8" floppy system needed to recover old game data

der Mouse mouse at Rodents.Montreal.QC.CA
Fri Oct 7 23:25:47 CDT 2005


>> I would like to be able to pay a reasonable amount -- say \pounds
>> 5.00 to \pounds 10.00 for a 3.5" disk with the same quality level.
>> Because my data is worth a lot more than that.
> Maybe what's needed is a lower density format and/or a redundant
> filesystem using the current media.  Surely a redundant filesystem or
> lower density encoding scheme that only tried to fit 88K on a 3-1/2"
> HD floppy diskette would succeed in having MUCH greater longetivity.

You can't do much about the width of a track, because head size is a
physical constant of the drive.  In principle current hardware could
put flux transitions farther apart or some such, but I don't know
whether the existing controller electronics can.

> Maybe there's a 'market' for something like that, i.e. a disk format
> where data is written redundantly on ten zones of the drive or
> something.  It really should be mostly a software problem.

It should be - but is it?  I don't know floppy controller chips enough
to say.

However, if you pick up any good book on coding theory you can read all
about adding ECC bits and then scattering data around so that (say) any
arbitrary ten sectors could be lost or corrupted without rendering any
data unrecoverable.  ("Lost" is an easier case; it's easier to correct
bits you know you don't have than bits you have but which you might
have wrong.)

For example, if you take (say) 13 sectors of 256 bytes each and think
of them as an array of 13 by 2048 bits, you can then apply ECC along
the 13-bit dimension, adding 5 error-correction bits, to get 18 sectors
of 256 bytes each, with the ability to correct any N sectors that are
wrong, for some value of N (which I'm not a good enough coding theorist
to know, but I'd guess is around 4 to 6).

Repeat for each track and you've lost 5/18 of your space but you've got
the ability to lose up to N sectors per track without losing any data.
More complex schemes can spread data out even more....

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse at rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B



More information about the cctalk mailing list