Dan's Data letters #2Publication date: 2-Oct-2002.
Last modified 03-Dec-2011.
After reading your comments on your new Canon digital SLR, I decided that perhaps I should ease up on my pro-emulsion, anti-digital campaign and give digital a try. I purchased the Canon S30 point-and-shoot and so far am quite pleased with the results. It's not the quality of my Nikon N90s but it is fast and much more portable.
My question is actually about CCD technology. I noticed that the S30's maximum resolution is 2048 by 1536, or an effective 3145728 pixel count on my "3.2 Megapixel" camera. Now, if I don't shoot in RAW mode (which is rare), I can scale the JPEGs down to as small as 320x240 (I think) but I had a thought at work the other day: does the camera actually use all 3.15 megapixels and then scale down (using some sort of raster arithmetic, I suppose) to the resolution, or does it just simply use a progressively smaller portion of the CCD?
Either way, I'm sold on the RAW format so it doesn't present much of an issue, but there are occasions when I will shoot in 1024 by 768 JPEG to conserve space.
If the camera did lower resolutions by using less of the CCD in a simple cropping sense, then you'd get a longer and longer effective focal length (more and more zoom) as you went to lower and lower resolutions. You'd only be looking at a rectangle in the middle of the sensor. This is exactly what "digital zoom" does; it just takes a crop of the normal imager area, and may interpolate it up to a higher resolution, but can't put back the detail that's been lost by using less than the whole sensor. You can do "digital zoom" in post-processing.
If the camera did lower resolutions by only using, say, every second output pixel, then that'd work. It'd be doing that after taking a full-resolution shot, though, because there's little point giving the imager the ability to disregard pixels when you can do it in the in-camera image processing stage, before the file's written to flash memory.
What I bet the camera is actually doing, though, is quite nice weighted scaling, not a dumb pixel-dropping routine. Some cheap and nasty digitals used pixel-dropping for lower resolutions, which resulted in really ugly images - see my old review here for a perfect example. Nowadays, image processing hardware is cheap and low power enough that cameras can do bilinear scaling and produce nice-looking results, and that's very probably what your S30's doing.
There are a number of other issues to do with sensor pixel counts, by the way. You may find this page interesting.
I need your advice on what I think may be a dreadful trend in spamming. The past few weeks, my beloved Yahoo e-mail account has recieved a disturbing number of messages that are replies from various mail servers telling me that the mail I've tried to send to email@example.com was undeliverable as joeblow doesn't have an account on that server, or some similar error.
Upon closer inspection of these e-mails it would appear that I've tried to send mail to joeblow informing him that if he's interested in watching siblings engage in various illicit acts, or has an odd curiosity involving the mating habits of various farm animals, he can see these things and many other oddities at a Web site that is conveniently linked in the e-mail. The really disturbing part is it would appear that I've sent this e-mail to a LOT of people.
Granted, I've been known to send some very interesting e-mails while under the influence of various legal substances, but my "sent mail" box does not have any evidence of these messages. I must conclude that some spammer has chosen my e-mail address to use in the "return address" field in their spam. I understand that spamming can be a profitable business, and I also understand that theses spammers don't want to put their address in the "Return Address" field, but why not just add an invalid address in there? Why use my address?
Can you tell me if there is a way to stop this? The e-mail headers don't seem to contain any useful information, other than that my address is indeed the return address. The links in the e-mails point to various personal pages on www.terra.es and unfortunately my Espaņol is not up to par, nor do I believe that Babelfish is sufficient to convey my displeasure with this practice.
(name expunged, to prevent any further address-guessing)
Yes, your address is being forged as the reply address for spam. This isn't new; it's just new to you. It happens to me all the time.
Spammers have been known to do this because they don't like someone. Joe@wherever.net tries to shut them down; ten million penis enlargement spams go out with firstname.lastname@example.org as the reply address.
Assuming you haven't been going out of your way to annoy spammers, it's not totally beyond comprehension that someone might have come up with your email@example.com address (letters changed to protect the innocent) by means of random keyboard-bashing. Type six alphabetic characters and any given sequence has a one in 308,915,776 chance of coming up; given the amount of spamming going on today and the clear qwertyuiop-ishness of many of the reply addresses, some dork may well have chanced upon your address. Or maybe they just plugged in an address from their faithful One Zillion Guaranteed Opt-In Customers CD, for whatever reason, and yours was that address.
Spammers are idiots. Who knows why they do things.
Fortunately, people who've got their act together won't be complaining to you. Everybody sensible knows that the reply address on spam means nothing; only newbies complain to it, and no competent sysadmin will pay any attention to their complaints. This isn't to say that an incompetent Yahoo admin might not ban your account, but I wouldn't lie awake worrying about it. If your cow-orkers are irked, and/or some manager decides you're a spammer and wants to fire you, that sucks, but an appeal to the company sysadmins (assuming such people exist) ought to clear things up.
Can I tell you if there is a way to stop this? Yes, I can tell you. No, there isn't.
Well, short of killing all spammers. I'm developing a plan involving red hot crowbars, but I haven't worked out the details yet.
Stopping the spam is hard, but sending complaints is easy. Check out SpamCop.
Last year I finally upgraded from my P133 to a Duron 900 - it was something of a shock to see a computer do things so fast! The motherboard that I got with it was an Asus A7VL-VM, and 128MB RAM to go with it. Everything was fine untill I recently bought a 256MB DIMM so Windows XP might be a tad faster.
An Odd Thing happened... Windows refused to load whenever the DIMM was inserted..
I then went through the whole house, trying out all the different PC133 RAM in the house. There is some 736MB of PC133 RAM here; 3x32Mb, 3x128Mb, and 1x256Mb.
None of the 32MB DIMMs work at all in this computer. Two of the 128MB DIMMs work in this computer - but not both at the same time - Windows hates me when I use both at once. The remaining 128MB DIMM is suspected of being corrupt. The 256MB DIMM is brand spanking new, but Windows hates me if I use it in this computer.
When I say "Windows Hates Me" I mean it gives me the good old NT BSOD with "Unknown Hard Error: NT.DLL" or something like that.
Anyway, seeing as there's little to nothing that I can do about this, I was thinking of upgrading to another motherboard. I was looking at the Jetway 867AS-H and the ECS K7S5A - both around $AU100. Notice that both motherboards support DDR and SDR modules - I have been told that such motherboards have stability problems. Is this true, or is it just an urban myth?
Your Odd Thing isn't that odd, if you've got a bum memory module. Very cheap RAM often has compatibility problems of one kind or another, and if the module's been mishandled, it could be sufficiently static-damaged that it isn't completely dead, but still won't work right ever again.
The various other compatibility problems you're seeing, though, make this look more like a problem with your motherboard, which is possible. Motherboards can be static-damaged too. Asus boards don't have a particularly high fault rate, but you can kill anything if you're talented enough.
Memory modules that work alone but not together, like your 128Mb modules, are common enough. The problem could be caused by the motherboard, or by the memory, or both.
If I were buying one of the two alternative motherboards you mention, it'd be the ECS one. I'd steer clear of the Jetway one. I don't know of any endemic problems caused by the "dual fuel" design itself.
I wouldn't buy either of them, though, if I had a choice. Since dual fuel boards are all earlier models that don't get much benefit from DDR memory, there's not really a lot of point to them. It's not as if you can use both kinds of memory at once.
If you hunt around, you could probably find yourself a top-class SDR-memory board - an Abit KT7, say - pretty darn cheap. That's what I'd buy.
Of course, if it's the memory that's toast, a new motherboard won't help at all.
I have a Linux server that I would like to keep running 24x7 - however the cheap ass power supplies you generally get in PC systems just don't seen to be designed to do this. The one in my system, after 2 weeks of constant operation, seems to have started making funny noises that haven't really stopped.
What I'd like to know is whether those quality power supplies you review are designed to be "always on", or are they just bigger better power supplies that are still made for use only 8 to 10 hours a day?
Your noisy PSU probably has a bad fan bearing.
Super-cheap PSUs are amazingly reliable considering their price, but I wouldn't want one in an important machine. The various flashy case modders' PSUs that I've reviewed are OK quality, but there's no reason to presume they'll last any longer than a decent boring grey-box PSU from, say, AOpen. AOpen's PSUs cost easily twice as much as the real cheapies, but they're still cheap, and I'd be quite confident about using them in a 24/7 box.
Of course, for true reliability, you need hot-swap dual PSUs, so you can remove and replace a failed PSU without turning off the server. Various server cases come with these PSUs, but they ain't cheap.
I have a vicious problem with my monitor... the image shakes like an absolute BITCH when the house is placed under high power load. The shaking pretty much renders the computer unusable.
I have come to the conclusion that it's not my hot water service, which many people link to monitors with this problem, but rather simple under-volting or whatever it may be from power use in our house.
Is there any way to fix this? I recently purchased a powerboard with basic RFI filtering, and that didn't do too much to stop the shaking. It helped a little though. If the problem is lack of power being supplied to our house, how can that be remedied?
This sort of thing is often caused by quirky building wiring. If it happens in a home, the quirk is likely to be an actual fault, like a piece of steel in a wall that's carrying current when it shouldn't. It could be worth getting an electrician in.
Assuming that's not the problem, though, it's possible that it's an undervoltage problem, I suppose. Again, that may be indicative of a serious fault somewhere, but if you just want better power for the monitor, then you can get it by using a power conditioner. Some power conditioners are scarily expensive, but cheap ones exist, and may solve your problem; see here for instance. Note that I have no idea whether this store is any good; it's just what a search turned up.
Also, try changing the screen refresh rate you're using. That may reduce or eliminate the shaking problem.
I've just been reading your Ethernet Networking Explained page. At the end, it mentions STP cable. What is this?
STP is Shielded Twisted Pair. It's like regular twisted pair network cable, but with an earthed shield around the conductors for better noise resistance. STP works the same as UTP, but you won't see it often used for normal networking, mainly because it's more expensive and less flexible. Furthermore, its better noise resistance isn't necessary for most applications.
Any twisted pair cable has pretty good noise rejection, because of the twisted conductors, which are used as balanced transmission lines.
Only two of the four wire pairs in ordinary Category 5 (Cat5) network cable are used for 10/100BaseT Ethernet; one pair has the positive and negative transmit wires twisted together, and the other pair has the positive and negative receive wires. Each negative wire carries an inverted version of the positive wire's signal, and twisting the cables means that any incident RF noise should affect each wire in each pair evenly. Since the noise will be the same on each wire, but the signals will be inverted, it's possible to delete whatever's the same on each wire and thus recover a low-noise signal. This technique allows quite long cable runs at quite high data rates without any shielding.
Earthing's also a bit of an issue for STP; shielded cables need their shield earthed at exactly one end, to avoid earth loop problems when the ground potential at the two ends of the cable isn't the same. This'll usually be the case, thanks primarily to mains power that's not in phase in different places due to the waveform-skewing effects of other loads.
STP cables should work OK with gear designed for UTP, but if you don't ground the shield somehow, the fatter, more expensive cable won't work any better. It may actually be worse; an ungrounded shield is basically a great big antenna for all kinds of RF, including noise from the cable it surrounds. It can actually worsen crosstalk problems.