Atomic I/O letters column #79Originally published in Atomic: Maximum Power Computing Last modified 16-Jan-2015.
I still think Video Floppy may make a comeback
We have been discussing CompactFlash cards versus Secure Digital Cards at work as to which is the best and why.
It looks like all the old digital cameras used CompactFlash and the newer ones SD cards, but then most digital SLR cameras seem to use CompactFlash.
Is it because the CompactFlash cards use parallel data transfer and SD cards use serial and therefore the CF are faster? But this contradicts the PATA versus SATA HDD transfer speeds.
This isn't a serial/parallel thing. Secure Digital cards do actually support a one-bit-at-a-time serial transfer mode, but that isn't really relevant here.
A CompactFlash card is, essentially, a little ATA ("IDE") drive. You can use a cheap pin adapter to plug a CF card into a PC PATA cable (the picture above is of the CF card connector on just such an adapter), and the card will appear to your computer as another hard drive. You can also get slightly more expensive adapters that turn a CompactFlash card into a SATA device, using the same sort of bridge hardware that does the same job for larger drives.
(There's also an upcoming, incompatible SATA CompactFlash standard called CFast, but that need not detain us here.)
The advantages of the CompactFlash approach are that it can give very fast transfer rates (CF cards can, in theory, use all of the standard ATA DMA modes), and the standard is open and free. Anybody can make CompactFlash cards or a device that accepts them, without paying license fees for the technology (though using the CF logo still costs money!).
On the down side, the CF card socket has lots of little pins which, if they get bent, are extremely difficult to straighten. CF cards are also relatively large, and difficult to incorporate into small devices like pocket cameras. And high transfer rates don't matter most of the time - only if you're shooting rapidly in RAW mode, or using the memory card for storage in some other high-data-rate application, are you likely to actually need more speed than the slower card formats offer.
SD and MMC cards (Secure Digital is MultiMediaCard plus some Digital Rights Management options that almost nobody ever uses) are more than fast enough for almost all photography, far smaller than CF, and have a simple connector that's hard to mangle. The SD standard isn't free to use, but it's also obviously not expensive enough to discourage many companies.
The bigness of CF cards means they can have more than one actual Flash chip inside, giving a higher ceiling capacity; CompactFlash has always had the largest-capacity cards.
Again, though, only professional photographers are likely to care about the very expensive maximum-capacity cards at any given moment, and Flash RAM technology can now fit multiple gigabytes on a single very tiny chip. So there's not much advantage there now, either.
I needed to replace the fuse in my Logitech Z-680 speaker system, so I bought the correct one, came home, turned the master switch off on the back of the subwoofer, turned the power off at the wall that leads to the power board, and grabbed my Phillips screwdriver.
I placed the fuse on the end of the tunnel, then nudged it the rest of the way into the fuse cavity, not touching anything, but accidentally rested my left arm on the heatsink. And I got one hell of a boot!
Now I'm not game to touch it at all.
Why did I get shocked, and how can I prevent it? It baffles me. A friend told me that I need to unplug the powerboard from the wall. But I switched everything off - isn't that enough?
The only way this could have happened without some kind of electrical fault is if undischarged capacitors in the speaker's power supply were connected to the inner terminal of the fuse socket.
This doesn't seem very likely to me - the fuse socket ought to just be in-line with one side of the power transformer - but who knows. If this were the case, then when you leant on the heatsink, your body provided a path to the amplifier chassis for the capacitor charge, and zap.
If this is what happened, then it would have happened whether the speaker was still connected to the turned-off wall socket or not. Devices with big caps in their power supply all ought to have something - usually a high-value resistor - that drains the capacitor charge over a relatively short time when the device is turned off, but not all of them actually do.
The other ways your shocking experience could have happened all involve one or another kind of electrical fault.
By far the favourite in this particular race is a wall socket whose switch interrupts the neutral conductor, not the active one. In that case, turning the wall switch off will leave the active conductor all the way into those devices live. There's no path back to your house wiring for the electricity to follow, so things plugged into that socket will turn off when you turn the socket's switch off. But anything directly connected to the active line, including one side of the fuse holder in your speakers, will be live and awaiting developments.
The earth conductor isn't affected by switches; it's connected all the time. If a power cord has an earth pin (and if the earth contact in the wall socket is actually connected to earth, which is another thing you shouldn't take for granted...), then its earth conductor will always be connected to ground while it's plugged in. The earth conductor is connected to the chassis of all grounded appliances, so touching the chassis and something that's live will allow current to flow through you.
(If you touch something that's live when you're not also touching something grounded - which includes just touching the ground, without insulating clothing in the way - you'll probably still feel a tingle as a small amount of current leaks through you in the general direction of earth by whatever paths it can find. If you feel a real belt, though, then there's a proper current path through you.)
Incidentally, this is exactly the kind of situation in which a "safety switch", or "residual-current device" should trip. Safety switches monitor the current in the active and neutral conductors, and cut the power if they notice a significant difference. That difference can only be caused by current passing from active to ground somewhere. I talk about this some more here.
I have a laptop. I need to take it into an alumina refinery. Inside the refinery are "pots", where alumina powder and large electric currents meet to produce aluminium.
Feeding the pots are bus bars. The bus bars have about 200,000 amps going through them. This creates strong magnetic fields.
How close can I get before my computer stops working? Or, how strong a magnetic field (gauss) can a laptop handle?
The strength of the magnetic field around the bus bars - or any other conductor - is directly related to the current passing through them. The equation is
B=(1.257 * 10^-6 * I)/(2piR)
...where B is the field strength in tesla (one tesla is 10,000 gauss), I is the current in amps, and R is how far you are from the centre of the conductor.
For 200,000 amps and a distance of only one metre, this still gives only 0.04 tesla, or 400 gauss. That'd try pretty hard to grab a spanner out of your hands, but it's barely enough to wipe a floppy disk, if that.
You need a few thousand gauss at least to wipe a hard drive. You'd get that if you were only 10cm or so from the centre of the bus bars, but the bars and their surrounds are probably too thick for that to be possible.
UPDATE: After this page went up, someone who's designed alumina smelter electronics e-mailed me to add some more details.
Apparently the busbar current (where he was working, anyway) is likely to be closer to 400,000 amps, giving field strengths up around 800 gauss, which "will hold a crowbar vertically with no trouble, and makes walking with steel-cap boots almost impossible". But which still won't wipe a hard drive.
That seems unlikely to matter, though, because that strong an external field is likely to apply enough bias to the ferrite cores in the inductors and transformers of laptop (and many other) power supplies that the PSUs won't work any more. My correspondent says they had to specify magnetic components "around three times the usual size to avoid issues due to this".
Plus, carrying your laptop through an area that's full of highly abrasive alumina dust is not the greatest idea.
So, as it turns out, there are actually some very good reasons why a laptop ain't gonna work well in an alumina smelter, especially if you insist on carrying it around near the busbars. But those reasons are not the obvious ones.
Since I have some money burning a hole in my pocket and a computer that could do with an upgrade, I decided to look around to see what my Core 2 Quad options were.
I soon discovered that I needed a new motherboard. The Asus P4P800 I have for my P4 2.8 GHz CPU is a Socket 478 board, and for modern Intel CPUs I need a socket 775.
However, I have some apparently rather difficult to satisfy demands for the motherboard. Since I bought a new AGP graphics card not long ago (a Radeon X1950 Pro), I would like a motherboard with both an AGP slot (for my current card) and a PCIe slot (for future upgrades). In addition, I need connections for at least four PATA devices and one SATA device (4 HDs and a DVD-rewriter).
After some searching, I could only find a single motherboard that would satisfy these requirements: the Asrock 4CoreDual-SATA2. This mobo, however, is suspiciously cheap compared to the popular P35 boards overclockers are using with Core 2 Quads.
Am I missing something that makes that makes this board much less desirable, other than the lower number of PCIe slots?
Also, would I still be able to use my current two 512Mb sticks of DDR400 memory with a modern board, or would I have to buy new memory?
I think the best option, really, is to get a whole new computer. That way you've still got your old one as a working PC.
If a second computer's of no value to you, though (and you don't want the hassle of selling it off), then a Via-PT880-chipset board like the 4CoreDual-SATA2 may fit the bill. I've got some serious reservations, though. The PT880 was a pretty cool chipset in 2004, but it's not much of a competitor today.
For a start, apparently quad-core CPUs are very far from being certain to work, at all, on many PT880 boards.
The 4CoreDual will also... probably... let you use your old AGP card. Apparently PT880-board AGP slots aren't perfectly bug-free, either. But if the card works, you will indeed have an upgrade path to a PCIe card later on... except that some PT880 boards only have an x4 PCIe graphics card slot rather than the standard x16. The manual for the 4CoreDual is unclear about this.
If this board only has an x4 graphics card slot, then PCIe cards won't run quite as fast as you'd expect, and some PCIe cards probably won't work at all.
The 4CoreDual will, at least, let you use your old RAM, though it only has two DDR slots plus another two DDR2 slots, and you can only use one kind of RAM at a time. So memory upgrading won't be an economical procedure - and the board will only let you have a maximum of 2Gb of memory, anyway.
I don't know about overclocking, either. Don't expect the 4CoreDual to have sterling support for running any CPU above stock speed.
If everything on the PT880 boards worked as advertised then they'd still be a reasonable budget option for upgraders. But the Curse of the "Crossover" Motherboard does seem to have struck here, once again; mobos that let you use both old and new hardware seldom actually end up being worth the trouble.
Personally, I'd save up until I could afford a whole new P35 or X38 or whatever system, and keep the old computer for backups or multiplayer games (or just sell it, of course).
Your need for at least two Parallel ATA sockets in your new computer, by the way, isn't a serious obstacle. Many new motherboards only have the one PATA socket (because PATA optical drives are still standard items), but you can get a PCI PATA controller card for $notmuch on eBay, and run your other drives from that.