Fourmilab's Coruscating, Actinic,
(partially) Nuclear-Powered Christmas Lights

by John Walker

December 10th, 2004

Interior view of Single String Test 'Tis the season to put up the Christmas lights and, as usual, confront the problem of dead series-strings of incandescent bulbs. Now, if you're an engineer like me, you've probably developed a pretty strong sense over the years about how the world works. This hindbrain reality detector will probably rule out ever finding a string of 12 bulbs in series in which three or four have burned out filaments, but in fact this is common when repairing such lights. What's going on? I'm not entirely sure—the most plausible explanation seems to be that when a filament opens, you can often get the string working again by tapping it against a window (or other hard object). This will sometime jog an open filament back into contact, so the string lights again until another filament opens, which can sometimes be remedied in the same way. When the thing resolutely refuses to light and you take it to the bench to repair, you find multiple dead bulbs, many with obvious open filaments when viewed through a 10× magnifier. On the other hand, the explanation may be tiny filament snipping gremlins from the fourth dimension who materialise inside the bulbs when the lights are stored in the attic waiting for next year. But this engineer is not going there.

Regardless of the cause, it is intensely irritating when one of these strings of lights fails. Each one contains 12 “grain of wheat” bulbs, driven with 24 Volts AC Exterior view of Single String Test from a transformer, so the voltage drop across each bulb is about 2V. These “light curtains” can be installed inside or outside, so the bulbs in the strings have heat shrinkable tubing along the base which makes it difficult to test for blown bulbs. Generally, it takes between 30 minutes and an hour to repair a dead string. First you replace all the bulbs with visibly open filaments, but usually there are one or two dead bulbs which look OK under the magnifier, and can only be found to be bad by stripping the wire on either side and testing with an ohmmeter.

Now, you might say, “Why be so bloody cheap? Just pitch a string when it dies and install a new one!” Well, first of all, I am bloody cheap, and I really don't like the idea of throwing away nine or ten perfectly good bulbs, along with the connector, wire, etc. because a few bulbs have failed. But in addition, replacement bulb strings aren't reliably available—it seems like they change the connectors every few years, so you still have to solder the connector from the old string to the new one. And of course what they really want you to do is buy a whole new set with the transformer, etc. and that is right out as far as I'm concerned.

Still, I do like the lights in the window, and neighbours remark upon how welcoming illuminated Fourmilab is when driving into the village on dark, foggy nights in the Christmas season. What to do? Prototype in Fourmilab Electronics Lab Well, how about that time-proven, all-purpose solution: smashing the problem to death with technology! In this case, replacing the incandescent bulbs with light emitting diodes (LEDs), making compatible strings which are completely interchangeable with the existing ones. LEDs, properly treated, almost never fail, and they also emit far more light per milliampere of current consumed than incandescents; they last forever, run cool, and use less power—win, win, win. Of course, they do cost more—quite a bit more, in fact, but that's a capital expense which can be amortised over the decades they'll last and is, in any case, easily recouped if you consider the time and irritation you'll save not having to repair or replace incandescents.

Based on extensive research (flipping through the Distrelec catalogue), I chose the Kingbright W7114PWC/H white (actually, it has a bluish cast) high-intensity LED. This device is typically operated with a forward voltage of 3.7V, which I derated to 3.6V since I didn't need the full intensity (which is almost blinding head-on). To power each string, I fabricated a head-end consisting of a full-wave bridge rectifier and a 100µF smoothing capacitor. Driven with 24 VAC (RMS), this produces about 37.8V DC with Fourmilab's typical line voltage of around 230V. Dividing by the 3.6V forward voltage of the LEDs, we find that 10.5 LEDs in series will yield the desired current. Detail: rectifier, capacitor, and first LED What, saw an LED in half? No problem—just connect 10 LEDs in series with two in parallel at the end of the string! Now, veteran electrical engineers will recoil in horror at the thought of connecting forward-biased diodes in parallel without a load-balancing resistor but, in fact, it works just fine. As long as you use diodes with the same part number from the same manufacturer, you can see a difference at low current, but when they approach operating load they behave identically.

Here, I've made a “space frame” circuit, soldering the leads from the string connector directly to the rectifier, and the rest of the string to the other side. The leads on these DIPs don't survive many flexures, so it may be wiser to fabricate the circuit on perf-board to keep the IC from losing its legs. The pictures above show the first replacement LED string installed with legacy incandescents; as the incandescents burn out, I'll transition over to all LEDs.

One difference between LEDs and incandescents is that incandescent lamps have an isotropic (all-aspect) light emission, while most LED packages strongly “beam” the light along one axis. In fact, the LEDs are so bright that this isn't a problem—even the off-axis emission from the LED is as bright as the incandescents, as you can see in the exterior shot above, taken through thick freezing ground fog. If the LEDs are mounted in more or less random orientation, their principal beams will be be distributed through space. This causes the wall of light to appear to “sparkle” as a moving observer passes through the beams of the various LEDs, which is a nice effect. My next experiment will be “frosting” the LEDs to diffuse the light by dipping them in a mixture of Epsom salt (MgSO4) and beer. Re-visit this page in a couple of days to see how that worked!

Switzerland
Electricity by Source
Hydro 59.5%
Nuclear 37.1%
Fossil fuel 1.3%
Other 2%

Partially Nuclear Powered?

Fourmilab's Christmas lights, like everything electrical in Switzerland, consume almost no fossil fuel. Electricity generation in Switzerland is about 60% from hydroelectric power and 37% nuclear. So more than every third electron that passes through Fourmilab's Christmas lights is kicked along its way by nuclear energy—green glowing cool!


This document is in the public domain. Permission to use, copy, modify, and distribute this document for any purpose and without fee is hereby granted, without any conditions or restrictions. This document is provided “as is” without express or implied warranty.


  Valid XHTML 1.0
by John Walker
December 10th, 2004