Naked Science Forum
Non Life Sciences => Technology => Topic started by: Quantum_Vaccuum on 29/09/2007 05:09:39
-
I was wondering, how can such a small memory card compile up to 2 gigs or more sometimes of data? And travel from one computer to antoher. And how do usb drives actually work?
-
Here's the WIki article that explains everything...
http://en.wikipedia.org/wiki/USB_flash_drive (http://en.wikipedia.org/wiki/USB_flash_drive)
-
They work by having a type of transistor called an MOSFET (metal oxide semi-conductor field effect transistor if you must know). These will conduct or not dependent upon the voltage on a 'gate' electrode. They are used in all sorts of things such as audio amplifiers. The ones in flash memory are special because they have two gates one of them is not attached to anything and is so well insulated that if you can charge it up it will stay charged for several years. So you can use this as a form of memory, charge it up to store information then you can read this information by trying to pass a current through the transistor.
How can you get so much information in such a small space? 'Just' by making everything very very small.
-
I hear that we made the leap from 1GB devices to economical 4GB ones simply by storing more than 1 bit per memory cell. A few years ago, a single transistor storage element was either "on" or "off" and so represented just one binary bit of information. Now, with better manufacturing processes and readout circuitry, we can store several voltage levels on one transistor. 4 levels = 2 bits, 8 levels = 3 bits. So with barely any more complexity (read number of transistors, or area of silicon), capacities have been doubled/ quadrupled "overnight".
-
I hear that we made the leap from 1GB devices to economical 4GB ones simply by storing more than 1 bit per memory cell. A few years ago, a single transistor storage element was either "on" or "off" and so represented just one binary bit of information. Now, with better manufacturing processes and readout circuitry, we can store several voltage levels on one transistor. 4 levels = 2 bits, 8 levels = 3 bits. So with barely any more complexity (read number of transistors, or area of silicon), capacities have been doubled/ quadrupled "overnight".
But would this not make them substantially more susceptible to noise and ageing?
-
Advances in insulation & shielding have helped greatly too.
-
But would this not make them substantially more susceptible to noise and ageing?
You would think so. The fact that products exist would seem to imply the problem is not too bad, or can be worked around. I suspect that either now or in the near future these memory devices will be using some redundancy and error-checking and correction technologies similar to those used on modern hard disks.
-
But would this not make them substantially more susceptible to noise and ageing?
You would think so. The fact that products exist would seem to imply the problem is not too bad, or can be worked around.
All it means is that the marketing guys don't think the issue is that important to undermine sales (since most of these problems will only arise some time after the sale, by which time they already have your money, and the guys with more conservative designs have already lost their share in the market because they lacked the specs).
This, unfortunately is the way of much of the IT industry. I have seen far more problems with computers that are a few years old than with those that are a decade old, because new designs are cutting margins. Same problem with lithium batteries - they push the limits, and the batteries have higher failure rates.
-
The term USB only refers to the way the data is transferred from place to place, remember.
The technology used for the actual memory is what really counts as USB only defines / limits the data transfer rate.
On the topic of multi level logic:
But would this not make them substantially more susceptible to noise and ageing?
Yes, in principle, but they have already reduced the logic level voltages significantly in order to improve speed and reduce heating problems. CMOS used to use +/- 12V, I seem to remember and now, after a long time with logic supplies on 5V, they're dropping and dropping. Heat dissipation is proportional to operating frequency so you have to do something about that by reducing logic levels.
Reducing size tends to reduce crosstalk as, also, does improved circuit layout.
There is definitely nothing magical about binary circuitry. It's all a compromise. If you can increase your information packing density enough, then you can afford to do more error handling. (Deja vue??)