Power struggle

Originally published 2004 in Atomic: Maximum Power Computing

Last modified 01-Apr-2008.

Let's talk about the endlessly fascinating, entertaining and, let's admit it, somewhat arousing topic of power filtering.

Most computer users are vaguely aware that it's a good thing to filter the AC going into your precious multi-gigahertz baby. Surges and spikes can blow it up, slumps and blackouts can shut it down.

Here in Australia, we generally have pretty good AC power. Hook up a fancy RMS voltage and phase meter to the AC supply and you'll see it wandering all over the place from the 230 volt sine wave it's meant to be, but that's cool; things that plug into the mains are made to cope with some degree of input lousiness. PC Power Supply Units are, generally, more tolerant of weird input than most appliances.

But when the guy next door starts playing with his new million amp welder, or your elderly fridge and freezer both click their squeaky old compressors on at the same time, or lightning strikes a suburb away, it's quite possible for big nasty spikes to appear on the mains which can toast PSUs, and possibly other equipment as well. Nearby lightning strikes are particularly nasty, since they can put large currents on phone and cable TV wires, frying computer gear that way.

So you buy a surge/spike protector powerboard and you're fine, right? After all, there are quite a few brands of powerboard that now come with an impressive "Connected Equipment Warranty", offering thousands or even tens of thousands of dollars of coverage if something plugged in through the powerboard gets toasted. How can you lose?

Well, you can lose technically, and you can lose legally.

The technical side is simple enough. No ordinary cheap powerboard ("cheap" definitely includes "$US100") actually provides very good protection from line current gremlins. It may protect you once from a big-ish spike, or several times from smaller ones, but it won't last forever.

Cheap surge filters are all based around components called Metal-Oxide Varistors (MOVs). MOVs pass current only when the voltage across them is above a set value, and they react to overcurrent in microseconds. A circuit breaker or fuse can take tens of milliseconds to trip or blow; that's much too slow for spike suppression.

Unfortunately, MOVs will only work a few times, at best. The more work they have to do, the closer to death they come. A surge/spike powerboard with a toasted MOV is now... just a powerboard.

Better surge/spike boards are meant to tell you when their MOV's died via a little light or even a buzzer, but they commonly, actually, don't. A surge/spike filter that's been in use for some years and still reports its MOV as perfectly healthy is, probably, lying.

Better surge suppressors use two other kinds of spike-sinking component. Gas arrestor tubes are much tougher than MOVs, but respond too slowly to be useful for many applications, including computer protection. But they're present in better power filters because they can handle the load of a really big surge, after some other component has (possibly) bravely given its life to intercept the first several microseconds of overvoltage.

And then there are Silicon Avalanche Diodes (SADs), which are between gas tubes and MOVs in toughness, but as fast as a MOV. They're more expensive, though, which is why cheap surge/spike filters never contain them.

The best surge/spike suppressors gas arrestor tubes and SADs, and possibly MOVs as well. Proper "power conditioners" go on to include a great big heavy iron transformer, whose 50Hz resonance (60Hz, in countries that use that mains frequency) and enormous inductance blocks surges and spikes quite well too. This makes the power filter large and heavy and easily as expensive as a mid-range Uninterruptible Power Supply (UPS), but if there's an aluminium smelter down the road from you, this is the kind of filter you want.

In case you're wondering what one of these "ferroresonant" power conditioners looks like, it looks like this:

I picked up this elderly Sola unit cheap on eBay. The delivered price was slightly more than a tenth of what younger versions of the same thing sell for second hand.

It's got an Institutional Greenish-Beige paint job and a magnificently retro label...

...which suggests that this thing may have fed a VAX or something at some point. But, essentially, this relic is very much the same as a modern ferroresonant conditioner.

If you think that probably means that not a lot of innovation has happened in the area of ferroresonant power conditioner design in the last 30 years, you're right. Inside, there's the gigantic transformer (which accounts for most of the thing's roughly 28 kilogram weight), a bit of wiring, and a couple of hefty capacitors.

That warning label isn't kidding, but the drain resistors across the caps are still working perfectly, so moments after the conditioner's turned off, there's no charge left in the caps.

Eagle-eyed readers will also be able to see, in the full sized version of the above image, that these are "soggy foil" capacitors. Against all reason, that's a real kind of capacitor, and these ones still seem to be OK. The whole conditioner, in fact, still works fine, as far as I can tell; in that first picture of it, it's running a thousand watts of photo lights.

It's possible that this conditioner doesn't have the greatest possible output waveform - old ferroresonant regulators deviated a pretty long way from a sine wave - and I haven't hooked a 'scope up to it to see. But when I plugged a desk fan into the conditioner, the fan motor only hummed slightly louder than it did from ordinary wall power, which is a good sign. And many non-motor devices don't care about power waveform, anyway; switchmode PSUs like those used by all PCs certainly don't.

You don't necessarily want to have a power conditioner sitting next to your desk, though, because they themselves hum quite loudly. The strong oscillating electric fields that make a transformer work also make it vibrate, and you can hear the timbre of the sound (and, in this case, probably also the output waveform...) change as the load on the transformer changes. Modern power conditioners have UPS-style warning beepers to tell you if you're overloading them. Old ones like this expect you to keep track of the load, and failing that, just know what an overload sounds like.

Ferroresonant conditioners also make a distinctive sound when ~~fired at you~~ turned on, as the simple tank circuit "fills up" with energy. Here that noise is. It's vaguely reminiscent of an engine starting, or twanging a ruler on the edge of a desk. It's not the sort of sound you expect a solid state device to make.

Anyway, if your local power is crummy enough, or your equipment is valuable enough, that you can justify buying a power conditioner, then rather than faff around buying tons of flaky powerboard filters for all of your electrical outlets, you can get an electrician to install one giant power conditioner at your fusebox. Now, the whole house should be bulletproof.

You can even get whole-house UPSes, but that's almost certainly overkill. UPSes with moderate filtering capacity just for your computer equipment will do for most users.

So that's the technical problem with surge/spike filters. The legal problem has to do with those increasingly popular Connected Equipment Warranties (CEWs).

You can read a sample Belkin Connected Equipment Warranty in PDF format here, or a Panamax one (also PDF) here. I'm afraid they go on a bit, so you might like to skip actually reading them.

I took the time to plough through them, so I know that first up, you'd better still have the receipt for your surge protector. The warranty probably only covers the original purchaser.

And if you can find a cheap power filter whose CEW covers new-for-old replacement of damaged equipment, I'd like to see it. Normally, they give you blue book used value for your gear. Blue book value for a tweaky 3.4GHz P4 box with a GeForce 6800 may be the same as for a Compaq with the same basic specs. CEWs also often leave the guarantor the option of having the equipment repaired, rather than replaced. And of making you pay to have your exploded computer shipped to the repairer.

If the guarantors really want to weasel out of the deal - which they won't necessarily, but we're talking about the insurance industry now, so anything's possible - they can insist on proof that every wire going into the toasted equipment was protected with their gear. Were you using an extension cord? Were you overloading the surge protector (did you plug your laser printer into it?). Hey, maybe the spike was just bigger than the filter was meant to deal with!

Did the filter protect you from a spike last week, burn out, and so not protect you from a spike this week? If so, then no payout.

Bit of a Catch 22 there, really. If the filter's not burned out then they can say you didn't have it plugged in at all, and if it is then they can say that it was dead already, and you should have stuck your head behind your desk every day to see whether the little orange light had gone out.

The take-home message from all of this is that quality power filtering isn't cheap. Fortunately, though, most people here in Australia don't need quality power filtering. To Aussies, I therefore say: Go ahead and buy dirt cheap no-CEW filter-boards, and be merry.