Atomic I/O letters column #72
Originally published in Atomic: Maximum Power Computing Last modified 16-Jan-2015.Interference over Ethernet?
I just read the article on Ethernet over Power and Power over Ethernet in issue 71 of Atomic, and it reminded me of something I've wondered about for a long time.
As a trainee technician that works in a lot of corporate environments where there are a lot of data and power cables running, I know that the two don't mix very well. The reason for this being that there is an obvious hazard if a power cable becomes exposed and manages to touch a worn down data cable at the same time there's a fire/electrical hazard. It's an unlikely event, but stranger things have happened.
There's also the fact that having power cables running in close proximity to data cables can cause attenuation to occur within the data cable.
So what I'm wondering is, how do the data pairs in a Cat 5e cable react when there's power running over other pairs in the cable? Is it simply the case that the 13 watts max running over an Ethernet cable is not enough to cause a noticeable degradation in performance, or is it something more tricky?
Paul
My new LaLoPoE (Lots and Lots of Power over Ethernet) standard will surely sweep the world!
Answer:
PoE is, as you say, quite stringently power limited - the absolute maximum is 48 volts at 400mA, so the worst case
scenario is only 15.4 watts going where it shouldn't.
That's not likely to ever happen, though, because the standard says that the supply shouldn't deliver any power if it detects a short or open circuit (defined, rather strictly, as anything outside the 15 to 33 kiloohms PoE uses to identify device classes).
It's still theoretically possible for one loose copper filament to spark somewhere and make a nuisance of itself, but it should be really, really difficult. The usual problems that arise when, for instance, a roadie mistakes the Cannon plug carrying mains power for the one that's coming from the speaker cabinets, should be completely impossible.
Yes, power cables next to signal cables can indeed cause problems. Interference, not strictly attenuation, but I know what you mean.
This is also impossible with PoE, though, because it's 48 volts DC. No alternating current, no changing magnetic or electric field, no capacitive or inductive transfer of the AC frequency to other conductors. The field changes if and when the amount of power being drawn changes, but this is unlikely to happen at anything like the frequency needed to cause interference.
This doesn't mean that dodgy PoE supplies couldn't deliver wobbly DC that causes interference anyway, but it'd have to be pretty hideous to do anything noticeable in normal computer environments. Radio telescope installations, yes. Ethernet LANs, no.
It's not at all the same thing as ye olde network cables installed over the fluorescent battens in the ceiling, which can be counted upon to drop three bits out of every five.
Swap begone!
I upgraded to 1Gb of RAM. I have WinXP Home and a 1.1GHz Duron. I don't want a swap file, for performance, so I changed my swap file to 0Mb in My Computer. And at My computer/Advanced/Performance Settings/Advanced/Virtual Memory, the total paging file size says zero megabytes.
In regedit I navigated to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management, and changed DisablePagingExecutive to 1.
But in the FreeMeter system monitor it says I'm using about 300Mb of RAM and still about 200Mb of page file.
What is happening? Is there another program to confirm the swap file is zero?
Glenn
Answer:
FreeMeter may be telling you about data that Windows wants to swap out, but can't, or something - you can't really
disable virtual memory on WinXP, only stop it from working properly.
It doesn't really matter, though, because you shouldn't be doing this in the first place.
It is not generally a good idea to try to run WinXP without a swap file. It's possible, but many programs (not to mention a bunch of default XP services) assume the swap file to be available, and may fail in ugly ways if it's not there.
Often, in particular, a program will request much more memory than it's actually likely to need, assuming that this is no big deal because it'll just get a chunk of swap file allocated to it and only use as much physical memory as it really turns out to need. With no swap file, though, every program that asks for 512Mb because it might possibly need it if you open 50 documents will get its own 512Mb inviolable slice of your precious physical RAM, which nothing else will be allowed to use.
Fortunately, generally speaking, the swap file does not actually slow Windows XP (and later) down. Heck, it didn't even really slow Win95/98 down that much. It's not having enough physical RAM that slows Windows down.
Windows can actually run faster with a swap file, because it uses it to unload seldom-used components to free up physical RAM that it can then use for, for instance, disk caching. WinXP's virtual memory management is actually quite smart.
With only a gigabyte of memory, you're not going to be able to run many apps at once without a swap file. The OS itself, even with System Restore and so forth turned off to avoid dire where's-the-swap errors, will eat a lot of your RAM.
Swapless XP is more feasible if you've got more than 2Gb of RAM, and it can be handy for specialised setups. Business boxes that boot from Flash disks that can't handle a lot of write cycles, for instance. Such computers are unlikely to suddenly be used for 3D games and video editing, so a hard ceiling on their memory isn't too dangerous.
For general purpose computing, though, swapless XP still a waste of time.
XP will never be blazing fast on your computer, but it'll run more than fast enough for most purposes as long as you've got enough physical RAM. If you're short of physical RAM, the only solution is to buy more of the stuff.
If you want to get tricky about it, then you can put your swap file - or your whole Windows install - on a RAM-based device, like a Gigabyte i-RAM card. That can let you wring out a bit more performance. It's goofy to do that with a computer that's worth less than an i-RAM with a couple of gigabytes on it, though, and there's close to no point to doing it at all if you haven't already installed at least the magic 3Gb of RAM on your motherboard.
Wanted: Two-dimensional fan
I have an Xbox 360. The heat sink above the GPU is not good enough but it's custom for the case, so I'm looking for a tiny fan I can put on top of the heat sink.
Do you have any suggestions? The gap between the DVD drive and the top of the GPU is only 1/4 inch.
Todd
Step 1: Buy one of these.
Step 2: Take your hatchet, and carefully cut a hole in your Xbox...
Answer:
Unless you feel like running the DVD drive outside the casing, there's not going to be enough room for a conventional
"axial" fan there.
There are two DIY options, though.
One is to use one or more centrifugal "blowers" rather than a fan. See a blower compared with fans on this page, for instance; most of the companies, like Comair Rotron, that make small axial fans make blowers as well. You should be able to get them from any good electronics store. If you put one or more blowers at the end of the heat sink where there's a space on top of the circuit board you should get some solid air flow over the heat sink, though cramming even small blowers in there could be a pain.
The other option is to just chop a big hole in the top of the 360's casing and install a large fan there. The standard 360 ventilation is provided by a couple of tiddly 60mm exhaust fans; even a quite low power 120mm fan hacked into the middle of the top panel should delver considerably more air flow, so you might well be able to just remove the standard fans.
There are plenty of after-market 360 fan products, of course, from bolt-on external fans of various kinds to dual-60mm fan modules that you can swap in place of the stock internal ones. You can't beat the airflow bonus you get from chopping a big old hole in the side of the casing, though. And if you put a dust filter over the new intake fan and remember to clean it, your 360 shouldn't turn into one giant fuzz bunny, either.
Positive, negative, cognitive, laxative...
I'm building a bench power supply from a kit and have hit a snag with the main filter capacitors. The two caps each have three legs, all different lengths, which is a problem because I don't want to blindly solder the third leg to either the + or -.
Googling gives me nothing useful, and it doesn't help when the brand of capacitor doesn't get one hit in Google.
The power supply kit in question is a Dick Smith K3206.
I did some resistance measuring with my multimeter between the legs and got:
From the middle to the minus, or the middle to the plus, it goes up to 40 megaohms, which is the limit of my meter.
From the positive to the negative it goes up to 3.9 megaohms, and seems to keep going up slowly.
Any help would be appreciated!
Komodo
Perhaps it's the world's ugliest transistor?
Answer:
The middle terminal appears to be connected to nothing because... it is. It's a dummy lead that's there only to stabilise
the cap. This design is fairly common for large electrolytics
- you solder the third lead to an unconnected pad on your circuit board, and it gives the big cap a firm tripod mounting
that stops it from wiggling around.
If there's no hole in the circuit board for the third lead, you can just snip it off.
Genuine "three terminal capacitors" do actually exist. They're not weird magic devices, though; they're usually two caps in one casing, or some other hybrid device.
You're seeing that slowly increasing resistance reading between the plus and minus leads, by the way, because the small current from your multimeter is charging the cap.
