Dan's Data letters #84Publication date: 12 Jan 2004.
Last modified 03-Dec-2011.
I am interested in the history of computers and that brought me to your web sight. I have a program card from a #1 ESS Phone Switch. It was the first programmable telephone switch (arguably the reason transistors were invented). The memory card is a sheet of aluminum with iron dots. The program is generated by magnetizing and demagnetizing the dots with a stylus. Each dot corresponds to a bit in the computer's memory.
Right now I have no way to read a card. I have heard about film that was used for this purpose. In my search, I found your web site. This seems like just the sort of thing you would like to be aware of.
Yes, magnetic field viewing film might do the job, though its sensitivity might be too low; I don't know. It wouldn't cost you much to get a bit of film to experiment with, though.
Following the shedi path
I'm about to convert my humble shed to an office and join the throngs of those who "work" from home. Well, it's a good excuse to sit around unshaven in pyjamas all day.
I currently have BigPond ADSL wired to the front of the house, about 80 metres from the shed in the backyard. As well as extending the ADSL to the shed/office, I'd like to share the connection with a second PC located in the house (for the kids).
Are there any benefits, performance or otherwise, to be had from setting this up in a specific way? e.g. Should I extend the ADSL line to the office, connect it to an Internet sharing box then run some CAT5 back to the house for the second PC? Or, leave the ADSL modem and sharing box in the house and do it the other way around? The "work" PC is Win2K and the kids run XP.
Also, If I extend the Foxtel cable to the office in the same conduit as the CAT5 and phone line, will I suffer interference between the cables?
No, there aren't any performance or other benefits to wiring in some particular configuration, beyond whatever's physically convenient, cabling-wise, for you. Your Net sharing device (probably a little anonymous plastic NAT "router" box, unless you've got good reason to use a Linux machine or something as your sharer) should be pretty much maintenance-free and tolerant of sitting in a cupboard, so you shouldn't need it to be in any particular location. As long as you don't need to tear up floorboards to get at it if its power supply dies, it can be anywhere.
You're also not talking about a remarkably long cable run, here; 80 metres is fine for plain old quality Category 5 Unshielded Twisted Pair, let alone the superior and now-common Cat5E. Expect no problems out to 150 metres of quality cable.
Talk to a cable supplier about the wire you should use, though; ordinary cheap cable for basic indoor applications may not be the best choice for use in outdoor conduits where it may have to deal with temperature extremes and moisture. You shouldn't pay a huge amount more even for top-spec "plenum rated" fireproof cable, and it might save you considerable aggravation some years down the track.
Back in the days when coaxial-cabled 10Base2 networks were all that home and small business users could afford (my old networking guide was written during the 10Base2/10BaseT changeover period), connecting different buildings together with an Ethernet cable was not a good idea. The coaxial cables have an earthed shield that can, but is not meant to, carry considerable current. Ground potential differences between buildings connected with this sort of cable can thus result in data corruption, hardware damage, electric shocks and fires, in increasing order of entertaining-ness. The UTP cable used for modern 10BaseT and 100BaseT has no ground conductor, so this isn't an issue any more.
You shouldn't have any interference problems if you bundle the pay-TV cable in with the data wire, but there might be attenuation issues over that distance; you'd have to talk to the cable TV people about it.
While browsing (haha) through a local computer store I came across some software called GhostSurf, which claims to allow someone to browse the Internet while masking their IP address even from their ISP.
As I understand how it, when one surfs the web it goes something like this:
User request ---> ISP ----> Web site
Web site ----> ISP ----> User
On the box, GhostSurf clearly states that it hides the IP address of the Web sites visited from one's ISP, although on their web site there is only an oblique reference to this.
So my question is - is this fraud?
No program will hide your IP address from your ISP; your ISP gave you your IP address. You can, however, quite effectively anonymise your Internet actions, so that neither your ISP nor the servers you're accessing can easily tell who you are.
If you use an anonymising proxy out there on the Internet somewhere, and ask it to show you hotnudetrees.com, and it does the DNS lookup to find that site, your ISP won't be doing the lookup. Your ISP's DNS server knows the IP address for hotnudetrees.com or can find it from the DNS system, but it doesn't know that's what you're looking at - unless your request for the site is sent to the proxy without encryption. If your connection to the proxy is encrypted, which it should be, the ISP won't know what you're doing at all.
I've heard tell that GhostSurf ain't all that it says it is, but there's little information about the program on the Web, so I've no firm opinion about it.
I have found a local Australian seller for Luxeon Star LEDs. They're a very reasonable price, too!
The site is the Alternative Technology Association, in their shop section. They also sell another kind of LED called "Superflux" which are made by Lumileds too, and they seem to be very good value for the money.
Note that the Superflux LEDs are just regular small-die LEDs with much better heat sinking, allowing them to run at as much as 70mA continuously for extended periods. This gives them around four times the power of an ordinary 5mm LED, provided you get rid of the heat somehow (generally not terribly hard, at only about 400 milliwatts per LED).
A one watt Luxeon Star is good for around ten times the power of a regular LED, though, and Lumileds are getting even higher power units with reasonable lifespans to market now.
The Superflux LEDs are also better suited to wide angle area lighting than to spotlight applications.
But, as you say, they're good value!
You say that most PCs only need a 300W PSU. I have two computers. One is a P4 3.2ghz, GeForce FX 5950 and has eight 7200RPM IDE hard disks. The other is an AMD 1.6ghz with a GeForce 3, seven 10,000RPM SCSI drives and one 7200RPM IDE hard disk. Both computers have 550W PSUs.
My question is, can I safely step down to a lower wattage PSU, so it doesn't sound like an aircraft starting up when I power on either PC?
Your drive-farm PCs may actually need their 450W PSUs, or at least a bit more than a 300W unit could deliver. The regular running power draw for these systems isn't likely to strain a quality 300W PSU, but those serried ranks of drives will suck a lot of juice when the machines are first powered on.
The SCSI screamers in the second machine may actually draw less power on spin-up, because they're don't expect to be spun down very often. Consumer hard drives expect to be in power-save mode fairly frequently, and so often have faster spin-up times, so their users won't be waiting too long when they come back to the power-saving computer and move the mouse.
Anyway, you should expect a modern hard drive to draw 10 to 15 watts when it's spinning, and 20 to 25 watts when it spins up. It may draw less, but you should be pessimistic when you spec these things out, if you don't have a sufficiently detailed spec sheet handy and can't be bothered wiring up a multimeter.
The power draw will be split between the 5V and 12V rails, but not in any generally predictable proportion. Different models of drive draw different fractions of their power from the two rails. A 50-50 power split (which means more current drawn from the 5V rail) is possible, and not too dangerous to assume.
Now, it's not too critical that a PSU hold all of its rails right up to spec on power-on; the system needs a lot of power then but it's not doing anything that mildly flaky power is likely to spoil. Still, 160 to 200 watts from spin-up of a bunch of drives, split between the 5V and 12V rails, on top of the draw from the rest of the system, may prove excessive for a 300W PSU. In particular, the 5.5V rail draw may push the aggregate load on the 5.5V and 3.3V rails over the line. The overload wouldn't be a severe one for a decent 300W PSU, and the system might gasp its way into operation and be fine once spin-up was over, but I'd still use a PSU with a genuine 350-plus-watt rating to get a bit of headroom.
(A couple of readers have pointed out that with a bit of jumper and/or BIOS tweaking, you can probably set the SCSI drives to spin up one at a time, in order according to their SCSI ID, and thereby avoid any power surge problems. This'll make the computer take a lot longer to boot, of course, but booting isn't something a server generally has to do very often.)
Regarding your noise problem - all of those drives, particularly the 10,000RPM units, need a decent amount of air flow to keep them cool. 7200RPM consumer drives are made to tolerate high temperatures (within limits), but a pile of them like you're using mean you really ought to shift a decent amount of air over them if you want them to last. High end SCSI server drives expect their own forced air cooling, even if you've only got one of them; it's hard to make a 10,000RPM drive-farm that doesn't make a racket.
You might like to explore the idea of attaching air guides lined with sound-absorbing foam to the intake and exhaust points on the cases. A pretty simple "muffler" design can make a big difference to the noise level, while still allowing OK air flow.
I am hoping you may be able to give me some clues about a rather baffling battery problem I have. My little nephew was given an el-cheapo remote control toy (a little plastic aeroplane that zooms around the floor) which takes two AAA batteries. My sister had put a new set of Energizers in the toy, but it didn't work, so she asked me to have a look. I couldn't get it to work either, so I tried the 1.2v rechargeable batteries from my cordless phone. Result: One happy two year old.
The obvious conclusion was that we had a dud set of Energizers, so we bought a new set of batteries (Eveready Gold, I think) but they didn't work either.
Eventually my sister went to a local electronics store and found that one brand of their batteries (Tandy Alkaline Enercell) worked, while others didn't.
Any idea what's going on?
My first guess would be that the toy has odd battery contacts, which don't manage to touch one end of most kinds of AAA. This problem generally appears the other way around - non-rechargeable batteries work, but rechargeable ones don't - because rechargeable batteries often have a flatter positive "nipple" than non-rechargeables. If the positive terminal doesn't stick out far enough, it doesn't touch some kinds of battery bay contact.
I noticed you used a smoke generator to test a rig you built. My question is what type of smoke is generated (CO2, cigar, cigarette, joint, etc) and is it safe for a running PC? Or was that test with no components (especially RAM) in there? I would think the static charge of a motherboard would attract the particles in the smoke and cause a problem (the same reason you never want to buy a used car from a smoker - it gets into everything electronic and wreaks havoc).
There are two basic classes of non-pyrotechnic smoke machine - ones that use water-based fluid, and ones that use oil. "Oil crackers" are more suitable for filling a large area with the kind of faint mist you need for laser light shows and such, but water based machines are the only choice for fairly thick, hanging smoke that humans have to breathe. They can also create "dry ice"-type low-lying fog if you pass the smoke through a chiller box.
Water-based smoke machine output will leave a residue, but only a very light one; a large amount of it might be a problem for hard drives and CD-ROM drives and such, but a quick burst isn't going to gunk up anything enough to do any harm. Actually, even oil crackers probably can't do a PC much harm, but I haven't really looked into that.
The smoke fluid used by most water-based smoke machines has as its main active ingredient triethylene glycol; it's pretty innocuous stuff. You have to really choke a room with it - like, put the smoke machine in a small sealed office and use up a half-litre of fluid - in order to leave a noticeable glaze on anything.
Note that while electronic gear that belonged to a heavy smoker for a few years may indeed be profoundly scungy as a result, only high voltage components (the yoke on the back of a CRT, for instance) will actually attract the smoke. Usually, smoke just gets laid down on the convection or forced-air ventilation path through the device.
The question I have is this: Have you run either tank through (shallow) water or snow, and if so how well does the inside stay dry? if it gets pretty wet, how do you feel you could best increase the water resistance of internal components?
I haven't gone boggin' with either of my Tamiya tanks, but I can tell you that they're not sealed, in standard trim. Splashproof, sure, and possibly OK for running in puddle-free snow, but I wouldn't drive them through water or mud deeper than their road wheel axles. Amphibious, they ain't.
You could address this problem fairly easily though, I think, as long as you didn't want to make them waterproof all the way up to the sprocket axle. Everything else could be, for instance, painted with "liquid electrical tape" for a decent waterproof seal.
You could also employ the old backyard basher's trick of putting the electrical components in balloons and sealing up the necks with Shoe Goo-type all-purpose sealer-glue. There are also highly water-repelling sprays and impermeable lacquers that're used on circuit boards by R/C boaters and other people who expect to find some water inside the hull of their toy.