25, and which will provide 5 amps (5000mA) of 12DC current. This is a good source of power for multi LED lights.

Each Cree LED drops 3.5 volts across it when operating at 700 mA current. A 12 volt source will, therefore, power only 3 of those LEDs in series, but strings of 3 LEDs in series can be wired in parallel up to a total of 5000 mA/700 mA = 7 strings in parallel. That means one of those 12 VDC power supplies can power 3 x 7 = 21 LEDs.

The 100 micromol PAR fixture we are designing requires 44 LEDs, but reducing that to 42 will have negligible effect, so we can use two circuits, each with 21 LEDs running at 700 mA for the fixture.

LED drivers are the electronic devices that connect the DC power to the LEDs to run them at a chosen constant current. There are many commercial LED drivers available, but, for 700 mA current, at a cost of roughly $5 per LED, more than doubling the cost of the LEDs for a light fixture. This suggests that a DIY driver is a better option.

A simple, inexpensive and effective LED driver can be made from an IC adjustable voltage regulator, and the most widely available cheap voltage regulator is the LM1084IS-ADJ, sold by Newark Electronics, NATIONAL SEMICONDUCTOR|LM1084IS-ADJ/NOPB|Linear Voltage Regulator IC | Newark.com

This tiny integrated circuit device can be mounted on the heat sink, using an electric isolator package, Parts-Express.com: TO-220 Silicone Insulator 10 pcs. | transistors transistor rubber ic heat bag to allow heat to be conducted to the heat sink, but not electric current. Adding one resistor, and possibly a couple of capacitors will enable this tiny device to act as a constant current controller for about 6 of the LEDs, so 7 of them will be required, but each costs about $3.00, or about $0.50 per LED, one tenth of the cost of a commercial driver. The voltage limiter is connected as shown:

The R1 resistor sets the constant current provided by the voltage regulator, and that current is equal to 1.25 divided by the value of the resistor in ohms. To get 700 mA for each of the two parallel strings, or 1.4 Amps total current would require .89 ohms, which is not a resistor value that is available, but 2-1.8 ohm resistors, VISHAY BC COMPONENTS|5083NW1R800J|Metal Film Resistor | Newark.com in parallel give .9 ohm, which results in 1.42 Amps or 710 mA current per LED. This simple circuit will work adequately as long as the DC power supply is kept within about 6 inches from the voltage regulator. If the DC power supply is located farther away, a couple of little capacitors are needed to avoid excessive fluctuations in the LED current.

 

[VaughnH]

Mounting and cooling LEDs

LEDs require a heat sink to keep them from overheating and destroying themselves. For a big light fixture, a heat sink made from aluminum channel extrusions, bolted side by side to give the width needed, works very well, and is much less expensive than finned heat sinks made for that purpose. It is helpful to use about 1/8 inch thick aluminum spacers between the channels to provide space for cooling air to flow around each channel.

After bolting the channels together, use silicon carbide sandpaper wrapped around a wood block to sand the faces flat, and polish them with finer grades of silicon carbide paper. The objective is to get as perfect a contact between the LED “star” mountings and the heat sink as is possible.

The “star” mountings have notches around them to allow holding them in place with #4 size screws, and flat head screws, normally countersunk, are the best type of screw for this purpose, because they cannot accidentally contact the very close by solder pads on the “star” mounts. Drilling and tapping the holes to fit #4 screws is very easy, using a cordless screw driver or drill.
EDIT: Apparently the star circuit board the Cree LEDs are mounted on do not insulate the center aluminum layer very well at the edge where the mounting notches are located. Using flat head screws can cause current leakage to the heat sink, which should be grounded, and shut down the light. It may be much better to use an adhesive thermal conductive material to mount the LED stars on the heat sink, and omit the screws.

First mark lines along the heat sink where the LEDs will be located. Then mark a spot 3/8” (9.5 mm) on each side of that line on the centerline for each LED. Use a center punch to make a small indentation on each mark. Drill a 3/32 inch diameter hole through the heat sink at each location. (Extreme accuracy is not needed.) Use the cordless screw driver, or cordless drill with speed control, to slowly tap the hole with a #4-40 tap. It works best to turn the tap a few times, back it off to remove the aluminum particles from the tap, then finish tapping the hole. Working with aluminum is much easier than working with steel, so it is very unlikely that you will break the tap.

Use silicon carbide paper, a fine grade, to flatten the areas where the drilling and tapping raised the metal, and polish again with finer grade paper. Clean the heat sink surface with alcohol until no more powdered aluminum rubs off.

To get better heat transfer between the LED “star” mounts and the heat sink, use a thermal compound on both the heat sink surface and the back surface of the LED. There are now many such compounds available, but one of the best for the money is Antec Formula 5, which comes in a big enough tube to mount about 50 or so LEDs. Fry’s Electronics stocks it. Use only a tiny amount on each surface, a dot of the compound, spread a little with the nozzle of the tube. Rotate the LED slightly as you lay it in place to further spread the compound, and attach the LED with 1/2” long #4-40 flat head screws. (The length of the screws isn’t important, but needs to be long enough to handle easily.)

The heat sink will slowly get hotter and hotter, with the light fixture on, unless a cooling fan or fans are used to help transfer the heat to the air. Mounting the fans above the heat sink, blowing down on the back of it, works fine. This gentle cooling breeze will also help keep the AC-DC converters and voltage regulators cool, if they are mounted on or above heat sink.

Use short lengths of 22 gage insulated wire to connect the LEDs in series, in groups of 3, for this particular design, and longer lengths to make the parallel connections behind the heat sink. Soldering is easiest using a solder with a small amount of silver in it, and resin core, Silver-Bearing Solder (1 Oz.) - RadioShack.com plus a small soldering iron, 15-Watt Soldering Iron with Grounded Tip - RadioShack.com

You now know just about all that I know about making a relatively cheap LED light fixture. Try it. It's fun!

 

[VaughnH]

This shows the underside of the heatsink, with the cooling fans, two of them, the 48 volt AC-DC converter, and the tiny little voltage regulator with current setting resistors. I used two current setting resistors so I can use a DPDT switch to select either 200 or 400 mAmps current. The lower current will be useful when I have to be away for a few days.

 

[Tom Barr]

Did you want to borrow the PAR meter to check the theory vs the actual?


Regards,
Tom Barr

 

[shoggoth43]

Some possible additions....

Make sure to use a Constant Current Source as shown. Just doing the math and using a voltage regulator circuit will not work long term. Most Voltage Regulator circuits are not precise enough and over time will "wander" enough to shorten the life of the LEDs, if not burn them out immediately. The circuit design listed will work just fine but is not the most efficient design. The closer you can get the input voltage to that required by the LED string the better for efficiency. With a small addition of a Potentiometer you may dim the LEDs further by adjusting the current they get. At 1.4A an LM317 VR chip is going to be quite warm, put a heatsink on it for maximum life ( assuming the one shown above is similar ).

LED drivers are available for reasonably cheap if you hunt around. EBAY has some that are perhaps 3-5$ for a buckpuck style design with and input voltage of up to 16V for 750mA. With some extra bits these can use Pulse Width Modulation to achieve dimming as well. These designs will typically be more stable across varying input voltages and more efficient than the Voltage Regulator chip used as a Constant Current Source.

15W is barely enough to solder these LEDs with so flux would be a good idea. A 25W soldering iron will be faster and more reliable. Especially if you attempt to solder to the LEDs when the star is mounted to the heatsink or if you go back and adjust/clean up your wiring. If mounted it will draw away a lot of heat requiring the iron to be used longer and thus upping the chance for damage and a dead LED.

LEDs are bright enough to get that "bottom of the swimming pool" caustic lighting effect you get with MH lighting. If you have too much of a spacing on your LEDs and they are too close to the surface of the water this effect rapidly turns into a bizarre strobing type effect which can be very distracting. Keep the spacing tight if possible and lift the LEDs as high off the surface to cut down on the angles the light can hit the water surface with.

-
S

 

[Philosophos]

I like the look of this, and the price. You've made a DIY costing $150 for something that normally seems to go at around $900.

I've got a question/request with this design. Do you happen to have a nice, elegant formula sitting around for calculating PAR from these bulbs, with relation to inverse square etc? I'm looking to crunch numbers and it'd be a lot faster for me to check over your formula than to construct it from the ground up.

If this system works out, maybe I could throw together a simple excel spreadsheet that'll do PAR and cost calculations for individual projects? I'd be making one for my self anyhow.

-Philosophos

 

[VaughnH]

Did you want to borrow the PAR meter to check the theory vs the actual?


Regards,
Tom Barr

Tom, Marica left it with me a couple of weeks ago, so I have already been using it. I'm planning to make a better "wand" to hold it, before I return it, and wait until you get moved so it isn't in your way. The data on the log-log plot I posted is from using the meter.

 

[VaughnH]

Some possible additions....

Make sure to use a Constant Current Source as shown. Just doing the math and using a voltage regulator circuit will not work long term. Most Voltage Regulator circuits are not precise enough and over time will "wander" enough to shorten the life of the LEDs, if not burn them out immediately. The circuit design listed will work just fine but is not the most efficient design. The closer you can get the input voltage to that required by the LED string the better for efficiency. With a small addition of a Potentiometer you may dim the LEDs further by adjusting the current they get. At 1.4A an LM317 VR chip is going to be quite warm, put a heatsink on it for maximum life ( assuming the one shown above is similar ).

I am certainly not an electronics expert and far from it, so I can't dispute this. But, the literature that describes the LM317 voltage regulator indicates that this should work well for a long time. I have the device mounted on the big heat sink (electrically isolated, of course) so it can't get very warm. I'm only using it for 830 mA current, too. Doubling that current could be a problem with heat, but I recommend using the low drop out voltage version, the LM1084 version, instead, and this will heat up a lot less, plus allowing more LED's to be driven by one voltage regulator circuit. I looked into using a pot to adjust the brightness, but I decided it was overkill, offering no advantage that I wanted, and low resistance pots are not widely availaible or cheap.

LED drivers are available for reasonably cheap if you hunt around. EBAY has some that are perhaps 3-5$ for a buckpuck style design with and input voltage of up to 16V for 750mA. With some extra bits these can use Pulse Width Modulation to achieve dimming as well. These designs will typically be more stable across varying input voltages and more efficient than the Voltage Regulator chip used as a Constant Current Source.

I was unable to find drivers that would cost less than about $5 per LED being driven, far too much to fit within the budget I was trying for.

15W is barely enough to solder these LEDs with so flux would be a good idea. A 25W soldering iron will be faster and more reliable. Especially if you attempt to solder to the LEDs when the star is mounted to the heatsink or if you go back and adjust/clean up your wiring. If mounted it will draw away a lot of heat requiring the iron to be used longer and thus upping the chance for damage and a dead LED.

This concerned me when I started soldering the LEDs, but I had no problems at all doing the soldering with that little iron. This was with them mounted on the heat sink. The heat sink prevents any heating damage to the LEDs. The single minor problem I ran into was when I unsoldered a few of the LEDs. It takes more heat to unsolder them than to solder them.

LEDs are bright enough to get that "bottom of the swimming pool" caustic lighting effect you get with MH lighting. If you have too much of a spacing on your LEDs and they are too close to the surface of the water this effect rapidly turns into a bizarre strobing type effect which can be very distracting. Keep the spacing tight if possible and lift the LEDs as high off the surface to cut down on the angles the light can hit the water surface with.

-
S

The strobing effect, of shimmer effect is distracting, but I found I quickly adapted to it so it doesn't bother me. It is interesting to see it. I used my digital camera and software to check for "spotlight" effects - bright spots of light right under individual LEDs. I couldn't find any. Apparently my 3 inch spacing is enough to blend in the outputs of the LEDs, at the roughly 20 inch distance I have the fixture mounted from the substrate. If that fixture were located at half of that distance I'm sure there would be some spot light effect, and fixing that would require closer spacing of the LEDs but running them at a lower intensity also, or the PAR would be excessive.

Conversely, if you tried this fixture mounted 36 inches above the substrate, it would provide far too little light, and most of the light would spill over onto the floor. For that type of setup I think using lenses on the LEDs would be the best idea, but I haven't tried that. All the lenses do is concentrate the light from individual LEDs into a smaller "cone" of light.

 

[VaughnH]

I like the look of this, and the price. You've made a DIY costing $150 for something that normally seems to go at around $900.

I've got a question/request with this design. Do you happen to have a nice, elegant formula sitting around for calculating PAR from these bulbs, with relation to inverse square etc? I'm looking to crunch numbers and it'd be a lot faster for me to check over your formula than to construct it from the ground up.

If this system works out, maybe I could throw together a simple excel spreadsheet that'll do PAR and cost calculations for individual projects? I'd be making one for my self anyhow.

-Philosophos

I don't think a usable formula can be devised for converting LED output to PAR. One LED will give a PAR that is inversely proportional to the distance from it, squared. That is certain. But, an array of LEDs covering the top of the tank, doesn't lose intensity per the inverse square rule. That is because the farther you get from the array, the more the other LEDs contribute to the intensity at any one spot. Until you get beyond about 20 inches, for my configuration, then the drop in PAR starts to become an inverse square relationship. Describing this with a formula would be difficult at best. Also, I'm not sure yet that my results can easily be related to other LEDs, Luxeon LEDs for example. The only data we get on LEDs is the total lumens output. And, the PAR for the array of LEDs depends on the density of LEDs that provide the light. I suspect that graphical "solutions" to determining what array will give the PAR you want is the best way to do this.

But, I do hope you can simplify this so a simple spreadsheet, or formula, can be used. I find it very confusing to try to get to the result I want using the methods I used. The cost for making one of these is also hard to predict, because so many little expenses keep adding up.

 

[Philosophos]

Your project is in its early testing stages, I'm sure the costs will get figured out better as a few of us try your method out. Your design is the first decent DIY LED that I've seen.

The formulas would probably be unwieldy, yes, but I'm willing to give it a try. With any luck I'll have a PAR meter of my own before long, and be able to contribute better. A lot of this looks like a matter of geometry as much as anything.

No doubt the bulbs make a difference; I think there needs to be some pressure to list PAR, LUX and spectrum plots on most brands of lighting. Still, maybe an acceptable margin of error can be built in, with some hints from K rating.

It looks like you've done a nice job with this lighting system. I wouldn't be surprised if your fingers start getting sore from replying to this thread.

-Philosophos

 

[SuperColey1]

Your design is the first decent DIY LED that I've seen.

I'll have to redesign my Luxeon one then (joke)

Vaughan I'm totally lost r.e. the actual measurements and science here but totally agree r.e. the overlapping of light being the key with LEDs. If they are linear the results (from plant growth) are pretty poor whereas the same amount spread like yours (and mine) are superior to comparitive light sources.

I was wondering how the PAR compare to the ADA MHs that Tom measured at AFA?

Also what wattage are your LEDs?

To add to the difficulty of PAR readings on LEDS there can be a pretty big difference from 1 LED to another in terms of output and spread because not all the dies are perfectly identical. They are also not quite as accurate to K as T5 lights with CRI of 8 or 9.

AC

 

[VaughnH]

The Cree LEDs use about 3.5 volts, and I am using 412mA current, so together the 24 LEDs consume about 34 watts. The actual power being used is higher, because of losses from the 48volt AC-DC adapter. My guess is that it is using a total of around 38-40 watts.

You can't use a single PAR reading to characterize the lighting, since the light intensity is a minimum at the substrate level (about 45 micromols) and goes up as you get to the top of the water column (about 100 micromols), plus it drops some near the ends of the tank and at the front and back of the tank. Since this tank already has a lot of plant mass in it, it wasn't possible to get good PAR readings all over the tank.

My PAR readings are comparable to what Tom measured in the Aquaforest tanks.

These "cool white" LEDs do not make reds pop out at all. They are much more like 10,000K lights than anything else that I have seen. I would prefer having a lot more red in the spectrum, but getting that uniformly over the tank isn't easy. One way would be to use the 3 emitter LEDs, which have three emitters mounted on one "star" PCB, with those being red, green and blue, but I'm not sure if that would give a good color balance in actual use, or not. They would cost a lot more, for sure.

 

[SuperColey1]

The Cree LEDs use about 3.5 volts, and I am using 412mA current, so together the 24 LEDs consume about 34 watts.

My PAR readings are comparable to what Tom measured in the Aquaforest tanks.

This is pretty interesting then. The AFA one that is pictured in Tom's post uses 3 x 150W MHs and therefore if we are talking about comparable PAR readings to 34 watts (assuming you would only need one over your tank) then we are seeing that 34W LED is comparable with 150W ADA MH.

Then allowing for Tom's readings where his own home lighting was 2-3x the reading of the ADA we can say that (using the higher of 3) that 34W LED is equal to 50W MH.

What I am getting at is that this goes some way to proving the reports I read of LEDs actually bleaching corals rather than not being powerful enough. Of course talking about the initial reefer's complaints that LEDs were inferior to MH due to their coral's dying observations

Also confirms my observations (mine has been running for 6 months now) that the 37W LEDS (15" above the water surface) are matching or bettering the 48W T5HOs (6" above the water surface) from growth observations

Mine is quite a poor comparison because I am actually using the timed lighting stagger to minimise the 'full' period to 4 hours and the others come on incrementally.

I would guess to summise my own setup that I only have 3 quarters possible W of LEDs to my old T5HO, have it raised much higher from the tank and am using less of the maximum through the photoperiod.

Maybe I should set them up for a month to run 2/3rd max for 3 hours either side of a 4 hours 'full' non burst and compare growth rates then

Keep us updated Vaughan. You have the benefit of the equipment etc that I have no access to

On the red part I have seen people use the standard cheapo 3/5mm red LED strips just to add a little red into the string. Innefective as useable light but enough to give a little colouration to red plants. I personally love the clear, crisp look of the cool whites. Kind of a clean looking appearance

AC

 

[shoggoth43]

For color balance you should be able to tweak it a touch. One of the big projects over at nano-reef was the LED group buy. Most of them are buying white and blue LEDs. They're then setting them up on different drivers and dimming each driver to get the color temp they want. I can't see any reason why instead of blue/white you couldn't do warm/cool white, or add some reds instead.

Incidentally, from the PAR measurements they've made the blues put out roughly the same PAR as the whites, but have far less visible light due to the spectrum they use. If you really wanted to mess with this, you could go all out w/RBG on separate circuits and tweak till you had a migraine. I would think it'd be easier to just do the cool white ( or whatever ) and then mix in a couple red/green/warmwhite to knock things back to the look you want. You would also be able to just put the reds over the plants you want to be really red.

Closer spacing of LEDs would make this less obvious, or perhaps putting the other colors "inside" the usual grid spacing so there's no real holes in your lighting, just extra "oomph" where you want to have some intense color. Also, if you are starting from scratch all this is much easier than trying to retrofit into something else. It's very tough to cram all this into a splash guard on an existing tank vs. a completely square lumiere or other box hood, especially on the custom all in one type tanks.

-
S

The Cree LEDs use about 3.5 volts, and I am using 412mA current, so together the 24 LEDs consume about 34 watts. The actual power being used is higher, because of losses from the 48volt AC-DC adapter. My guess is that it is using a total of around 38-40 watts.

You can't use a single PAR reading to characterize the lighting, since the light intensity is a minimum at the substrate level (about 45 micromols) and goes up as you get to the top of the water column (about 100 micromols), plus it drops some near the ends of the tank and at the front and back of the tank. Since this tank already has a lot of plant mass in it, it wasn't possible to get good PAR readings all over the tank.

My PAR readings are comparable to what Tom measured in the Aquaforest tanks.

These "cool white" LEDs do not make reds pop out at all. They are much more like 10,000K lights than anything else that I have seen. I would prefer having a lot more red in the spectrum, but getting that uniformly over the tank isn't easy. One way would be to use the 3 emitter LEDs, which have three emitters mounted on one "star" PCB, with those being red, green and blue, but I'm not sure if that would give a good color balance in actual use, or not. They would cost a lot more, for sure.

 

[VaughnH]

Your project is in its early testing stages,....

-Philosophos

Unfortunately, this proved to be true. Yesterday the lights came on when the timer told them to. A few minutes later as I was getting ready to add the day's doses of ferts, the lights dimmed. I quickly checked and only 6 LEDs remained on. Those were the 1st through 6th from the + connection, in one 12 LED series circuit, with the other 12 LED circuit all off.

I disconnected it and stewed about it the rest of the day. This morning I removed the light and checked each LED individually - all are working. So, the problem is in the power system. But, I'm at a loss as to how 6 of a string of 12 LEDs can light up, and the others stay dark, with the LEDs all working when tested individually.

Any ideas? To help this along, I have 24 LEDs in two circuits of 12 each, with the two circuits in parallel, powered by a 48 volt DC supply through a IC Voltage Regulator acting as a constant current controller, set to 412 mA of current. I have a switch that lets me drop that current to 206 mA. I tried it and nothing changed except the intensity of the light. One more fact - it is very hot today and was very hot yesterday, but the air conditioning keeps the inside temp at the aquarium around 80 F.

EDIT: I found 3 questionable solder joints in the series strings of LEDs, and resoldered them, but it still doesn't work. Now, I'm wondering if this is entirely a matter of poor soldering in the strings, with the joints gradually increasing in resistance until that became too much. This:

15W is barely enough to solder these LEDs with so flux would be a good idea. A 25W soldering iron will be faster and more reliable. Especially if you attempt to solder to the LEDs when the star is mounted to the heatsink or if you go back and adjust/clean up your wiring. If mounted it will draw away a lot of heat requiring the iron to be used longer and thus upping the chance for damage and a dead LED.

may be the problem with the soldering too. I had a lot more trouble with soldering today than I had originally.

Another possibility is that, since the "- "connection on the LED string is ground, and the heat sink is grounded, any leakage of current to ground before the end of the string would cause some of the LEDs to be off. If that is in one of the parallel strings, that particular string will get most of the current, leaving the other string without sufficient current to light the LEDs. But I can't pick up a leak to ground yet with an ohmeter.

 

[shoggoth43]

Once you mount them it's a PITA to do anything soldering to them. I would guess that once you hit the "leak" to ground everything past that should drop. You could do the binary split method to find the leak. Cut the pair apart. One string works fine yes? Drop your amperage if you can for the testing of the other string. If it doesn't work, cut it in half and test each half. If one is good, work on the other. Cut the dead part in half and keep going as above. This should let you find the offending LED, or LEDs, faster without disconnecting everything.

One important note on heatsinking, you may want to make sure you can dump the full amperage into one string if you have two for just the above reason. As long as you can deal with the heat you should be ok if one string drops and the other has to take the full load. For the drivers I'm getting that would potentially be 1.5A into 13 LEDs ( yikes ) if one string drops.

I have not really run across a way to make make a parallel string play nice in the event that one drops. It seems like you'd be looking at burning out the other string in short order unless you can soak all of the heat or come up with some sort of LED matrix which lets you bypass an individual dead LED. What little I saw of that looked pretty complicated. Again, this is for the reefy guys and I see no need to blast 750 mA into my individual string, but it's a potential gotcha. You could maybe do a ballast resistor in each string to kind of balance the two, but you're still stuck if one dies.

The stars themselves should be electrically isolated for just this reason. You shouldn't have to do that, they should be built that way already. The only way you could accidentally ground them out is if the LED itself is poorly soldered onto the star and is jumping to the base metal. Unless your LED - solder work is somehow contacting any metal screws holding the star to the heatsink I can't really think of a way you'd be grounding it out unless there's a problem with the star itself.

Maybe that gives you some ideas of where you can poke around? I have no idea. Usually once I've overheated an LED it's dead. I mount my LEDs with screws and nylon washers to eliminate the screws as a possible ground path. There are heat conductive adhesive pads which are not electrically conductive you might be able to use on the voltage regulator. They have them available for the stars themselves though so that's another possible option. Maybe the thermal compound is causing some issues? Silver based compounds are good for heat but they will conduct electricity as well.

-
S

Unfortunately, this proved to be true. Yesterday the lights came on when the timer told them to. A few minutes later as I was getting ready to add the day's doses of ferts, the lights dimmed. I quickly checked and only 6 LEDs remained on. Those were the 1st through 6th from the + connection, in one 12 LED series circuit, with the other 12 LED circuit all off.

I disconnected it and stewed about it the rest of the day. This morning I removed the light and checked each LED individually - all are working. So, the problem is in the power system. But, I'm at a loss as to how 6 of a string of 12 LEDs can light up, and the others stay dark, with the LEDs all working when tested individually.

Any ideas? To help this along, I have 24 LEDs in two circuits of 12 each, with the two circuits in parallel, powered by a 48 volt DC supply through a IC Voltage Regulator acting as a constant current controller, set to 412 mA of current. I have a switch that lets me drop that current to 206 mA. I tried it and nothing changed except the intensity of the light. One more fact - it is very hot today and was very hot yesterday, but the air conditioning keeps the inside temp at the aquarium around 80 F.

EDIT: I found 3 questionable solder joints in the series strings of LEDs, and resoldered them, but it still doesn't work. Now, I'm wondering if this is entirely a matter of poor soldering in the strings, with the joints gradually increasing in resistance until that became too much. This: may be the problem with the soldering too. I had a lot more trouble with soldering today than I had originally.

Another possibility is that, since the "- "connection on the LED string is ground, and the heat sink is grounded, any leakage of current to ground before the end of the string would cause some of the LEDs to be off. If that is in one of the parallel strings, that particular string will get most of the current, leaving the other string without sufficient current to light the LEDs. But I can't pick up a leak to ground yet with an ohmeter.

 

[VaughnH]

Thank you for looking at this for me:

Once you mount them it's a PITA to do anything soldering to them. I would guess that once you hit the "leak" to ground everything past that should drop. You could do the binary split method to find the leak. Cut the pair apart. One string works fine yes?

Yes, one string works, but not fine. Only 6 of the 12 in series lite up. I can easily drop the current in half, unground the second string, and do this form of troubleshooting. Good idea!

Drop your amperage if you can for the testing of the other string. If it doesn't work, cut it in half and test each half. If one is good, work on the other. Cut the dead part in half and keep going as above. This should let you find the offending LED, or LEDs, faster without disconnecting everything.

Before doing that I need to buy a 25 watt soldering iron, RadioShack has them.

One important note on heatsinking, you may want to make sure you can dump the full amperage into one string if you have two for just the above reason. As long as you can deal with the heat you should be ok if one string drops and the other has to take the full load. For the drivers I'm getting that would potentially be 1.5A into 13 LEDs ( yikes ) if one string drops.

One reason I have the LEDs running at 400 mA is so a failure of one string won't cause the other to be overheated. It will only get 800 mA.

I have not really run across a way to make make a parallel string play nice in the event that one drops. It seems like you'd be looking at burning out the other string in short order unless you can soak all of the heat or come up with some sort of LED matrix which lets you bypass an individual dead LED. What little I saw of that looked pretty complicated. Again, this is for the reefy guys and I see no need to blast 750 mA into my individual string, but it's a potential gotcha. You could maybe do a ballast resistor in each string to kind of balance the two, but you're still stuck if one dies.

I tried to devise a ballast resistor configuration that would let me avoid problems when some LEDs quit, but the circuit continues in operation, but nothing I tried would work like I wanted it to. Eventually I just settled on using them at 400 mA.

The stars themselves should be electrically isolated for just this reason. You shouldn't have to do that, they should be built that way already. The only way you could accidentally ground them out is if the LED itself is poorly soldered onto the star and is jumping to the base metal. Unless your LED - solder work is somehow contacting any metal screws holding the star to the heatsink I can't really think of a way you'd be grounding it out unless there's a problem with the star itself.

My screws are flathead screws, with the conical back surface, so the only contact they make with the star is a line contact. I checked all of them with a magnifying glass to be sure nothing was close to contacting.

One thing has been causing me some worry: I don't use a "splash guard" between the light and the water, since the LEDs are so far from the water, and the fans keep the area ventilated when the lights are on. But, I wonder about condensation at night possibly being the cause of shorting to the heatsink. However, if that were to happen the water should soon evaporate and the short go away. It didn't.

Maybe that gives you some ideas of where you can poke around? I have no idea. Usually once I've overheated an LED it's dead. I mount my LEDs with screws and nylon washers to eliminate the screws as a possible ground path. There are heat conductive adhesive pads which are not electrically conductive you might be able to use on the voltage regulator. They have them available for the stars themselves though so that's another possible option. Maybe the thermal compound is causing some issues? Silver based compounds are good for heat but they will conduct electricity as well.

-
S

My thermal compound is silver based, but is there as a very thin coating, so I can't see it "bleeding" around the star to short out the contacts. I did my magnifying glass inspection looking for that too, and didn't see anything even close to contact. I do wonder about the quality of the star mounting, but I have no way to check that out. I know they are mounted not by Cree but by 3rd parties.

My plan tomorrow is to buy the 25 watt soldering iron, resolder all of the LED connections, then try it again. If that still doesn't work, I will try the "binary" troubleshooting, probably with a jumper wire to ground the string short of the end. Gotta be careful not to overheat anything doing that, but I think it will be safe, except for possible voltage regulator overheating, but it is mounted on the big heatsink too (electrically isolated).

 

[VaughnH]

Hip, hip, hurrah! Thanks to the troubleshooting tips from Shoggoth43, I found the problem! First, I disconnected the ground connection to the string that didn't light up at all. Plugged it in, and nothing changed. So, I reconnected that ground, and disconnected the ground to the string that had the 6 still operating LEDs. Plugged it in, and the same 6 continued to be the only ones lit up. This had to mean there was a grounded LED in that string. So, I removed the mounting screws for the last lit up LED and placed a piece of paper behind it to keep if off the heatsink. Plugged it in, and the previously unlit string came on fully, and the string that had the 6 still lit ones was dark. Absolute proof that that LED was shorted to ground at the "-" end of it. I still can't see anything wrong on it, nothing to show that it is grounded, but there is no question that it was grounded. To double check this, I then regrounded both strings, but left the faulty LED separated from the heatsink. Plugged it in, and both strings are fully lighted!

My tentative guess is that moisture between the star mount and the heatsink may have been involved, or the star mounting is not well done by the assembler. Tomorrow I will figure out what next. Any suggestions? Is there a relatively easy way to electrically isolate that one - with a mica shim, for example? Or would it then overheat?