Slashdot Mirror
Nickel Sensors Could Raise Hard Disk Capacity
Makarand writes "Tiny filaments of nickel, thinner than a wavelength of visible light, acting as magnetic
sensors may expand the storage capacity of hard disks many times. Although, technologies
exist to increase hard disk capacity, reading data bits reliably from such disks has proven
difficult because as data bits become smaller their magnetic fields are weaker and difficult
to pick up. Nickel filaments are capable of picking up of these weak magnetic fields
using a phenomenon called "ballistic magnetoresistance" which is not completely understood.
As the sensors are only a few atoms wide the electrons travel along a straight line
in the conductor greatly enhancing the binary signal picked up from the data bits.
These sensors could also be used to detect biomolecules in low concentrations."
Computer evolution is starting to look like biological evolution (sorry to all the Creationists...)
Back when computers were akin to our old 1-cell relatives, it didn't take much to have a serious jump in usefulness. But now that they (and we) are vastly more complicated, significant improvements in individual aspects of the technology don't seem to affect the whole system as much, so they seem so much less exciting.
As I see it, progress is going to be coming more and more in small steps, taking much longer to affect a huge change.
But please, feel free to prove me wrong, I'd love to see the kind of jumps in usefulness that compters experienced back in the 80's.
Chaos, panic, disorder...my work here is done.
Hmm. Several points come to mind.
1) So what's so unreliable about current storage? Disks can and do eventually die, but so do ***ALL*** mechanical devices. The magnetic lifespan of a disk is not clearly the limiting factor in the life of a hard drive. Half of the drives I replace die as a fail to spin up properly--not something we like to see, but an indication that the short life of the magnetic states aren't the most unreliable part of a hard drive.
2) I don't see any indication that this is 'fragile' technology, on the macroscopic scale. Sure the signals are smaller, but once you reliably detect them they can be amplified ad nauseum, and reliable detection is what this is all about.
3) Large scale enterprise storage in our current realm of thinking, requires high speed access and high reliability, and does NOT involve single drives. Hardware RAID5, RAID 1+0, RAID 5+0 (I've seen it done!) etc. is the way to get high reliability and high performance. Having a single hard drive, even one that's 100% reliable, isn't a reasonable storage solution for mission critical data, and so consequently there's not a lot of demand for a 100% reliable hard drive.
"People who do stupid things with hazardous materials often die." -- Jim Davidson on alt.folklore.urban
Microscopic $/MB is great, but only if you use all those megabytes.
"They that can give up essential liberty to obtain a little temporary safety deserve neither liberty nor safety."
Maybe that says that we should be concentrating on fast solid state storage rather than trying to minimize the stuff we have now. Hard drives are currently the slowest and most unreliable equipment in our PCs today. Sounds like thats the bottleneck scientists should be working on.
It's easier to fight for one's principles than to live up to them.
Carbon nanotubes? Like fullerenes, except they're long cylenders instead of spherical. And they conduct electricity quite well.
Repeal the DMCA!
You hit the nail on the head. I have a specialized need. Capacity wasn't an issue; for various reasons, I'm using two 128MB modules, and that's all I need. On the other hand, the write limits don't really concern me. As I'm using custom software, I can trade off the number of write cycles in exchange for reliability. I basically limit it to a maximum of one write per location per hour, and start moving things around on the disk if a single section consistently gets more writes than that. Again, also, I can check after each write whether it worked; if not, I just mark that byte bad, and go to a reserved backup area. Assuming my usage patterns stay relatively similar, or even double, the number of write cycles shouldn't be an issue for roughly a century.
;-) to 5% at 1M. Clearly not quite linear, but I'm not too worried. At 10M writes per location, 65% of locations were still writeable. Not ideal, I'd admit (that brings a 128MB disk down to ~83MB), but again, that would take a millenium with my usage patterns. (Currently, I'm reserving 16MB out of 128MB for repcing bad sectors. 16% bad sectors occured at around 4M writes.)
Furthermore, I did destructive testing on a smaller unit (32MB) to get a sense of how accurate that 1M write cycle number is, since it did worry me for a bit. 1% of the write locations (byte sized) failed after 200k writes, going up to 2% at 500k, skyrocketing
So you're right in that this isn't necessarily feasible for a standard PC. On the other hand, with different usage patterns, you could easily just use a RAM disk. A good sized linux installation is reasonable in 2GB; this is under a thousand dollars in compact flash, with direct compact flash to IDE adapters. 2GB of ram, likewise, is only a few hundred dollars. On boot, make a ramdisk. Only write back to flash if you get nervous; or do it once an hour, and last a century. (Note also that most IDE drivers do bad sector remapping; it's not ideal, because it uses 512B blocks instead of the 1B blocks that flash actually fails in, but it's zero modification to existing code.) Needless to say, it would be more expensive than normal to build a computer this way, but it really wouldn't be off the scale.
I've had this sig for three days.
Ballistic doesn't refer to any new physical principle. It's the fact that the nickel layer is just a few nm thick that removes statistical properties such as resistance since there just aren't enough atoms in the thin layer for electrons to collide with, hence ballistic electrons. Prepare for ballistic transistors when they grow sufficiently small.