On Power Switches (Part 1)

I've long wanted to write about power switches, but I'm pretty sure it's going to come out with more than its fair share of "get off my lawn, you kids" old-manishness. Forewarned is forearmed.

Let's go back a few decades. In the 1980s, and back beyond that into the mists of time when chips were wood and bits were for horses, the power switch used to look like this:

It's big and chunky and it makes a satisfying "clunk" when you switch it. Oh yeah, and it's on the back of the machine:

That's the same specimen as above -- on the back of a very stately Macintosh Classic, circa 1990. Does anyone else remember groping blindly around the backsides [ahem] of printers, computers, and anything else big and electronic, hoping that you'd find the power switch before you found something sharp or electrified or both? I stuck my fingers in more than a few sockets this way.

There's really no reason for putting the switch on the back of the machine other than having it close to the power jack. If you were to crack open the case and follow wires, you'd see that both the power switch and the power jack were built into the power supply, a metal-shielded box inside the computer that neatly contains almost everything high-voltage. Its job is to take 120 volt household current and drop it to the usual voltages that computers use -- 12 volts, 5 volts, and usually a few more. Oh, and to turn on and off when you throw the switch built into it. Here's what a power supply usually looks like:

One refinement on this was (gasp) to move the switch to the front of the box. Here's how the front of most early 1990s-era computers looked:

(Well, okay, that's actually a pretty strange example. ANO Office Automation must've been looking for a way to stand out from the crowd and they decided the best thing to do was to build the computer upside down. But anyway.)

That big square button on the front is the power switch. We've gone from a toggle switch to a push button, but it's still working the same way. It still switches 120 volts right from the wall before anything else can happen to it. Behind the plastic faceplate, here's how it looks:

See that ring of fuzzy goo around the switch itself? That's finger grease, lint and hair from the decades of sweaty computer types who turned this machine on and off.

Undoing those two screws and pulling the switch out of the case reveals this:

The switch and the wires leading up to it are shrink-wrapped in insulating plastic because this is the only place inside the case where you'll find 120 volts AC. The next highest voltage inside the case only 12 volts -- and that's DC, meaning that you could touch the pins without feeling a shock. If you touch an exposed 120 volt AC line, you'll definitely know about it.

Inside the machine, that black cable goes into the power supply (excuse the blurry picture and festooned wiring):

Remember that Macintosh in the first picture? Well, I fully blame Apple for what came next. I remember when the first set of Apples came out with the power switch mounted on the keyboard -- a soft-touch little button, and when you pushed it, that famous Apple "Bing!" noise invited you into a new generation of computing. None of this old-fashioned knife-switch stuff. I was wowed by this innovation just like everyone else. And of course, the other manufacturers had to follow suit, so IBM-compatibles started using the ATX standard in 1995 (according to an extremely reputable source), which introduced a "soft" power switch.

To be fair, this was a big improvement in usability. Sure, there was the esthetic bonus of the clicky little power button, but there were serious benefits too -- the computer could also shut itself off, for example. Previously you had to tell the software to shut down, then when it decided it was done wrapping everything up, you had to throw the switch yourself. The old power switches were no different from the computer's perspective than hitting the power cable with an axe à la an offended Pete Seeger; if you shut the power off in the middle of something, you could do real damage.


Here's how the new switch looked:

If that switch looks too wimpy to do any real switching, it's just because it *is* too wimpy to do any real switching. This is the downside to the change: it takes brains to have the computer turn itself on and off, and those brains must receive power even when the computer is turned off. This switch doesn't control any more power than a button on the keyboard does. It just tells the real switching apparatus that the user wants it to do something.

This new design requires "vampire power", a.k.a. "standby power" -- power drawn by a device even when it's turned off. In the case of the ATX power supply standard, there is a 5 volt supply that's always given to the motherboard, whether or not the computer is turned on.

This is the power jack that goes from an ATX power supply into the motherboard; near the top you can see the purple "+5V Standby" line that powers enough of the motherboard to handle power-on controls.

Power supplies typically aren't very efficient. There are many different designs, from the ages-old transformer-and-rectifier to the modern switching supply. The ATX standard currently recommends an 80% efficient design, with a 70% efficiency minimum required for compliance; previous requirements were lower.

So in summary, almost any computer that was made in the last 10 years is always drawing power, even when it's turned off. And not just computers, but more broadly anything with a clicky little power switch, or the ability to turn itself on or off (e.g. by remote control or after a period of inactivity), or a clock. And everything with a wall wart for power -- since any power switching happens after the power conversion, there is always some power lost in the conversion, regardless of whether the device is active or not.

Beware the little clicky button on this printer.

I'm going to stop there, but I'll pick up this theme later with some concrete measurements of common devices and some broader stats about power use across countries and continents. This is how the design decision to go to a "soft" power switch has big implications for power infrastructure and the amount of power taken by these devices is surprising.

More soon! (Part 2 is now online here.)