Dan's Data letters #34Publication date: 4-Mar-2003.
Last modified 03-Dec-2011.
The thrill of the case
I have three computers in my house, all of which work fine. The third computer is in the basement. Whenever I touch an unpainted surface on that computer's case I get an electrical shock. I don’t mean a little static shock, I mean, enough to make you jerk your hand back, and it happens each time. I also get the same result from touching the screen on the monitor. This happens even if I unplug the monitor. I still get the shock just from it being attached to the computer.
My first instinct was that there must be something wrong with the computer. After some troubleshooting, nothing seemed to be wrong. So, I unhooked the system and carried it upstairs and plugged it into the outlet where my other computers are. Up there, it doesn't shock me at all. So, I bring one of the computers from upstairs down into the basement, and it starts with the same shocking behaviour.
I am concerned that I might have some bad wiring in my basement and maybe even a fire hazard. Do you have any suggestions as to what could be causing this and how to fix it?
Since the computer behaves itself when it's plugged in upstairs, the problem is pretty much certain to be a lousy earth in the basement.
Earth connections at household power sockets are never perfect grounds, unless you've completely powered down (and, possibly, unplugged) everything in the house. If you connect an AC voltmeter between a power socket's earth pin receptacle and, say, a long metal stake driven into damp soil outside, you'll measure a small potential difference. That's because the earth wires run in the same conduits as the active and neutral wires, and current-carrying conductors run past ground planes in earthed appliances as well; this inductively and capacitively couples the earth to the active, inducing a small current in the earth. Various devices also leak a little current to earth through things like electro-magnetic interference filter capacitors. The aggregate effect of this normal leakage can, if you're unlucky, be enough that your house's Earth Leakage Circuit Breaker (ELCB, "safety switch"), if you have one, will keep nuisance-tripping, but it shouldn't be at all dangerous.
If there's rather stronger coupling of active to earth, though - generally because of failed insulation in a conduit, or poor wiring somewhere, or a defective appliance - then things that're meant to be earthed can, indeed, give you what I like to call "an exciting tingle", and what a friend of mine who once used a switch I thus described without jumping in the air before he flicked it called "AN EXCITING #&@^ING TINGLE?!"
You should get an electrician in to look at this problem. It may, indeed, be a fire risk, or it may deteriorate further and end up giving you a really dangerous belt.
I was recently reading your letters and I'm confused about where you advised Tim to use twisted pair cable to connect audio equipment over a long distance because it's "naturally noise-resistant".
My understanding was that when Cat5 cable is used with network adapters, a reference signal was sent through one wire, and data through the other, and therefore noise would affect both equally so the receiving device could reverse any noise in the data.
But the audio device you suggest using the cable with wouldn't know or care about this. How do twisted pair cables stay noise-free?
You're almost describing a balanced connection, but not quite. In balanced interconnects, you have two signal conductors, and one ground. The two signal conductors each carry the same signal, but they have opposite polarity - when the waveform's rising on one of them, it's falling on the other. So one conductor carries a signal that's exactly the same as whatever you want to feed down the cable, and the other one carries that same signal, but inverted.
The idea here is that noise will affect each conductor the same, so that when you un-reverse the inverted conductor's signal and combine it with the un-inverted conductor's signal at the other end of the line, the noise will cancel itself out.
In any case, that's not how Ethernet network cables work. Standard 10/100BaseT RJ45-connector cables use only four of the eight terminals on the connector, and they use them for two simple two-conductor loops, one for transmit and one for receive. No fancy balanced stuff is going on.
Twisting two conductors around each other, as opposed to running them parallel to each other, turns them from one big (if narrow) loop into lots and lots of little loops, each successive one being reversed compared to the one before it. The multiple reversed loops reduce common mode noise (potential differences between the conductors and ground), because the polarity of noise created by external interference is reversed in each loop, and thus it tends to cancel itself out. Twisted pair cable certainly isn't perfect, but it's a lot less interference-prone than the same cable without the twists.
I do have one favour to ask though. I am thinking of splurging on an iPod MP3 player and saw this "battery charger" that some guy made using a battery pack of 8 AA cells. While it seems simple and well within the scope of my electronics "skills" I began thinking it was a little too simple, mainly because it doesn't have any form of block to prevent the iPod's battery inadvertently charging the battery pack on the charger. I assume something like this would be necessary.
The guy who made this said that it will charge an iPod to around 80% capacity with 8 1.5V AA cells wired in series. Looking at the iPod specifications (PDF file here) the iPod gets a 12V input at up to 1A from its power adaptor. Would using two 9V batteries wired in series give it more charge than the 8 AAs, or would 18V be too much? I couldn't find anything on the power output of 9V batteries.
I haven't been able to find any further info on recharging a battery with another battery - most sites I've come across cover recharging from mains power or solar cells but that's about it. Trying to figure out the diode (if any) is required to prevent backwards charging from occurring is a bit beyond my knowledge so I was wondering if you might be able to shed some light on it for me?
Reverse charging is unlikely to be a problem - there's charger hardware inside the iPod between the FireWire socket and the battery. But if you manage to connect your battery pack backwards, you may blow something up. The iPod is unlikely to expect reverse polarity input, since it's not possible to reverse the polarity of a FireWire connector's power wires, without stripping some wires. Ordinary AC adapter barrel plugs are often made to be easily reversed, so devices that run from ordinary AC adapters usually have reverse polarity protection.
9V batteries wouldn't be a good choice. They'd sag well below the 18 volts you'd expect, into a significant load like an iPod, but they might still deliver more volts than the iPod wants to deal with. 9V batteries also have miserable capacity, especially under heavy load. The eight AA alkalines will also sag a bit under load.
Personally, I'd consider using a pack of ten NiMH or NiCd cells, which have 1.2 volt per cell output and won't sag significantly under this load. Ten 1.7Ah NiMH cells will cost you a bit, of course, and you'll have to charge them, but you can use them over and over again, and one charge for the NiMHs should give you a bit more than one charge to the iPod.
Alternatively, C or D size alkalines won't be pummeled as much by the iPod's charge load, and will cost you less per charge. They're bulkier, of course, but I wouldn't be surprised if you got more than ten near-full iPod charges from a set of eight D cells, or from a couple of six volt lantern batteries in series, for that matter. You could use a rechargeable 12 volt sealed lead acid "gel cell", as well, if lantern batteries just aren't heavy enough for you.
With or without rinse aid?
After reading a letter asking how to clean the keyboard (and your reply suggesting to pop the caps off and use a vacuum) I got around to thinking. I have seen (in newsgroups, surprise surprise) people writing that it is fine to pop your keyboard in the dishwasher for a clean as long as you let it dry properly afterwards.
Well? Whaddyareckon? It sounds harebrained enough to be true, but I'll let you try it first.
Cleaning plastic parts in the dishwasher would definitely work - all of the keytops in a plastic mesh orange bag, or something, and the halves of the main casing just sitting with them in the top basket. Just a little detergent, no super-pre-heat, no dry-out cycle at the end (it'll warp the plastic), no worries. You can wash keytops in the washing machine, in similar fashion; apparently tying them up in an old sock works well.
Washing a whole keyboard, however, I'm less sure about. It depends on the design of the 'board, but most keyboards contain plenty of crevices where water could linger for a long time. You'd want to leave the 'board to dry in the sun for rather a while, I think. Yes, plenty of people on newsgroups report that washing a whole 'board that way worked for them, but you can find people on newsgroups who believe all sorts of things.
If you've got a keyboard that's so filthy that it either has to be cleaned or thrown away, and you've no desire to have at it with cotton swabs and monitor wipes and so on, then by all means fling the whole thing in the dishwasher. What the heck. I wouldn't try it on a keyboard that was worth something, though.
Hunting an ISA snipe?
Like so many people I still have an old 486DX2-66 lingering around. Looking at SmoothWall, I thought "Why not make good use of it?"
However my ADSL modem is a USB type. Is there a way to get USB on an old 486? I can't seem to find any ISA USB cards, only PCI. Do they even exist?
Is there a way around this, or is it impossible to connect a USB device to a 486 ?
ISA USB cards used to exist, and I think at least some of them had x86 drivers. These days, though, they're extinct; someone somewhere might have them for sale, but I don't know who.
I think you'd do better to buy a DSL adapter with an Ethernet port; then, you could connect it to the 486, or any other device you like, via an ISA network card.
Giblets, and heat
I'm considering buying a new set of innards for my computer. I do like playing the latest games, but I'm not hysterical about frame rates, so I decided against spending a larger fortune on a 3GHz system with GeForce FX Ultra, et cetera.
Instead, I decided to get a 2400MHz/Athlon XP 2400+ type system, probably AMD, although the CPUs seems to be pretty well matched around this price spot. I've heard rumors that the XP 2400+ is usually somewhat overclockable as well. I've noticed the 2400 seems to run with an FSB of 266 which is a bit low, but...
The idea is then to get an nForce2 board and utilize the dual-channel DDR. Because I also use this machine for software development, VMware and other slightly hefty things, I really want at least 768Mb of RAM, which made me decide to spring for 1Gb. I read somewhere that for dual-channel DDR, you should match the RAM speed to the FSB of the CPU, which I don't understand. Anyway, if I want to overclock the thing, do I increase the FSB of the CPU? How far can I realistically expect to push it? Does it make sense to buy 333MHz RAM or even 400? The reason for considering high speed RAM is mostly that it'd make it easier to yank out the CPU and stuff in a faster one, once they drop in price.
It seems the Radeon cards are very good value for money right now, although it does seem that if I want to play Doom 3, I really should consider going for the 9700/9700 Pro. Lastly, I currently have a 300W PSU. Is this good enough, or should I shoot for something bigger?
Slightly off topic: Recently, I've been using a utility called VCool to idle my AMD 1300MHz CPU. This dropped the almost-idle temperature from 47 to 31 degrees. Is there an easy way to calculate how much power this saves me? I've been wondering about this for a while, since electricity prices here in Denmark aren't that low (about 1.45 Kroner, 21 US cents, per kilowatt hour). There are some advantages to having an AMD computer, though; I haven't really needed to turn on the heat all winter ;p
As you say, the Athlon no longer has the huge price/performance advantage it used to have. It's still somewhat better value than the P4, but the difference isn't very impressive.
You should always have RAM as fast as your CPU's Front Side Bus speed, if only because a RAM bus running slower than the CPU bus will bottleneck the system. RAM running faster than the CPU is pretty much wasted. Doesn't hurt, but has no real impact on performance; some RAM benchmarks will report impressive numbers, but practically no real world tasks will show more than a minuscule improvement.
The nForce2 motherboard you get should allow you to adjust RAM and CPU bus speeds separately, and keep the PCI and AGP speeds at stock while you do it. But even if it doesn't, if you're starting from a CPU bus speed of 266MHz and get 333 or 400MHz-capable RAM, you'll be fine to wind both buses up together when you overclock.
You can't expect a huge amount of overclockability from an Athlon XP 2400+, but it's a 2GHz CPU in the first place; what do you want from life?
2100MHz is almost certain to be achievable, 2200MHz will probably be fine, 2300MHz might be possible, but I wouldn't bet on it. Since 2300 is less than 10% faster than 2100, you're unlikely to even notice the difference, so don't worry about it. Wind it up a bit, just to give you the warm fuzzy feeling that you've done so, and enjoy your fast PC.
There's some sense to getting 400MHz-capable memory, if you're keen to upgrade in the future, and considering the amount of RAM you're going to be buying. PC-3200 isn't currently that much more expensive than PC-2700, after all.
NForce2 boards, used with Thoroughbred-B core Athlon XPs, allow you to adjust the multiplier freely without fiddling with the CPU to unlock it first. Which is nice. But unless you want to do the same tricks as the real nutcase overclockers (reduced multiplier, massively boosted FSB...), you won't see any benefit from 400MHz RAM with the XP2400+.
As far as video card choice goes - yes, the newest hottest game always wants the newest hottest video card. It's the usual trade-off. Bear in mind, however, that the humble GeForce4 MX still isn't a lousy chipset, and MX cards are cheap compared with even mid-range full GeForce4s, let alone top-end Radeons. You'd probably end up spending less if you bought a budget card today, and a Radeon 9700 when Doom 3 comes out - and you'd end up with a spare video card!
If your 300 watt PSU is a decent unit, it'll be fine. If it isn't, it won't. Sorry, but that's as specific as I can be.
As regards your CPU temperature question, no you can't easily tell how much power you're saving, because you don't know how much power the CPU was consuming before. Redline power for your CPU's about 68 watts, but when there's nothing going on it won't be consuming that much. At a guess, I'd say you were saving 10 to 20 watts, but that really is a guess. If you can get your hands on an AC clamp meter, you could clip it on your PC's power cord and see what you can see.