What is it?
A non-magnetic plate or platter forms the ‘disk’. This disk is coated with very fine iron oxide (commonly known as rust).
Because it is a ferromagnetic material (ferro means iron), it can be magnetised. A magnet has two states - North pole one way, or reversed. If the magnetised material is pointing one way it can
represent a binary ‘0’ and a binary ‘1’ if it is the other way. So a magnet can store a ‘bit’ of data. what we are looking for in today’s disks however is several
Gigabytes of data. As the disk is revolving, we can magnetise lots of little islands and as each island passes underneath the read/write head we can have a stream of bits.
A practical modern magnetic hard disk has a single platter. It has a great many tracks - concentric rings from near the
centre of the platter to near the outer edge. The faster the disk spins, the faster the disk can transfer bits of data with the read/write head - this governs the data
transfer rate. As there are a lot of tracks, the read/write head needs to be able to move radially between the innermost track and the outermost, although some disk
drives may have several heads. The head needs to be able to move rapidly from one track to another - this governs the the data latency time - the average time between
requesting a particular track and the head being placed over it.
To get more data bits onto the disk, the islands or magnets need to be smaller - and this can be done by arranging the
magnets to stand on their ends, perpendicular to the disk surface. Also, the head needs to be closer to the disk to get the resolution - 1 bit at a time.
The head doesn’t touch the disk surface but flies across it a bit like aquaplaning but with air. The surface of the disk needs to
be extremely smooth and the air in the drive needs to be extremely clean, else the head may contact the disk - which would cause a ‘head crash’.
As it is near-impossible to guarantee 100% good islands of magnetic material across the disk, disk drives split the tracks
into sectors and use at least two methods to get around this - redundancy in recording and avoiding sectors with too many errors. Redundancy is usually a few extra bits for each sector so that a cyclic
redundancy checksum can be used to identify that errors have occurred and to correct up to a certain number of errors. Apart from the extra bits used for error
correction there are additional bits used to identify sectors and tracks (known as soft formatting).
|
