SparkFun Forums

[SilverLogic]

Okay, so I'd like to try an experiment, I am trying to create a grow light made entirely of LEDs, something similar to the following:

http://www.shop.sunshine-systems.com/Th ... -98562.htm

There is a blog I've found where someone has created their own:

http://topledgrowlights.com/2010/10/cho ... right-led/

In his blog he mentions using 3mm 1w LEDs between the 450 nm – 470 nm, and 630 nm – 660nm range.

I am a novice programmer, I've created my own AVR prototyping board, so I have some decent knowledge of electronics (although most of my expertise falls in graphics), but I don't even know where to start with this. I was thinking about a using a proto board and just attaching 100 LEDs in parallel, but would this work? To be honest I don't know where to buy the LEDs that I am going to need, I found this one website, but the only ones they have in 630nm are 5mm, and in the aforementioned blog recommends 3mm.

http://www.superbrightleds.com/cgi-bin/ ... =/leds.htm

Any input would help, I am just trying to save some energy and a few bucks on the electric bill, and I don't wanna burn down the house

[phalanx]

I'm working on a design for a friend of mine that has a 75g reef tank which has similar requirements to your application except for the longer wavelenghts (red) which do not penetrate water very well. The array has 100 Luxeon Rebel LEDs in a 50/50 royal blue / cool white ratio. Since we were concerned about cost, we purchased the LEDs directly from Future Electronics which is the primary distributor for Lumileds. We also found that there can be a considerable cost savings if you place an order by telephone instead of buying through their online site. We were able to get the royal blue LEDs for nearly half their advertised price by calling them.

While driving LEDs isn't very difficult, you run into a number of problems when trying to drive 100 LEDs at 1W or higher. If you put the LEDs in parallel you will need a resistor for each one and it would be in your best interest to get a power supply as close to the forward voltage of the LEDs to minimize power loss in the resistor. Using a resistor as ballast is a fast way to lose any efficiency advantage LEDs had over incandescent.

The best approach is to use a switch mode constant current driver. These are effectively a switch mode power supply that are modified to produce a constant current instead of a constant voltage. There are a number of off the shelf and DIY solutions available. Meanwell makes a commercial constant current unit that has up to a 48V (I believe) output that will let you drive about 13 LEDs in series if they have a 3.5V forward voltage. That means you will need multiple units to drive all 100 LEDs and at $40-$50 each, the Meanwells can get very expensive. Facing that same situation, I opted to design my own controller based around the Supertex HV9961 which runs off of 120VAC and can provide string voltages over 90V to allow for up to 25 LEDs to be connected in series. I put 2 controllers on one board and have 2 boards for the entire array.

Here is the controller with only 1 channel populated:


Here is the controller turned on with 24 cool whites at 750mA (about 2.5 Watts per LED).


Here is the entire array. Notice the large heat sinks that are necessary to keep the array cool. I'm still in the process of soldering down the wiring so I don't have any pictures of the entire array turned on. The gap in the center is to accommodate a brace in the fish tank.



I hope that gives you some idea on what is involved in making a large array of power LEDs. It's not the easiest project on Earth but it is certainly feasible.

-Bill

[wonderfullyrich]

I apologize if this is a tangent from the original project, but I've got a similar project that I'm trying to figure out how to spec and build. I'm a newbie to LED lights, but figure this has to be achievable.

Similar to the grow light and the tank lights, I'm trying to drive about 80 violet (420-470nm) LEDs for the purpose of making a Bili Light. The point isn't to use them here, rather I'm trying to do a test build here and figure out what parts and instructions I need to assemble them in Uganda (or Eastern African in general) where new born jaundice is a larger problem. Several people have made these before, theres a student at Rice that made one and I've seen a listing at Makers Faire in Durham (using the same Royal Blue Luxeon Rebel LEDs), but I haven't been able to get in contact with anyone who might know what parts I need and give me guidance on assembly. I'd like to build this based on a 12volt battery (readily accessible).

Any chance you can help me out? I get your point about being in parallel and getting enough forward voltage and I can see that being my biggest stumbling block. Can this be done with a buck to buck converter and the HV9961?

Rich

[phalanx]

Hi Rich,

There are 50 of the Royal Blue Rebels in the array picture I posted which fall into the spectrum range you are looking for (min 440/max 460). If you want to power 80 of them from a 12V battery, you will need some kind of boost controller to get the voltages high enough to reduce the number of controllers that you need. If you used a simple buck controller like the HV9961 in my application, you would need one controller for every 3 LEDs in your array. The HV9961 is only designed for buck applications. What you should be looking at is something along the lines of a Supertex HV9911. This controller can be configured to run in boost mode so you can generate string voltages much higher than your input voltage. Take a look at the application note on the website for that controller. It goes through a design iteration for both a continuous and discontinuous conduction boost controller with significant voltage output levels. The math may look a little daunting but it's manageable and you should be able to substitute in values for your application. To keep the component sizes reasonable, I would try to limit a single controller to 20-25 LEDs each which would require you to run 4 controllers for your array of 80.

Don't forget that in power conversion applications like this, the ideal formula says Power In = Power Out. If you are running a string of 20 3W leds, you will be consuming 60W of power. This means that your 12V battery will be sourcing about I = 60W/12V = 5A of current for 20 LEDs. Multiplying by 4 strings gives you about 20A for all 80. Due to efficiency losses, in the real world that number will be a little higher. You will have to make sure your power source is capable of sustaining that kind of load and all connections to and from it are up to the task.

One way to make this easier is to reduce the number of LEDs in your array. The higher efficiency binned rebels put out quite a bit of light. You may not need 80 for your application.

-Bill

[anhingus]

hello, rich -

at the moment i have 72 leds in parallel on a board, 5-volt wallwart, one
resistor as current limit. works well.

looking at specs on your royal blues, try using a 12-volt wallwart, an LM7805
voltage regulator, and a 5-watt multiturn wirewound pot for your resistor.

making assumption you are considering an automotive battery for where your
project is headed.

used to be, maybe still is, american electrical stuff had problems working with
voltage convertors in certain african countries so the good folks just plugged the
110-volt stuff into what was supposed to be 220-volts and all was fine until the
day they actually got 220-volts.

anyway - cheap test - use an LM7805 in series with 12 volt wallwart, put a
ten or 20 of your leds in parallel, use a wirewound pot for ballast. use an ammeter
to set current.

you like? scale up.

good project. good luck!

[phalanx]

anyway - cheap test - use an LM7805 in series with 12 volt wallwart, put a
ten or 20 of your leds in parallel, use a wirewound pot for ballast. use an ammeter
to set current.

you like? scale up.

You can't even remotely source enough current to fully drive 10 or 20 power leds in parallel with a setup like that. If you could, with 10 LEDs the 7805 would dissipate 70W, the resistor would dissipate another 15W, and 10W for the LEDs giving you a grand total of 95W of power. This equates to having about 8A available from your wall wart which isn't going to happen. Scaling up you will need an 80A 12V supply for 950W of power dissipation when only 80W are coming out as light. This gives you an efficiency of 8.5%.

Parallelling power LEDs with a single ballast resistor is never a good idea. Parts from within the same production run can have widely different forward voltages which could result in one LED hogging current and prematurely burning out.

If you need to check if a power LED works, most multimeters can get them to dimly light using the diode check function.

-Bill

[anhingus]

anhingus suggests that ohm's law is linear, diodes ain't.

last he heard, the 'd' in led stood for diode, but that was some time ago.

LM7805 - check the national data sheet - can be used as a constant current source

prudent, methinks, to breadboard stuff before commenting.

[mistergreen]

phalanx,
Nice to see an aquarist on here.. I have freshwater aquatic plants though.

[phalanx]

anhingus suggests that ohm's law is linear, diodes ain't.
last he heard, the 'd' in led stood for diode, but that was some time ago.

I (first person) fail to see how this invalidates anything I said. A power LED's I-V characteristics are nonlinear but its forward voltage is nearly constant while in conduction over its normal current handling range.

LM7805 - check the national data sheet - can be used as a constant current source

I'd check again. The LM317, which is the adjustable version, can be used as a current source. While linear regulators have their place, controlling current to more than 1 or 2 power LEDs is not it. They are far too inefficient when you have a large disparity between your source voltage and LED forward voltage.

prudent, methinks, to breadboard stuff before commenting.

I would argue that it's also prudent to understand what is being discussed before offering solutions that in a best case scenario will simply not work or in a worst case scenario could quickly burn up a number of expensive parts. There is no need to breadboard your solution since even a cursory look into the numbers shows it being far from realistic. Parallel circuits, ballast resistors, and linear regulators are poor design choices for high power LED applications.

-Bill

[phalanx]

phalanx,
Nice to see an aquarist on here.. I have freshwater aquatic plants though.

Hi mistergreen,

I wish I could say the tank this is going above is mine but I simply don't have the time to maintain a salt water tank. I'm building this as a favor to a friend of mine. I have a low maintenance 46 gallon freshwater tank but without any live plants.

-Bill

[wonderfullyrich]

This definitely is putting me on the tracks.

I figured out that the newborn needs 30 μW per cm2 per nm over a 24-72 hour period. This should help me figure the necessary number of Royal Blue LEDs needed to do this, but I'm slightly lost in how to figure how much light I'd actually need based on 30 microwatts per centimeter squared per nanometer?

For a rough estimate is it literally as simple as dividing the microwatts of the LED by the surface area it's shining on to see how and multiplying the number of LEDs to get the necessary 30microwatts? I've completely left out the distance issues between the light and the surface, focus of the light, but am I on the right track? If so then to meet the requirements (assuming a rather large child with a body surface area of 2700 cm2 and assuming I'm only likely to reach about 50% of the surface area) then I'm looking 16-17 of these royal blues (using a LXML-PR01-0425 with 700mA current the specs sheet says it has a typical performance of 740 microwatts). 30μW cm2(740μW/(2700cm2*0.5)) = 16.444 LEDs

Bill I'm understanding you are saying if I was to use an array of 80 of these that I'd need 20A to drive this. Anhingus as I read it you are saying I'm going to need something much much less. Right?

Rich

[phalanx]

Hi Rich,

I don't have much experience with the formulas involved for optical calculations so I really can't give you an appropriate answer without a lot of background research on my part. I've never seen a unit that is microwatts/cm^2/nanometer. I do know that it will not be as simple as taking the microwatts of the LED and dividing it over the area you need illuminated. You will have to account for the efficiency of the optics you use, the amount of light that actually falls on the body, the energy distribution within the pattern coming from the optics (weaker towards the edges), and reflections inside the target area. Someone more familiar with the issues may have to chime in.

Anhingus's suggestions do not work for your application and will do nothing more than cost you extra money if you choose to pursue making your Bili Light. Using his method would require your battery to source 80A of current while only a small fraction of the overall power is transformed into light. The rest would be wasted as heat (and it would get pretty hot). This ignores the fact that the components and arrangements he suggested physically cannot handle the power dissipation in this application.

If you use a proper switch mode power supply, for quick estimates you can calculate power draw by using the ideal formula: Power_in = Power_out. Assuming you are running 80LEDs at 3W each off of a 12V battery, you get: 12V*I = 3W*80. Solve for I and you get I=20A which means your 12V battery will have to source 20A.

In the real world there will be inefficiencies in the power electronics which assuming you can achieve 85% efficiency makes the equation look like this: (12V*I)*0.85 = 3W*80. Solve for I and you get I=23.5A which is a more realistic estimate of the current drawn from your 12V battery. If you reduce the current to the LEDs and run them at a lower current level, you will proportionately reduce the current demand from your 12V battery.

-Bill