Flipping a bit
 

How do computer flip a bit (change a 0 to a 1)? Is it due to physical flaws in the chip? I was recently involved in a discussion involving an algorithm that was essentially treated as correct because it had a 1/2^80 chance of being wrong, and the reasoning behind accepting such an algorithm for computing purposes was that a computer makes an error every 2^50 bits or so. Anyone familiar with the more technical details of this, I'd appreciate your input.

Re: Flipping a bit
 

Computer memory has transistor flip/flop switch circuits that are defined to be in one state or the other unless reset by the CPU. In practice individual bits are not set, but rather a group of bits such as a byte or a mulitple of bytes depending upon the the CPU and memory architecture which defines the smallest and largest addressable unit of memory. Of course regarding storage hard drives, the information is stored magnetically, but is also occasionally susceptible to errors.

Re: Flipping a bit
 

Hiya Goodfriar,

I think what you are quoting is taken out of context.

The only way in which I could think that there is a bit flop would have to be, by an introduced error or approximation in the algorithm, which may have been done for efficiency, ie speed purposes; the only other way would be, a rounding issue due to storing information to a specific precision. For storage, this precision must be determined and some numbers in the continuum of numbers, of course do not allow this.

Altough a bit flop is possible, at hardware level, though very rare, the software would usually detect this and give an exception to the the user. This is achieved in the normal course of event by using redundant bits to validate a pattern. The most basic of this is known as the parity bit.

Re: Flipping a bit
 

Thanks for the responses. I realize it is both extremely rare and also detectible,I was more concerned with physically what causes the error, e.g. magnetic interference, entropy, etc.

Re: Flipping a bit
 

Below is an article that goes into a bit of detail. Basically, high energy particles and waves are the cause. The earth is constantly being bombarded by all manner of particles and cosmic rays (high energy x rays). Also, the ground and building materials contain minute traces of radioactive elements which spontaneously decay, releasing alpha particles and other radiation. If one of these encounters a transistor, they may have enough energy to change its bit state.

Interesting, the problem is getting worse as chips components get smaller. Smaller components use and hold less energy, which means they're more easily disturbed by incoming energy.

http://www.edn.com/article/CA454636.html

[ QUOTE ]
Trouble Spots
The four common sources of SEUs are low-energy alpha particles, high-energy cosmic particles, thermal neutrons, and poor system design (Table 1). Low-energy alpha particles are generated by the radioactive decay of trace uranium-238 and thorium-232 in quartz filler used in mold compounds, or from polonium-210 in lead bumps that...

[/ QUOTE ]