Thread: True Errors in Memory and Storage

Answers to your questions might be found here:

http://britneyspears.ac/lasers.htm

Almost universally now, computer memory blocks have extra bits used for error detection and correction, (so a 32 bit word may actually be stored in 36 or more bits). The actual "value" of a memory location is the value "calculated" to be there by the actual data bits and the output of the EDC, (Error Detection and Correction), unit. EDC systems vary, some are able to simply detect an error, (simple parity type things). Others are able to detect multi-bit errors, and correct them - you pays your money etc.

There are two basic types of error. Hard usual arise from manufacturing faults and are always there. Soft are transient, sometimes just a borderline section of the crystal lattice which works some times and not others, (temperature effects etc). An interesting class of faults is the impact into a critical part of the lattice by, for example, cosmic ray particles. Certain components are built much larger than they really need to be to in order to reduce the chances of an impact on a critical point, (i.e. the particle can hit the lattice, but there are enough atoms in the lattice to make the impact negligible). These type of components are used often in space, and for nuclear hard military hardware.

Typical EDC systems employ Hamming codes. Search for this, or start here...

http://www-groups.dcs.st-and.ac.uk/h...s/Hamming.html