I was walking the streets of midtown Manhattan when I came across a large discarded film viewer on the sidewalk. I had a leatherman with me so I decided to open it up and see what was inside. This being New York, no-one paid any attention to me. When I got it open, I found that the light source was a large panel with 247 LEDs organized as 19 rows of 13 columns. I grabbed it and continued on my way. I assumed that all the LEDs were white, but when I got it home I found that it had alternating red, blue, and green LEDs. Presumably, the purpose was to rapidly visualize adjustments to color balance in the film viewer. It was exactly the part I was looking for to build a lamp with arbitrarily adjustable color. A full-color lamp lets one set the color tone of the room (in this case, my office). One can make suble adjustments around white, or set a radical tone with a deep color.
There were a couple of problems with the lamp, of course. One was that the red channel had blown out. I replaced the 83 red LEDs with new ones. A second problem involves how the panel was wired up. Within a color channel (red, green, blue), all of the LEDs are connected in parallel. Getting high brightness out of this source requires 5V at nearly 1A for each channel - 3A total. A second problem is that the red, green, and blue channels all share the same common anode. This topology rules out many inductor-based constant current current drivers. I settled on using the STCS1, which is a constant-current driver which uses only drain-switching to adjust the current, and which accepts PWM dimming.
I had some mousepads lying around that I had purchased from an electronics surplus shop. They seemed perfect for "haptic" control over color and brightness. I did some research and discovered that they use the PS2 mouse standard. Now I just had to write a PS2 mouse driver. Via web search, I found a couple of articles (this is one) that describe the PS2 protocol (the mouse takes 4 physical inputs: +5, ground, clock, and data. Pull clock and data high with a, say, 10K resistor). I wrote a driver (which is a little tricky because one must ensure that the driver won't hang on bad input), hooked up the mouse ... and nothing happened. I re-read the articles, fiddled around for a while, and discovered that the mouse requires an initialization process: 1) wait for the mouse to report that it is "ready", and then 2) send a command to put the mouse into "stream" mode. Once in stream mode, the mouse generates reports of physical interactions. After that, developing the code was just regular programming.
I wired everything up, turned on the power, and was able to get fine-grained color adjustments by moving my finger on the mousepad. The next problem I ran into was heatdisappation in the STSC1's. They come in a SO-8 form factor, with a metal pad underneath for cooling. I soldered them to SOIC adaptors, which was probabaly a mistake since it covered up the metal pad. At full power (1A), the LEDs started flashing - which I traced to thermal shutdown protection. I put heat sink compound and heat sinks on the drivers, and that resolved the problem for the blue and green channels. The red channel continued to have problems - because the voltage drop is lower. I discovered that by using .5 ohm power resistors in the red channel, I could limit the current / voltage drop for the red channel enough to avoid overheating the driver (in retrospect, just using these inexpensive fractional-Ohm power resistors for current limiting might have been cheaper and easier). I finally settled on using three .5 ohm resistors for the red channel, and limiting the current to .67A.
Below is a picture of the three main parts of the full-color lamp: the circuit board, the mousepad, and the LED panel. I put a diffusing panel (made by sanding a sheet of lucite) on top of the LED panel to better blend the light.
Here is a closeup of the board, showing the heat sinks on top of the LED drivers, the current-sense resistors on the left, and the red-channel current limiting resistors in the middle. The extra capacitors are from the test of a theory that the LEDs were flashing because of an electrical instability. That theory proved false. On the bottom right is a connector for a serial terminal, which I used for debugging.
This picture is of the lamp assembled into a box. I drilled a huge number of ventaliation holes. the right mouse button turns the lamp on/off; the led on the mouse panel is red when the lamp is off, green when it is on.
Here is the lamp tuned to blue
Green
And warm white.
