Atomic I/O letters column #43
Originally published in Atomic: Maximum Power Computing Reprinted here March 2005. Last modified 16-Jan-2015.
One switch for everything
I have an idea for a construction project. It's fairly simple really... modding a standard mains powerboard so that is controlled by the presence of power on a computer's USB port. Regular computer users might use this to automatically power up printers, speakers, network equipment, etc, while Atomicans could use it for water pumps, 240 volt cooling fans, etc.
I envisage installing a relay into the powerboard (with 5V coil and 240V contact rating), and connecting it to the computer via a USB cable. Although I could probably work out how to build this project without your help, I'd really like to avoid the possibility of killing myself.
How about it?
Daniel
Answer:
As you say, this can be done with a relay, and if you take adequate care
with isolation then it shouldn't be dangerous. You can also buy the thing
pre-made
for not
much - well, if you live in the States you can, anyway (find bigger
versions in the "Power Strips" section
here).
A more intrinsically safe version of the same idea, though ("intrinsically safe" is a good thing to hear in the description of something you're considering building yourself...), has two sockets on it (from either of which you can hang a powerboard) and powers up the second socket when it detects power being drawn from the first. This keeps mains away from USB, and can be used with all kinds of gadgets, not just computers. It's good for making a central power-up solution for a big home theatre setup, for instance.
It's still eminently possible to crispify yourself while building and testing such a device, of course, but at least a failure while it's running ought not to barbecue your computer.
You'll be happy to learn that this product also already exists, as a packaged kit (basic version here).
HSF bling
Looking at all these new and wonderful coolers based on copper, I am wondering what is the next step in air cooling. Then I thought, why not use silver!
Silver is a better thermal conductor than copper, though I know it's expensive. But why not put a thin plate of silver under copper heat sinks to improve the thermal transfer? Just like some hybrid aluminium/copper (Zalman 6000ALCU, Volcano 9...) coolers. Why not do the same with silver?
If a little smear of Arctic Silver can reduce your temperature by a few degrees, what can a whole plate do?
Ken
Answer:
Fine silver actually only conducts heat about 10% better than copper. Putting
a plate of the stuff on the bottom of a copper heat sink is, therefore,
pretty much pointless; you lose a little thermal conductivity in the joint
between the two metals, there'll be some galvanic
corrosion issues, silver itself is prone to corrosion, and fine silver
is a pretty impractical material - it's terribly soft. That's good news
for mating it to the top of a CPU, because it'll squish onto the top of
the die and get a good thermal contact, but it makes it hard to mount fine
silver onto anything else.
Interestingly, there's an alloy of 72% silver and 28% copper called "CuSil", which is much harder than fine silver (so is sterling silver, which only has a maximum of 7.5% other metals in it, usually copper), and is also supposed to have considerably better thermal conductivity than copper or silver. There's some doubt about whether this is actually true, though.
In any case, there have been a few gimmicky coolers over the years that've had silver plates on the bottom. The most famous is Noisecontrol's old "Silverado", which mated a silver disc with an aluminium heat sink, largely for marketing reasons.
Size on disk, size on C64 Datasette...
In Windows XP, what is the difference between an object's "size" and its "size on disk"? A file or folder's Properties always shows two sizes, sometimes the same, sometimes different. Why?
Chris
Answer:
The size is the actual amount of data in the file; the size on disk is the
amount of disk space it takes up. The second number is bigger because data
on disks is divided up into "clusters", which are the smallest allocation
units available. The standard cluster size for NTFS disks is 4096 bytes;
you'll notice that all files on such a disk have a "size on disk" spec that's
a multiple of 4096 bytes. Save a one byte file to such a disk, and it'll
take up 4096 bytes; save a 4097 byte file without disk compression, and it'll take up 8192 bytes.
This extra space is known as "slack" space. A file whose size is an exact multiple of cluster sizes, though, won't waste any space. It also doesn't matter if a file's fragmented.
Slack space used to be a big deal back in the days before FAT32, when disk space was more expensive and the FAT file system used 32 kilobyte clusters for partitions between 1Gb and 2Gb (and couldn't handle partitions larger than 2Gb at all). Nowadays, though, huge disks are cheap and sensible filesystems like NTFS can use small clusters even for big drives, so slack space very seldom becomes a problem.
Memory card booting
Is there any way to put WinXP Pro on a flash drive so that the computer will start instantly?
Since it's only read access, will the flash drive last, or run out of its "million" (or whatever) capacity read cycles?
If PDAs can start up quickly, why can't PCs be made to do so?
Sunil
Hot nude CompactFlash action!
Answer:
It's possible, but impractical.
Possible, because it's easy to adapt a CompactFlash card into an IDE (ATA) device; all you need is a passive pin-adapter.
Impractical, because of the following reasons.
An XP Pro install is likely to be more than a gigabyte in size. CompactFlash cards have plunged in price lately; as I write this, 2Gb cards are down to $US140 or so on eBay. This is still around 150 times the price-per-megabyte of a hard drive, though (a ratio that's stayed pretty constant for a while, now).
You can get those some-people-say-they're-dodgy Magicstor 2.2Gb CompactFlash cards for around half the price of well under $US200, but they're Microdrive clones, not flash memory, which of course defeats the purpose.
And, after paying for Flash memory, you wouldn't get instant startup anyway. CompactFlash cards in ATA mode only support PIO 4, which uses more CPU time than any of the DMA modes that all current ATA hard drives support, and can only shift data at a theoretical maximum of 16.7 megabytes per second, which is rather slower than the sustained read speed of various high density consumer drives these days. The card may also have its own overhead that makes the real transfer rate a lot lower.
A memory card has negligible seek time, which can be very handy for some things and certainly doesn't hurt for system startup. But the low transfer rate will kill the advantage.
And, even if that didn't bother you, you'd still need a hard drive to put your swap file on, because NT-series Windows flavours can't run without swap no matter how much physical RAM they've got. You can't put the swap file on the CompactFlash card, because of the limited write cycles the memory supports. You can, if you want to be devious, put the swap file on a RAM disk, but that doesn't actually improve system performance significantly.
PDAs have much smaller operating systems than desktop PCs. The total flash ROM capacity of a current high end Pocket PC is likely to be 128Mb. You'd probably need 10 of them to store just the files for a smallish WinXP install.
PDAs are also made to be able to suspend and resume their operation at a moment's notice, preserving the contents of their dynamic RAM with a trickle of battery power. When you "start up" a PDA you're usually just waking it up into full power mode again, not booting it from scratch. PCs can do the same sort of thing with sleep and standby modes.
Still better than drinking it
My friend was having an uninvited party in my home theatre room with his girlfriend. He spilled wine cooler into my Denon AVR-5700 receiver. I was not happy.
After about a week I took the receiver in to an authorized Denon service centre. They've now had it for three months, and have no clue what's wrong with it. They say that all the metal is corroded and even eaten away by the wine cooler in the small area that it dripped in, that I waited too long to bring it in, and that it may not be fixable for a reasonable price.
The repairer's suggesting all sorts of new expensive parts. Is this guy incompetent, or does wine cooler really corrode metal and cause that much damage to circuitry?
Jonathan
Answer:
After a week, yes, it could be that bad.
Wine cooler contains wine and fruit juice (rumours concerning the presence of human or animal urine remain unconfirmed), both of which are acidic. It's dilute enough that you could probably get away with leaving it on an average circuit board for a few hours, but after a week it's quite plausible that the board will be toast. I don't know the actual corrosion capability of an average wine cooler compared with more common electronics-killing liquids (cola, for instance), but all acidic liquids are bad news.
Plain water, however, is not. Lots of gadgets that've been doused with water, even while they were turned on, can be rehabilitated by just taking them apart and drying them out. The standard treatment for a gadget that's been accidentally drenched in Coke or whatever is to wash its parts thoroughly with clean water, optionally followed by squirts of alcohol into any crevices, to displace the water.
But if there's already a patch of rusted-out circuit board, and the board itself can't be cheaply replaced, you're often pretty much screwed.
