Designing and Building a LED Fixture

[CL]

It's lookin' sweet so far. Good luck! That heat sink looks awesome!

[malaybiswas]

Originally Posted by Hoppy

More parts purchased, at Radio Shack:


A DPDT rocker switch - $4.35 (tax included)
A PC board for resistors - $4.00 (tax included)
Roll of hook-up wire - $7.20 (tax included)

Total cost so far: $70.50, and still no LEDs purchased! Of course I had to buy a little soldering iron and some solder, too, but those don't count, being "tools".

Looking good. Just curious about what you are planning with the resistors.

[Hoppy]

Originally Posted by malaybiswas

Looking good. Just curious about what you are planning with the resistors.

The resistors became:

This is the current limiting resistor array. The DPDT switch will let me switch from about 350 mA to about 750 mA current. And, the resistor circuit is arranged so if a LED burns out it won't cause a huge current through the parallel LED array.

[Hoppy]

Today I finished the power supply/cable assembly.



On the left, on the end of the cable is the "plug" half of the hard drive extension cable, with two pins carrying 12 volts, and the other two carrying 48 volts, with the green ground wire loose, to be attached to a ground screw on the back of the fixture, to ground the heatsink.

The cable is the only 4 conductor wire I could find, at HD, and it is phone hookup wire, #22 wires. Any thoughts about whether that is enough copper to carry 1.4 amps? (It was cheap!)

[malaybiswas]

Since you have separate power supply for LEDs and fan, why are you connecting the wires?

Typically 22 gauge is ok for the fans. For LEDs most manufacturers recommend 18 gauge.

[redman88]

Originally Posted by Hoppy

Today I finished the power supply/cable assembly.



On the left, on the end of the cable is the "plug" half of the hard drive extension cable, with two pins carrying 12 volts, and the other two carrying 48 volts, with the green ground wire loose, to be attached to a ground screw on the back of the fixture, to ground the heatsink.

The cable is the only 4 conductor wire I could find, at HD, and it is phone hookup wire, #22 wires. Any thoughts about whether that is enough copper to carry 1.4 amps? (It was cheap!)

CAT 5 Cable should be able to carry that load.

[Hoppy]

Originally Posted by malaybiswas

Since you have separate power supply for LEDs and fan, why are you connecting the wires?

Typically 22 gauge is ok for the fans. For LEDs most manufacturers recommend 18 gauge.

The wires are not interconnected at all. The cable is a 4 wire cable, so I have two wires for each power supply. The 48 volt supply also has a ground wire, which is a straight connection to the ground prong on the 110 VAC plug. I ran that one to the shield wire of the phone cable, in order to have a ground for the heatsink.

CAT 5 cable was one alternative I had, but I didn't find any without plugs on both ends, greatly driving up the cost. If 22 gauge is really too small I will switch it over. I only wasted $2 on the 6 feet of phone cable.

EDIT: I just did some more research: CAT5 cable uses 24 gage wires, which are smaller than the 22 gage wires I now have. Also, I found that the maximum voltage drop I will get on my wires is about 0.8 volts, where 18 gage would give me 0.2 volts drop. I can't see this as significant, so I will leave the cable as is, until testing shows some problem needing a solution.

The only advantage to CAT5 cable is that I could use two pairs of wires for one circuit, and the other two for the other circuit, giving me more current capacity. Since I have a spare Ethernet cable I may salvage it just for that reason - assuming 2 24 gage wires are more copper than 1 22 gage wir.

[malaybiswas]

What's up! any progress?

[confuted]

Sorry I didn't catch this earlier.

LEDs, as you know, are semiconductors. In normal materials, like metals and carbon, resistance goes up as the temperature increases. Heating losses go as I^2*R, and V=I*R, so at a fixed voltage, as the temperature goes up, the current goes down -- the process is self limiting. Semiconductors, however, work the opposite way - as the temperature goes up, the resistance goes down, causing MORE current to flow, causing it to heat faster ... you get the picture. BOOM. Or melt. Or fizzle. Something along those lines.

That's why you put current limiting resistors in -- which you've done. But you shouldn't connect LEDs in series, since they aren't all identical (they look identical, but manufacturing tolerances say otherwise). Some LEDs will have slightly lower or higher resistance than others, or a different forward voltage, or different temperature coefficients. You can't count on all the LEDs in the chain drawing the same amount of power - especially as they start to heat up. It's entirely possible to have one of your LEDs run away and start a fire.

I'd strongly advise you to connect all of your LEDs in parallel, with a current limiting resistor for each, not in series (or any combination of series and parallel).

Note: Though I am an electrical engineer, I take no responsibility for your project, whether you follow my suggestions or otherwise.

[Hoppy]

Progress: Still waiting for DealExtreme to ship my LEDs. Hopefully that will occur tomorrow. And, the more I think and learn about LEDs the less I like my resistor setup for limiting the current. So, I'm scrapping it in favor of a limiter shown in Instructables, which uses a NFET and a NPN transistor to control the current. I can make two limiters, one for each series string of 12 LEDs, for about $5. That sounds like a winner to me.

Confuted, you have it backwards. Parallel LEDs can easily have one LED run away and destroy itself, followed by all of the others. LEDs in series will always all have the same current going through them. Also, put 24 700 mA LEDs in parallel and you need to supply that circuit with 16.8 amps of current, which is a problem with small DC power supplies. It is much easier to provide 48 volts at 700 MA than 3.5 volts at 16.8 amps.

[JimmyYahoo]

Hoppy, cant wait to see how this turns out. Been following some LED threads over on the reef forums and had the urge to try this over freshwater. My bank account is happy you beat me to it. Its great to be able to watch the progress and have detailed explanations along the way.

Any thoughts going forward about dimming/automation/color supplementation?

[Hoppy]

Originally Posted by JimmyYahoo

Hoppy, cant wait to see how this turns out. Been following some LED threads over on the reef forums and had the urge to try this over freshwater. My bank account is happy you beat me to it. Its great to be able to watch the progress and have detailed explanations along the way.

Any thoughts going forward about dimming/automation/color supplementation?

One man's bank account loss is another man's gain?

I have thought about how to get the great plant and fish color enhancement that GE9325K lights give, perhaps by adding a few red LEDs, but I doubt that I will try it. I have found that my eyes adjust to whatever lighting I use, and eventually the differences caused by the light spectra differences fades away in my perception. As far as dimming goes, I will use a switch to be able to toggle between 350 mA and 700 mA current, but I have no interest at all in "dawn to dusk" style lighting. For now I'm going to be very happy just to get this light to work, and get enough data to be able to predict, with reasonable accuracy, what configuration of LEDs will give a specific PAR value.

[Hoppy]

Here is the schematic for the constant current "driver" I plan to try to make today:



These are all tiny parts, with the 4 semiconducter parts costing about $2.50 total, and the resistors costing about $3.00 total. I can probably salvage a piece of the circuit board I used for my ill-fated resistor setup for this device, and use the DPDT switch I already purchased. I'm still thinking about the need for heat sinks for the MOSFETs due to the high current through them. I would just mount them on the LED heat sink, but I suspect that would make the heat sink electrically hot.

[malaybiswas]

Originally Posted by Hoppy

One man's bank account loss is another man's gain?

I have thought about how to get the great plant and fish color enhancement that GE9325K lights give, perhaps by adding a few red LEDs, but I doubt that I will try it. I have found that my eyes adjust to whatever lighting I use, and eventually the differences caused by the light spectra differences fades away in my perception. As far as dimming goes, I will use a switch to be able to toggle between 350 mA and 700 mA current, but I have no interest at all in "dawn to dusk" style lighting. For now I'm going to be very happy just to get this light to work, and get enough data to be able to predict, with reasonable accuracy, what configuration of LEDs will give a specific PAR value.

In my experience and opinion red LEDs tend to saturate the color which is more straining to eyes than pleasing (at least to me). I like the combination of blue though.

[Hoppy]

Originally Posted by malaybiswas

In my experience and opinion red LEDs tend to saturate the color which is more straining to eyes than pleasing (at least to me). I like the combination of blue though.

I noticed in reading other threads on other boards that adding a color can give different effects than you expected. That was why I started with just white ones. Also, it would probably be best to look for a LED that gave a color more like the red-violet of the 9325K bulb, rather than red. Finding that wouldn't be easy.