PiLight: How it all started

The first really notable LED-related project I did started back in the year 2009, when two friends came up to me and asked whether I could help them out with the sound & light of a local club they wanted to start. The venue was already known to me (as it has been a club location before), and for monetary reasons of course as much of the equipment still in place should be re-used.

One of the previous owners most notably had made a huge investment and had installed lots of led tubes into the club furniture, groups of red, green and blue, hidden behind acrylic glass. I had seen them in action (and even worked the old lighting system) before, but at that point had no idea how they were actually wired or controlled. Nevertheless, I agreed, and this picture best shows what I found:

Many projects start off with a certain level of messiness.

That's about 80 pieces of Halogen Transformers, 12V AC. The bunch of cables coming in at the top, that's the multicores going to the actual LEDs, and the lot of cables hanging out at the bottom... well, that's the remainder of the previous owner ripping out the 220V regular lighting dimmers that were previously used to power the whole system.

Now i won't go into all of the problems that this posed (the most important problem I'm gonna skip here is: which transformer powers which LED? Just let me say: we eventually figured it out...) but would like to focus onto the electronic problems that this setup posed:

Those transformers were AC units, meaning: no recifiers and no filters. That led to a constant flickering of the connected LEDs at 50Hz.
Lighting dimmers usually are designed for ohmic loads (which fits most regular lamps). Transformers are inductive loads and most available dimmers are only able to drive a fraction of their announced power for such types of loads.
LEDs cannot be dimmed by varying the applied voltage. The effect that regular voltage dimming has to LEDs is that they will be far too bright at lower dimming levels.
An 80-channel dimmer bank would have been pretty expensive. (We worked out that the original system obviously grouped the LEDs and only consisted of four dimmers with 6 channels each)

So, besides the fact that we set out to reconstruct the original system, I started lots and lots of research in order to improve upon what would initially work.

One of the very first things I learned was that the key to LED dimming is Pulse Width Modulation, or short: PWM. So whatever I was about to build needed to integrate into a club lighting environment (that means: would require a DMX input) and had to output a PWM signal at the required 12 Volts and would need to output enough power for some meters of LED tubing (measurements indicated up to 1.2 Gigawatts, err, I mean Ampere).

At that point I was lucky enough to find Hendrik Hoelscher's DMX board. It's a basic board driven by an Atmel ATmega8515 microcontroller that can be used for various lighting purposes, including LED dimming. The sample schematics showed an ULN2803 chip used for driving high-power LEDs, but the spec sheet showed that the maximum load of that chip was 500mA, which was not enough. After a bit of searching however i found the "big brother", the ULN2064, which (despite the limitation to four instead of 8 channels) was able to drive up to 1.5A. Awesome!

So I ordered a batch of the DMX boards, the necessary components, a few ULN2064 chips, grabbed a few leftover LED tubes, and put it all together.

Assembled DMX Transceiver board with virgin inputs & outputs

After programming the mictrocontroller initial testing was performed with regular LEDs and the internal standalone demo modes:

DMX Transceiver with regular LEDs connected directly

Next, the testing circuit was hooked up to an old DMX lighting controller (left over from one of the previous owners... we had no intentions using it again initally, but still it turned out quite useful later on):

DMX Transceiver connected to DMX lighting console

Since everything worked smoothly at that point, the next step was to hook up a sample ULN2064 to the transceiver, and attach a sample strip of LEDs. An old computer power supply was the only one that would provide enough power, so I hooked that up as a power source. I don't have pictures of that stage, only a short video:

This video also shows the software that we used to control it all, DMXControl, and a first sample effect I clicked together for a quick demonstration.

Now that the whole proof of concept stuff was successful, all I had to do was wrap it up into some kind of form we could really use. The DMX Board was able to drive nine individual outputs, and one ULN2064 could amplify four channels each. That lead to the final channel count of 36. The fact that we would be able to cover 72 channels with two these devices also was pretty convenient, since that pretty much matched the amount of LED tubes in place.

In addition to the 4 DMX boards I also soldered a long row of 72 terminals onto a strip of perfboard, which would hold the ULN2064s and a row of fuses, too.

Also, a PC power supply was not suitable for the final installation, so it was swapped out for a professional Mean Well 320W switching power supply that was able to drive a whopping 25 amps in total. This was the most expensive part of the installation.

Here's a few more pictures of the process:

Laying out the components: DMX boards on the left, terminal perfboard with ULN2064s still in packaging on the right, power supply on the bottom. Also shown: 9V AC transformer used to power the DMX boards