I did some more work on the light module for Ava’s desk last night. The goal is to construct a full-spectrum array comparable to a 500W halogen work light, to illuminate her work surface and help stave off the Seattle grey-sky winter blues. I estimate that it should be possible to produce a comparable level of light using only 150W. Of course LED-based lights are available at all brightness levels now, but I am not aware of any commercial full-spectrum units yet – and after designing this project it is pretty easy to see why. Give it another year or two and I think this sort of light will be available off-the-shelf, but right now it is still a challenging hobbyist project. Perfect for me to play with!

As per usual, I have gotten myself in a little deeper than I expected. Every LED project I’ve done previously starts with a fixed-voltage DC power supply, then I add appropriate current regulation for each LED circuit. This time I knew I was going to be drawing a great deal of current, so I thought it would make more sense to run all the LEDs off 120VAC. This is not as crazy as it sounds: it’s the same thing all those LED christmas light strings do, only I’m using much more powerful emitters. Instead of running all the lights in parallel, regulating each one independently, I decided to run all the LEDs in series with a single regulator: dividing a single power loss across all the lights, instead of multiplying it by the number of lights.

I started with a bridge rectifier and a large capacitor; the output, I expected, would be roughly 120VDC. Each of the Bridgelux emitters drops 8.9V, so I wired up a dozen of them in series. Then I threw in an LM317HVT linear regulator, rated for 40V / 1.5A, and configured it as a 1-amp current regulator.

This did not work. The lights came on briefly, then something went “pop”, and the lights went out. In the second or two before I managed to disconnect the power, two of the LEDs flared on and went “pop”, leaving a black smudge across the lens. Uh oh.

Further testing revealed that the output from the rectifier was nothing like 120V – instead I was getting something close to 170V! Now of course the whole point of alternating current is that the voltage alternates, swinging back and forth constantly. When we say that wall outlets supply “120 volts” or “110 volts”, that’s the RMS average – at any given instant, the outlet voltage may be anywhere between -170V and +170V. I expected that my little rectifier/capacitor setup would smooth out the peaks and valleys, giving me a nice steady 120V, but it appears that I’m drawing so little current that the voltage just stays high all the time. This is just a guess, really – it would all be so much easier if I had an oscilloscope.

I think the first “pop” was the LM317HVT, which is designed to handle a maximum of 40V difference between input and output. On encountering a 60V difference, it must have fried itself closed; with no remaining current regulation, the LEDs started overloading, and thus the bright flash and the pop as the magic smoke escaped.

I ordered a bunch of replacement emitters, replaced the two damaged ones, and added six more to the chain. Now I have eighteen of the Bridgelux units wired up in series, for a total drop of 160.2 volts. I also threw in a 10-watt, 68-ohm resistor to help out with current regulation. This seems to work, and it is almost hilariously bright. I should be able to replace the LM317HVT now and keep the current set where I want it.

If I can get this to work, the next step is to augment the “neutral white” spectrum with three other frequencies. “White” LEDs have a big spike in the blue range and a softer, broader lump in the yellow/green; this setup stimulates all three of our color receptors more or less equally, so the light looks white, but it doesn’t actually illuminate colored surfaces evenly. The reds are very weak, and there’s a noticeable gap in the cyan as well. My solution is to interleave colored LEDs with the white ones, adding 660nm (deep red), 627 nm (red), and 505 nm (cyan). I’m sure somebody, somewhere, has tried something like this before, but I haven’t been able to find any trace of it on the web, so I don’t really know how it’s going to come out. I expect that I will have to tweak the color balance somewhat, but in the end I hope to have a very bright light with all visible frequencies reasonably well represented.