On April 22nd John Moore followed up his April 18th presentation (see Meetiing Schedule item for that date) with an actual demonstration of networking. It was an illuminating (and successful) demo, but the notes from his previous presentation cover all that needs be said.

Here's a staggering collection of links to Linux sources of every description: http://loll.sourceforge.net/linux/links/ At the end of the meeting, with too little time to elaborate, Will Wilgus spoke of the importance of quality power supplies for today's hardware and invited members to join him in developing a testing facility. He has provided the following details:

PROBLEM:
Digital circuitry works (and it's usually fantastically reliable else we wouldn't even be capable of putting millions of little electrical switches on a piece of sand and then grousing when it all doesn't work perfectly all the time) only because comparisons can be reliably made between voltages. The reference voltages against which comparisons are made (in most cases), and the power necessary to run the switches (in all cases), come from the 'power supply'. Both Intel and AMD have become insistently picky about the requirements of computer power supplies for systems using their chips (particularly the Pentium 4 and the Athlons). They've always been tight on voltage regulation, but it's gotten considerably tighter in recent months. They now insist power supplies meet much stricter specifications than the ATX standard requires.

BACKGROUND:
If your supply is batteries, basic electrochemical reality keeps the voltages here rather than there, and keeps the noise level down, at least until they go flat. If it takes power from the wall, big conversions have to be made. Power to all those bitsy switches in the chips was intended by their designers to be DC (the wall doesn't provide that) and very clean too (no noise (high frequency, low frequency, whatever) beyond very tiny amounts) and very constant (no voltage changes beyond very tight limits) regardless of how many disk drives, add-in cards, USB devices, etc etc etc are attached. This means that a wall powered power supply has to do a great deal. It has to
1) convert AC to DC,
2) block all the noise and trash on the AC line,
3) not generate any noise and trash itself,
4) take widely varying loads in stride without producing any noise, allowing any noise through, nor changing the output voltages,
5) keep those output voltages almost exactly constant,
6) fit into the available space in a PC chassis, and
7) do it all, reliably for a long time, for a reasonable price.

It's not easy to design such a beast. Even if you have such a design, manufacturing every unit to properly execute that design is also not so easy. Nevertheless, some companies do it well. PC Power and Cooling has a sterling reputation -- see their Web site. Regrettably, few computer makers get their power supplies from PCP&C (or some equivalently high quality maker), and few of us want to spend more than necessary.

COMMERCIAL CHICANERY:
That's why there are very large numbers of power supplies on the market from mysterious makers in exotic places with low labor costs. Power supplies are pretty much invisible in a marketing sense, and computer vendors find them a tempting place to save. Many of them don't even tell their users what specifications they were intended to meet, much less whether they meet the ATX standard power supply specification. I've never seen one, nor heard of one, that even claimed to meet either Intel or AMD's recent power supply requirements.

THE RESULT:
If your supply isn't up to the necessary snuff, what will happen? Well, at worst, your computer won't work -- no boot. Or, if the supply is really bad or fails, your computer might get toasted as wall current gets applied to delicate digital circuits. Or even a fire, burning down the house. At best, you'll have times at which those bitsy switches in the chips can't work quite right (due to the bad power they're getting) and things will have to done again since internal checksums won't come out right. If, that is, there are error detection / correction mechanisms watching the circuit that's having trouble. Your disk drive may seem to have slowed down as data is reread over and over again trying for an error free transfer. Avoid this and you ought to get a considerable speed up. In the middle ground, bad data might get written to disk. If this annoys the IRS, you are likely to be unhappy. If the bad data is in some binary executable file, your system might crash, be unable to handle some file type, stop booting, ...

THE SOLUTION:
So how can you avoid all these not so good outcomes? Test your supply -- whoever made it -- to see what it actually does, and get a good one if you've already got a lemon, or even a marginal citrus. Antec sells one, but it's pretty rudimentary and unsatisfactory. That's the reason for the suggestion I made at the April Linux group meeting to the Hams in the group. Design a PC power supply tester and offer to test folks' power supplies for compliance with the ATX (vastly the most common type of supply) and even with the more stringent Intel/AMD specifications. These are high amperages, even if not so high voltages, and so it's not quite like testing an AA battery at the store by pushing on the 'button' and watching the cool bar display tell you 'how good the battery is'. Hence the suggestion to the Hams amongst us, they being more conversant with both safe handling of such, but also with such exotica as RF noise and transient decay.

HELP US, OH HAMS:

Contact me at htbcomputer@yahoo.com if you're willing to participate.