The 2007 RGB Christmas Light Project: Overview & Design Philosophy
(Entries below are in reverse chronological order, so scroll down to read the beginning of the story.)
Update - 11/20/07
Paul from the East Coast found this website and ordered a handful of pixels. He writes:
Hey John, thanks again for everything! They work great, look great, and I already have about 5 years worth of projects planned for them. :)
Joel from the Midwest bought 32:
DAMM those pixels are bright! Had to run down the UPS man at 7:30 but finally have them. Well worth the wait. Thanks for your efforts, John.
I'm working on a controller and breakout boards for the lights right now. The roofline has six discrete light runs, so I need to distribute power and data in a way that looks neat and clean (or better yet, is nearly invisible).
The guy who did our landscaping installs Christmas lights during the winter. He offered the services of his boom lift, which I think I'll take advantage of. The roof peaks just over 30' above the ground. I'm not a fan of wobbly ladders.
Update - 9/28/07
Earlier this week my overseas factory delivered 30 assembled pixels for my review. I've very pleased with the results. Videos and photos have been posted in this forum:
http://www.christmasinshirley.com/forum/viewforum.php?f=33
Look for posts by 'JEC' for more details and discussion.
Here's a short sample video I shot last night:
http://www.youtube.com/watch?v=FF1YUCV_JAw
A stack of circuit boards awaiting testing:
Each board has 2 x RJ45-8 (aka CAT5) connectors on board. This allows the pixels to be daisy chained together without needing to solder each joint by hand.
Update - 7/25/07
A few days ago a set of prototype circuit boards arrived from the factory. I spent a few minutes soldering the components in place, then applied power and data for a first test.
The results were very impressive. The six LEDs of each color are extremely bright and the color fading is smooth and responsive.
I found an empty 5 gallon paint bucket and inverted it over the lit pixel. The bucket was handily illuminated. I'm cautiously optimistic that the entire display will work very, very well.
I'm making one modification to the circuit board: the original design has simple 5 pin headers for data and power. I'm revising the schematic to include 8 conductor CAT5 connectors. This way, I can chain pixels together using off-the-shelf network cable. The twisted pairs will ensure the integrity of the data, plus I have three conductors each for power and ground.
Thanks for visiting - check back often for updates.
And if you're interested in grabbing copies of the circuit board yourself (I'm working on quotes for overseas fabrication) drop me a line.
John
Update - 6/14/07
Score!
DMX Reception and LED dimming are now merged peacefully together. Only task left is to edit the code so that each pixel's start address can be programmed on the fly. This way, the same firmware can be burned into each pixel. Later, after assembly, the addresses can be programmed. This can be repeated as many times as necessary for different pixel patches, configurations, etc.
Update - 6/08/07
Things are really moving along!
I've learned that the PIC I'm using (the 16F688, chosen purposefully because it contains both an EUSART and an internal 8 MHz oscillator) is spinning along happily.
Over the past few weeks, I've written rock-solid code for DMX-512 reception and LED dimming. The most exciting part is that I've dispensed with the external crystal, which will save ~$0.70 per pixel in the final design.
Currently, these routines are running in smaller separate programs. My task now is to integrate one with the other so that they'll coexist peacefully.
Also, I've ordered samples of several different types of LED, which should begin to trickle in over the next week or so. The goal of this project is not to illuminate an entire stage or side of a building. Rather, I just want bold, saturated colors for each of the pixels.
So I can dispense with the (nifty but money and power-sucking) Luxeon multi-watt LEDs and work with something smaller.
Plus, I'm experimenting with a few waterproof enclosures which should allow for excellent light diffusion.
More updates to come. Thanks for taking a moment to stop by.
Background - Posted 5/7/2007
Two years ago, and on the spur of the moment, I built a set of color changing pixels to decorate my balcony with at Christmas time. They were controlled by a regular DMX-512 signal generated by an iMac G5. You can read about my adventures by clicking on the '2005' links posted here.
My wife and I spent the end of 2006 preparing for the arrival of a new baby girl, so not much happened on the technical front. I did, however, consider many new designs for the pixels. The first versions worked well, but were very expensive, very un-waterproof and cranky to install.
After much mulling, sketching and re-hashing, I've arrived at a new design to be debuted live December 1, 2007. It seems appropriate to design a larger system to match the new house.
Some highlights:
Each pixel will be contained in a 6" acrylic globe. To make the concept shown picture above, I placed one globe on the front porch, then snapped the picture. Using Photoshop, I copied and pasted the globe over and over along the roofline until I was happy with spacing.
The row of pixels on the upstairs balcony probably will be eliminated. And the horizontal roofline to the up-right of the front door will have pixels.
In all, I think I'll need about 130 units.
Each pixel will contain a Microchip PIC 16F688. It's the smallest, cheapest of the PICs which still contains a USART. So the DMX512 reception will happen in hardware. The PIC contains an internal 8 MHz oscillator. Testing this summer will determine if it's stable enough to capture DMX or if an external crystal will be required.
Pixels will be connected together in strings of up to 32.
Each pixel will contain 3 each red, green and blue LEDs. The LED colors will be connected in series and driven by the PIC through a small FET (probably a 2N7000, but I still need to do some testing). Since the total current through each LED string won't exceed 20 mA, it should be easy to find some small SMT FETs to drive each string.
Each pixel will have a proper RS-485 receiver chip (probably an SN75176 unless I can find something cheaper). Shielded signal lines are a good thing.
Changing LED brightness will be based on Artistic License's superb BAM algorithm (look at App Note #11). It's deviously simple and consumes very little processor time.
The system will be driven by either a larger PIC or one of Parallax's new Propeller chips (8 x 80 MHz, 32 bit processors running parallel!). Either the PIC or the propeller will generate DMX sequences, color fades and so forth.
If you're interested in grabbing some of the circuit boards and/or programmed PICs for yourself, send an email to 'lights AT response-box.com' and we can talk.
More soon!
John