Slashdot Mirror
Fluorescent Lights Magically Activates iMac?
bats asks: "In my computer room at home, I have several machines -- and a fluorescent desk lamp. Among my various boxen is an iMac DV (slot loading) circa 1999. Its configured to go into power saving mode, but respond to wake-on-lan packets. The weird thing is this: If I flip on the fluorescent desk lamp, the sleeping iMac will suddenly wake up! This happens with 100% consistency. The desk lamp is plugged into a power strip and into the wall. The iMac is plugged into a UPS and then into the wall. The network switch for the room is near the desk lamp (1-2 feet) but the iMac is some distance away (8-10 feet). My question is: WTF?! How the heck does the iMac know when the light comes on? It seems like it must be some power spike in the AC or noise on the network interface. However, the power strip and the UPS should block an AC spike and the chance of electrical noise in the cat-5 looking like a wake-on-lan packet seem more than miniscule. So again I ask you, dear AskSlashdot reader, WTF?! I await conspiracy theories, pseudo-science, wild rantings, and hopefully, the right answer."
My guess would be cat-5 noise. I've had problems with my computer turning on erratically whenever I plugged/unplugged a port, or jiggled a loose connection. For whatever reason, the noise on the network woke it up. I finally just turned off WOL, since I wasn't really using it for anything... just playing.
Try putting more distance between the hub/switch and the light, and see if it still causes it to happen. I wouldn't imagine the light could cause much interference over more than a few feet or so. You might also try hooking up another machine, and monitoring all the electrical activity on the network. Myabe something will poke up.
Experiments must be reproducible; they should all fail in the same way.
> The wake-on-lan must just detect ANY activity on the ethernet, not a valid packet.
IBM has a White Paper, Wake up to Wake-on-Lan which describes the specific format of a Wake-Up packet (6 bytes of F (XX'FF') followed by a 48-bit target address repeated at least 8 times). Even if the target address was a broadcast address, I don't see how this could be reliably generated by the starting of a fluorescent lamp. Plus, your average LAN has a whole bunch of traffic going on all the time anyway.
I have an iMac DV (graphite), and I DON'T have it plugged into a network, and it's NOT configured for Wake-on-Lan, nor is it configured to wake on phone rings, and I don't have any infrared devices that I know about, and I have an optical mouse, so it won't wake if the mouse gets moved, and it still wakes up when I vaccum my living room or turn on my slide projector.
The answer to your dillema is pretty straightforward: when you turn the light on, your *optical* mouse is noticing a change in the pattern on the desk, and interpreting this as the mouse being moved to wake the Mac.
The way to test this hypothisis 100%? Unplug the optical mouse, let the bugger go to sleep, and turn on the light.
Easy, neh?
P.S. It does that in a few of my setups (I work in the dark, with a Kensington Flylight over the keyboard usually) -- but the Apple optical mouse seems much more succeptible than my Logitech cordless optical. Chances are this is because the Apple mouse is mostly transparent.
Exactly true. Shielded Cat-5 is more expensive. It is more difficult to find a supplier. And, in this case, shielding would almost certainly not help. The noise is definitely not coming from some electromagnetic connection with one wire inside one of the twisted pairs.
The problem is probably a ground loop; probably the noise is being conducted along the ground wires. More directly grounding the components to each other may help.
Also, it seems to me that Wake-on-LAN is a technology that is usually not implemented well. I've seen cases where it was a selling point, but caused problems.
One thing that helps stop Cat-5 noise is grounding the unused four wires. There are 8 wires in Cat-5, and only four are normally used. (It is possible to buy adapters that allow one Cat-5 to carry Ethernet for two computers, but that is not normally done.) There is electrical capacitance between one Cat-5 pair and the others. When you ground the unused two pairs, you are grounding most of the common-mode noise that would otherwise be experienced on the two pairs that are used. In a low-noise environment like the one mentioned, grounding the unused wires would be the equivalent of having shielded Cat-5. This is something you can try without cost if the Ethernet signal travels through an accessible wall or other connector.
However, as I said, the problem sounds to me like noise conducted along the ground wires, a "ground loop".
The Cubes had a similar problem to this, for some reason the power switch circuitry was susceptible to RF noise generated by flourecent lights and other noisy electrical devices. Other G3 iMacs have also demonstrated their weirdness. It isn't static on the electrical wiring itself but RF noise being generated by the RF generated by the flourecent light. Call up Apple and ask them in there is a tech article about the power switch RF interference problem. It's a probablem that cropped up in a couple of different Mac models though not necessarily all of those models.
I'm a loner Dottie, a Rebel.