After moving into this house I purchased a large order of X10 switches and set about installing them. In the course of 2 weeks I had installed about a dozen wall switches and other modules. I was aware on some peripheral level that this carried with it a potential for problems. If I had started with just a few switches and added slowly after I was sure that what was there was working it would have been very easy to debug knowing that the problem was probably with anything new. But I had to wire it all up at once.

I installed the switches and did not immediately set about actually doing anything with them. It was another 2 weeks or so before I even connected the X10 interface to the computer and even longer before I started automating anything. During this period there were no strange events, none of the lights turned themselves on or off without my permission. Everything was looking good for the future.

This period of calm did not last. Slowly I started noticing that certain lights in the house were switching themselves on. The programming logic for the lights was still very simple and there was nothing in the log of the Mac running the show at all. Neither the CM11a interface I originally had connected nor the LynxPLC that I connected later saw any spurious signals at all. The frequency of the strangeness and the number of lights affected grew larger until probably half of the wall switches I had installed were participating in turning themselves on at random moments.

I replaced the switches that were most affected. I set things to different house codes, I physically disconnected the switches one by one trying to isolate one with a problem. I switched off the circuit breaker to the coupler/amplifier. None of this made any difference at all in the problem.

The only pattern that I was able to discern was that it did not happen after I had gone to bed. The lights in the bedroom never switched themselves on in the middle of the night. But they did during the day once while my wife was napping. This was unfortunate but not nearly as bad as it might have been had they repeatedly done so during the night.

Recently Michael Ferguson, the author of XTension was in town and dropped by with his power line analyzer. Going through the house we found 1 phase of the house to be very quiet, with a very reasonable amount of noise on it. But the other phase carried 3 to 4 VOLTS of noise at any given moment. This is as much power as the CM11 puts out for a good signal. It was frankly amazing that anything was working X10 wise in the house at all!

Walking through the house testing every outlet and unplugging everything one device at a time yielded nothing.

Finally standing in the living room looking up at the ceiling in exasperation I noticed one of the very high up recessed fixtures that contained a CF bulb was burned out. This had happened several weeks ago and I was just using the other regular bulb for lighting until I could purchase a ladder long enough to reach it to replace it. They were both in too tight to get out with the suction cup extendable wand thingy.

It dawned on me that even if the ballast couldn't ignite the tube, the circuitry was still active and might be unhappy with the arc not being established, so I switched them off. and the noise on that phase disappeared! These lights were X10 controlled by a florescent friendly SwitchLink 2385I which was working just fine and not one of the switches that ever turned themselves on. We turned the lights on and the noise returned, we turned them off and the noise went away.

HA HA we thought, it's the burned out CF bulb thats doing it! But Michael, calling on his debugging ability honed to a fine edge by NASA, decided that further testing was necessary. I removed the Switchlink from the wall box and wired the lights straight through so that there was no switch involved and turned the circuit breaker back on. The light came on (except for the burned out CF of course) but the noise did not reappear So it had to be this brand new $40 switch that was faulty! My reaction, of course, was that wasn't possible, it was a brand new switch! Michael wishes you could have seen the look on my face as I asserted that it couldn't be the brand new, overpriced switch.

But that was another assumption that needed testing. We took the offending switch up to my test bed and wired it up to control an identical branded CF bulb (which, by the way, was a "consumer electric" 60 watt equivalent reflector flood CF bulb purchased from Home Depot) and the switch worked fine. There was no noise created by the switch when controlling a good CF bulb at all.

This was even stranger, which was it the bulb or the switch or were we missing something else? We replaced the switch into it's original circuit and powered it back up and the noise returned again.

The Offending Switch



The Offending Bulb

CONCLUSION: The horrific, X10 stopping noise was created by the combination of a burned out CF bulb and an otherwise perfectly good SwitchLink switch. I cannot test any further until I purchase a taller ladder and can get down the CF bulb and test it together with that switch and another known good one on my test bed which I will do in the very near future as it is a little dark in my living room now.

Neither of the 2 elements caused any problem on their own, but conspired together to cause no end of pain. This should serve as a reminder that any preconceived ideas are a roadblock to debugging. The fact that a bulb is burned out does not mean that it's ballast is not energized and working. And the fact that the switch is brand new doesn't mean it's not the thing thats causing the problem.

The switch above that had problems with CF bulbs has been happily controlling my garage lights since I last wrote here it has been working fine, and since they are magnetic type ballasts it's not having any difficulty at all. Then suddenly a week or so ago I noticed that the lights in the garage were taking longer to come on. A closer inspection shows that the switch that has been on/off only now accepts dim commands and does soft starts. This is probably not good for my ballasts but they dont really seem to mind, and will actually dim a little bit. So some cosmic ray hit the chip inside this switch and turned it from on/off only to a regular dimmable switch. Wish I could turn one the other direction as the dimmable ones are generally cheaper. Obviously they use the same firmware in the PIC inside and just have a flag or a bit or an external connection or something that sets them to one mode or the other. And whatever it was has failed or flipped in this. Here's a little animation of the switch above dimming the non-dimmable lights in my garage...