The goal of this project is to develop an open-source firmware and controller for kinetic sculptures, mobiles and moving lamps.
So, once I had successfully obtained and decoded the data from the RGBW remote, I got a bit distracted and made an RC car:
So, not exactly a lamp controller, but hey, it is fun!
For this project, I intend to use 1W RGB LEDs that come in a "star" configuration. These are different from "addressible" LEDs in that they have no build-in electronics. What they have is six pads that connect directly to the anodes and cathodes of a red, green and blue LED die. Driving these is tricky, because each color LED has a different voltage drop. You could use resistors for current limiting, but you would be generating a lot of waste heat. One trick I've used in the past is to tie them all in series and then use an adjustable voltage bulk converter to drive them as one:
Although this works, it requires the bulk-converter to be tuned to a very precise voltage and I later found out, also carries a risk of thermal runaway, especially if the LEDs are driven near their maximum ratings. Of course, this method also makes controlling the individual LEDs from a microcontroller impossible.
For this reason, I set out to find a constant-current supply for my LED projects. One part I came across was the DM100-AC, which can often be found on eBay for around six bucks. There is little documentation on these boards, but they seem to have quite a following on a particular Christmas decoration forum.
These boards are quite sophisticated, combining a STM microprocessor, two MAX1487 RS-485 transceivers and a three channels of 300mA constant-current LED drivers. The supply voltage is from 9V to 24V and the board features built in rectifiers and a large electrolytic capacitor for power filtering, so you could power it using an AC wall wart, although DC works fine too (there seems to be a variant without rectifiers, the DM100-DC, available as well):
At the top, there is an unpopulated header which I presume is for programming, conceivably one could load new firmware into these boards, but sadly none of the unused GPIO pins are exposed as pads on the PCB, limiting the possibilities of using this as a generic dev board. This header does provide access to 5V and GND, which I was hoping I could use to drive the Arduino Nano. However, the power regulator is a 78L05, which is rated for 100mA, so probably not enough head room.
In my next installment, I'll discuss the interfacing of the DM100-AC to the Arduino.
So, I was able to find Vincent's github. It was hiding in plain sight. Although his objectives were most similar to mine, his project demonstrates one of the difficulties in dealing with these Milight bulbs: there are actually several 2.4 GHz radio chips out there that can be used for interfacing these bulbs. The bulbs and remotes themselves seem to have a radio called the PL1167, but radio I bought for this project is something called the nRF24L01+. Browsing the datasheets, these radios appear to be completely different beasts, but apparently some really smart fellow managed to coax them into speaking the same language. As clever as it is, this mimicry does make understanding the how things work a bit harder.
Anyhow, since Vincent's code targeted the PL1167, and I had this other radio, it wasn't able to run his code directly, but looking through his code, I was able to determine why I had originally been unsuccessful with OpenMili. There are actually four different models of MiLight bulbs and four different remotes. I offer a picture of three of those:
OpenMili only supports the RGBW model and I had been doing my testing with the RGB remote. Luckily, as I own all three, switching to the RGBW wasn't a problem. I am now able to get some captures going:
Not bad! (a big hint: If you are using OpenMili, "xr" starts a capture). I guess this is a start, but it does show that the original OpenMili repo isn't complete and does not support all the styles of remotes. It seems like Chris Mullins WiFi Gateway project may have significantly expanded on the original OpenMili work, so I'll have to explore that next and see whether I can pick out the bits of code I need to get the other remotes to work.
Or, I could simply proceed using just the RGBW remote for now. We shall see...
Cloned the openmili project and made my first attempt at trying to get the MySensors Nano RF Shield to detect any sort of packets from my remote, no dice. Still many questions to be answered, but getting a better understanding of the task that lies ahead.
There are several related projects out there that might provide clues, such as this and this and this, but one key differentiating factor is that these projects on sending commands to such lamps, rather than what I want to do, which is receive commands from a remote. Essentially, I am trying to make my own Milight-compatible lamp, which is compatible with the proprietary Milight remotes and Milight gateway, which I already own.
The only lead I've found so far that appears similar to what I am trying to do is this, but while there is a mention of a github page, a link to it is nowhere to be found.
Okay, everyone. I've created a concept video to demonstrate the goals I have for this project:
What is shown in the video is the mechanical prototype of my motion lamp you may remember from an earlier project. This video is a concept of the controller I want to make. The remote control does not actually work, nor does the lamp have any real electronics in it.
The lamp is hardwired to the DM-100AC, which in turn is hardwired to the 12V adapter. The DM-100AC is simply going through its built-in demo cycle, so there is no Arduino and the remote is there just for show.
The lamp is being turned by a microwave turntable motor in the base. This is a 120V AC motor with a separate AC power cord and no control at all. The plan is to replace the AC motor with a small stepper motor so it can run off the same 12V DC as the rest of the electronics. It would also be under the control of the Arduino, so the I could use some of the extra buttons on the remote to vary the speed.
The mechanical mount for the LED arm needs to be totally redesigned. Right now it is sits in place loosely -- half way though the video you see it catches a ride with the turning globe and flashed the LED at you. It shouldn't do this, but it looked kinda cool so I left it in the video.
There's still a lot of work to be done to make this lamp robust enough for actual use, but it looks kinda nice in the video!