14.1 How Hard Disks Work
All hard disks are constructed similarly.
A central spindle supports
one or more
platters, which are thin, flat, circular
objects made of metal or glass, substances chosen because they are
rigid and do not expand and contract much as the temperature changes.
Each platter has two
surfaces, and each surface is coated with a
magnetic medium. Most drives have multiple platters mounted
concentrically on the spindle, like layers of a cake. The central
spindle rotates at several thousand revolutions per minute, rotating
the platters in tandem with it.
A small gap separates each platter from its neighbors, which allows a
read-write head mounted on an actuator
arm
to fit between the platters. Each surface has its own read-write
head, and those heads "float" on
the cushion of air caused by the Bernoulli Effect that results from
the rapid rotation of the platter. When a disk is rotating, the heads
fly above the surfaces at a distance of only millionths of an inch.
The head actuator assembly
resembles a comb with its teeth
inserted between the platters, and moves all of the heads in tandem
radially toward or away from the center of rotation.
|
Platters are cheaper than heads. That means some drives have an odd
number of heads, leaving one surface unused. For example, the Maxtor
D540X uses 20 GB/surface technology and is available in 20, 40, 60,
80, 120, and 160 GB models. The 20 and 40 GB models use one platter
with one and two heads, respectively. The 60 and 80 GB models use two
platters with three and four heads, respectively. The 120 GB model
uses three platters with six heads, and the 160 GB model four
platters with eight heads. Maxtor could have produced 100 GB (three
platter/five head) and 140 GB (four platter/seven head) models, but
apparently decided there was no market for such intermediate sizes.
In theory, using an odd number of heads allows using platters with
one defective surface, but we don't know if drive
makers do that. Probably not. It'd cost more than
just pitching platters with a defective surface.
|
|
The small
separation between the heads and surfaces means that a tiny dust
particle could cause a catastrophic head crash, so these components
are sealed within a head/disk
assembly , or HDA. The sealed HDA contains air
filters that allow air pressure to equalize between the HDA and the
surrounding environment. Opening an HDA other than in a factory
clean-room is a certain way to destroy a disk drive.
Each surface is divided into concentric
tracks that can be read from or written to by
that surface's head. Each surface on a modern disk
drive contains thousands of tracks. Each track is divided into many
sectors, each of which stores 512 bytes of data.
Old drives used the same number of sectors on every track, typically
17 or 26. Modern drives take advantage of the fact that tracks near
the outer edge of the platter are longer than those near the center
by storing more sectors on the outer tracks. All tracks
that are immediately above and below each other form a
cylinder. If a drive has eight surfaces, each
with 16,383 tracks, that drive contains 16,383 cylinders, with eight
tracks per cylinder. The concept of cylinders is important because it
determines how data is written to and read from the drive. When a
drive writes a file that is larger than one track, it fills the
current track and then writes the remainder of the file sequentially
to the next available track within that cylinder. Only if the
capacity of the current cylinder is exceeded does the drive move the
heads to the next available cylinder. The drive writes data in this
fashion because selecting a different read-write head is an
electronic operation that occurs quickly, while moving the heads to a
different track is a mechanical operation that requires significantly
more time. The heads write data to the surfaces in
exactly the same way that data is written to a floppy disk or
magnetic tape. Each track contains myriad discrete positions, called
magnetic domains,
that can each store a single bit of information as a binary 0 or 1.
When writing, the head exercises a magnetic flux to alter the state
of a domain to a 0 or 1, as appropriate. When reading, the head
simply determines the existing state of a domain.
Because they reside in such close proximity, it is nontrivial
for a head to locate the correct track and sector. Early drives used
a stepper-motor assembly similar to that still used on floppy drives.
A stepper motor simply moves the heads to where the track is supposed
to be, without reference to its actual location. On stepper-motor
drives, thermal expansion and contraction gradually cause the
expected locations of tracks to drift out of alignment with their
actual physical location, which required frequent low-level
formatting of the drives to return them to proper alignment.
Stepper-motor hard drives were last produced in about 1990.
Later hard disk drives used a voice-coil actuator mechanism in
conjunction with a dedicated servo surface. For example, a drive that
had eight surfaces used only seven of these to store data, and
dedicated the eighth surface to servo information that helped locate
the correct track. A voice-coil drive does not seek to an absolute
track position. Instead, the head actuator assembly seeks to the
approximate position where it expects the track to be located. The
servo head then fine-tunes the positioning by locating the servo
track that corresponds to the desired track. Because all tracks in a
cylinder must necessarily be aligned, locating the correct servo
track automatically also locates the correct data cylinder. Early
voice-coil drives were effective and not subject to thermal drift,
but designers hated wasting an entire surface and head on servo data.
All current drives use embedded servo
information, which means that no surface is dedicated to servo
information. Instead, servo data is interspersed with user data on
normal data tracks, which allows every surface to be used to store
data. The hard disk drive connects to the PC via a
controller interface. Early hard disk drives used a separate
controller card that installed in an expansion slot and connected to
the drive via ribbon cables. All modern disk drives, IDE and SCSI,
have the controller embedded in the drive itself. A ribbon cable
connects the drive to a connector located on the system board or to
an expansion card that provides a connection point. Hard disk
interface cards are not actually disk controllers, and are properly
referred to as host
adapters. They do not contain disk controller
circuitry, but simply provide a connection point between the system
bus and the disk controller embedded in the drive. The
disk controller serves as an intermediary between the system and the
hard drive. When the system needs to read data from or write data to
the drive, it issues commands to the controller, which translates
those commands into a form understandable by the drive. The drive
then supplies data to the controller during read operations, and
accepts data from the controller during writes.
|