4

Spinrite saved my ass many times. It refreshed the low level disk format (sector definitions), something that DOS couldn't do.

But on 80s TV shows, Steve said that it was necessary to rewrite the entire disk periodically, to "refresh" the fading magnetic domains. And it does appear that after time, ST220 disk sectors did used to just die.

But if that's true, how come it isn't necessary now?

Or was it ever?

Improve this question
11
  • 4
    Well it definitely won't be necessary on SSD drives.
    – Alan B
    Commented Jul 31 at 12:48
  • I'd like to make a Steve Gibson fan club. He did some amazing stuff in security.
    – Miss Understands
    Commented Jul 31 at 12:53
  • 7
    "Was it ever" is a retro question, but "why isn't it now" is not. @AlanB That's because it is likely that SSD controllers scan and refresh the fading memory cells themselves without user intervention like modern hard drived but the old hard drives didn't.
    – Justme
    Commented Jul 31 at 12:59
  • 1
    @JessFuckett Actually most of flash wearing is caused by erase process rather than write. You can always write individual bits low but erase happens in larger blocks. The charges do fade over years if left untouched, and flash wears out on read too - it's called "read disturb" and flash wear leveling algorithms need to handle that as well. So much for predictable life for SSDs. Yes, unrelated to magnetic media but somewhat similar.
    – Justme
    Commented Jul 31 at 14:05
  • 3
    For Gibson, the term 'spin doctor' is doubly appropriate.
    – dave
    Commented Jul 31 at 17:02

1 Answer 1

Reset to default
5

Old-style hard disks and floppy drives were formatted by using the "ordinary" write process to write synchronization and sector identification marks in a manner that was blind to whatever had been on the disk. During normal operation, to write a sector, the drive will listen for the synchronization mark that's followed by an identifier for the proper sector. It will then turn on the write head and blindly write a sector's worth of data.

Some utilities, of which I would expect SpinRite to be one, will read each track of the disk, reformat it, and then rewrite the track with the data it had just read. This would probably not be needed unless one wanted to change the sector interleave after getting a faster computer, but hard drive sectors did sometimes fail; it's unclear to how often failure was caused by gradual degradation and to what extent it was primarily caused by areas of the disk whose coating wasn't quite as good as it should have been, or an inferior region of coating which allowed magnetic domains to spread unusually fast. I would expect that on some, perhaps many, hard drives in the 1980s there would have been areas that would work more reliably if periodically refreshed, but there would also have been many hard drives that would not need such treatment. Today a sensible attitude toward storage media would be to regard any media that can't reliably hold magnetic domains indefinitely as meriting immediate replacement with media that can, but technology was nowhere near as refined in the 1980s as it has become in the decades since. Even if one replaced a hard drive that wasn't fully trustworthy, there would be practical no way of knowing whether the replacement would be any better.

Improve this answer
13
  • 1
    Older drives might have had some drift in the exact position of a track, resulting in offsets over time that would make it harder to read the bits. Newer drives use servo tracking to stick to the track precisely so drift doesn't happen.
    – Mark Ransom
    Commented Jul 31 at 16:31
  • 2
    Low-level format was real. There are three things you need to do to a freshly-coated magnetic disk: (1) record track and sector markings, (2) check for sectors that cannot reliably be written and re-read, (3) write an empty file system. In computer systems that don't confuse meanings, these are called formatting, bad-block check, and initialize, respectively. Formatting on site was once needed because removable disk packs were not necessarily sold for a particular geometry. With fixed disks, and especially with integrated HDAs, 'low level' formatting can be done at factory.
    – dave
    Commented Jul 31 at 17:00
  • 1
    @dave: For years, drives have often had tracking and synchronization data written at a physically depeer layer on the disk than user data, at lower density, in a way that allows it to be read independently of the user data. Equipment to record these markings needed to be extremely precisely calibrated, but a drive which could quickly move the head in and out in small increments could get by with much sloppier tolerances, since a drive could move somewhere near where it needed to go, read where it was, adjust its posiition, read where it was, and maybe perform a third adjustment if needed.
    – supercat
    Commented Jul 31 at 19:34
  • 1
    @dave: The magnetic field produced by the drive head when writing is so finely concentrated that it won't be able to affect magnetic flux states deeper within the media, and the drive will by nature adjust its behavior to accommodate any manufacturing variations in the mechanism. This approach reminds me a little bit of the way phosphor dots are put on a color CRT. The inside surface is coated with a resist layer, the screen is fed a signal to draw one color (eg. red), the parts of the resist layer that were hit with the beam are chemically etched, and then red phosphor is applied...
    – supercat
    Commented Jul 31 at 19:38
  • 1
    ...to the parts of the glass where the resist was removed. The process can then be repeated for green and blue. Even if the position of the shadow mask relative to the screen surface is only specified e.g. +/- 1mm, the phosphor dots will be perfectly aligned with the shadow mask, because their position will be adapted to match whatever the shadow mask position happens to be.
    – supercat
    Commented Jul 31 at 19:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .