Author Options:

Large-scale LED build request for advice? Answered

As I was putting up my Christmas lights over the Thanksgiving weekend, I had an idea (cue ominous background music). The idea evolved from a wish to simplify installing the lights on the 2nd storey roofline into a set of fully-addressable RGB C7 lights that would make Clark Griswold proud. As a proof of concept, I want to replace the boring string of 25 multi-color LED lights over the peak of my garage (24 ft of roofline). I looked at some available products that could be used as a starting point.
From the adafruit store:
12mm Diffused Digital RGB LED Pixels – 4 strings = $180
20mm Clear Digital RGB LED Pixels – 6 strings = $270
Available many places:
GE Color Effects™ 50 LED Color Changing Light Show - $99

The Color Effects strings would be expensive when you consider the 200+ feet of roofline and guttering that I want to cover, and would difficult/impossible to sequence across the whole house. While the RGB pixels from adafruit would give me the most flexibility in creating animated sequences, the prices work out to $2.00 - $2.50 per light. To make this a reality, I need to do this on a $1.00 per light budget for the pilot project. Preferably $0.80 per light for the whole house.

So, I want to light the peak of my garage with 48 RGB LEDs spaced out over 24 feet (6 inch spacing between lights). The string would be controlled by an Arduino and a handful of TLC5916 constant-current sink drivers. Each group of 8 RGB LEDs would have a group of 3 TLC5916 chips, so I am looking at 18 chips. I am not worried about the programming, and there are plenty of examples to look at for the TI chips. I plan to use a power supply from an old computer to provide the juice (up to 3 amps with every LED lit). If this was something that fit on my desk, I would dive right in and start experimenting. However…

I am worried about the 24 foot run from the first light to the last. That is a long way for a signal to run to latch the chips. Do I need some sort of repeater for the latch signal? That is a long run for 5V power. If I am switching 144 channels on and off quickly, will there be voltage drops that could introduce bugs into the system?

Does anyone have experience with the sort of large-scale installation I have in mind? Please help a newbie by offering any cautions, advice, and examples to build on. I will definately document the build and post instructions if I manage to sneak this past the wife.  Thank you!


A few tricks:

Use a much slower SPI bus than you would normally expect to do, and whack the driving lines from a HC125 buffer chip.

If you're cascading Dout-Din, stick another buffer for the lines on each cluster of chips - again HC125s are great for that - put a nice small pull up resistor on the lines - I'd try a 470 Ohm, or maybe 100Ohm.


Every addressable pixel I've found is over $1 each.

my ws2801 strips from sparkfun were 45/32 leds, and they were spaced on one inch (ish) centres...expensive for a roofline.

The ge strings can be hacked to have a controller control more/less of them and more customizable settings.

I really want to create lights that look like the C7 lights on the rest of the house so that they still look like traditional Christmas lights, so spacing them out is important to the look. The tighter spacing of some of the LED strips would be too much. The cost could possibly reach the $1 point with the design I have in mind, even with the bulbs, LEDs, and the wood to mount them. The pixels aren't really "addressable" like the other ones I mentioned above or like you describe, and they would need to have software PWM. The microcontroller would have to keep refreshing the string to mix the colors, while the more expensive versions handle that with the chip built into each pixel. My biggest concern is that there could be problems I don't understand in working with very long electric runs and trying to send a signal pulse that far. Everything I have done with my Arduino and LEDs has fit in the palm of my hand.

Any time you need to send pwm over a distance you run into a lot of problems with loss, noise, attenuation, etc.

The strings that have chips on each segment do so because then they cascade and amplify the instructions being passed down to the next segment, there's technically never a long run of serial data running down the pipe, just several short segments. It adds a propagation delay but it's imperceptible.

So, do you want 'changable colours along an entire string' or 'addressible bulbs'?

If its just dimming 3 colours down one strand it's super easy.

I want to change each bulb independantly to create dancing patterns of holiday goodness.  I love the idea of each bulb being "smart" by itself and passing the serial data on to the next bulb, but that is what contributes to the higher cost, which is my biggest limiting factor.  I think that by using a "sort-of-smart" cluster of bulbs and finding a way to pass the signals on to the next cluster reliably will get me down to the price point I am looking for.

I think that trying the HC125 buffer suggested above might just work.  I also will try putting the controller in the middle of the string rather than at the end, which would cut the maximum distance to 12 feet.  I think that the longest signal run will be something like 24 inches between clusters.

This will not be ready for this Christmas season, but I will definitely have some sort of pilot project ready to go to demonstrate later this year.  If that works, I will have to start looking at the rest of the house.  What sort of budget will that project need? How many controllers?  Will I need wireless serial to sync the whole house into a single pattern?  How could I sync the lights to music?  Will my wife veto the project?  Will I slowly slip into maddness trying to de-bug the thing?  Tune in next week (or year).

The monster has been there all along. We have merely unleashed the beast and pointed it towards a major city for a friendly rampage. GRARR!

A note on the project status: I have been crunching the numbers seriously, trying to determine how much money I might need to budget. I was hoping to hit $1 per light, half the cost of the best commercial alternative. It looks like the price for the pilot project might be closer to $1.50. As it turns out, the wire could be as much as one-third the cost of the project, 50 cents per light. Ugh.