How is HDD data stored?

Hard disk drives (HDDs) have been the predominant form of high-capacity, non-volatile data storage in computers for decades. But how exactly is all that data stored on the platters inside an HDD? Here we’ll look at the key components involved and how they work together to allow data to be written, read back, and preserved.

Magnetic Storage Fundamentals

At its core, an HDD relies on magnetism to store and access data. More specifically, it leverages the ability of magnetic materials to have their magnetization altered by an external magnetic field. HDDs use platters coated with a thin film of magnetic material, usually an alloy containing nickel, cobalt, and/or iron.

These magnetic platters can have their magnetization changed very precisely at microscopic scales, allowing each tiny region to represent either a 1 or 0 digitally. Patterns of magnetization are written to and read back from the platters by electromagnets known as read/write heads.

Magnetic Domains

The magnetic film that coats each platter is made up of tiny magnetic domains about 10-100 nanometers across. Each domain has a magnetic field pointing in a certain direction, determined by the material’s properties. Without any external magnetic fields, the domains’ magnetization directions are randomly oriented, so the platter’s overall field is canceled out.

But when an external field is applied by a read/write head, the domains realign so their individual magnetic fields point in the same direction, combining to create a detectable overall field. This process of realignment forms the basis of writing data to the platter.

Reading and Writing Data

For reading and writing data, the read/write heads are positioned just above the surface of the platter. Electromagnets within the heads generate magnetic fields that penetrate into the magnetic material coating on the platter. This allows data to be written by reorienting the alignment of magnetic domains.

For writing, the external field from the head orients the domains to represent either a 1 or 0 bit value depending on the direction. Reading is achieved by detecting the orientation of the domains – the head senses either the presence or absence of a magnetic field from the domains, which the drive interprets as 1s or 0s.

Magnetoresistive Heads

Early HDDs used inductive read/write heads consisting of a tiny electromagnet coil wrapped around a core. They operated by inductively sensing the changes in the magnetic fields emanating from the platter. But modern drives use magnetoresistive (MR) heads, which offer better sensitivity when reading data.

MR heads feature materials whose electrical resistance changes based on the presence and direction of an external magnetic field. This magnetoresistance effect enables detecting very small magnetic field strengths from the disk surface. Materials exhibiting higher magnetoresistance ratios enable more sensitive data reading capabilities.

Data Organization

With an understanding of the magnetic properties that enable data storage, next we’ll look at how the platters are organized and formatted to actually store file data in an ordered and efficient way. This largely involves dividing and arranging the platter surfaces into discrete tracks and sectors.

Tracks and Sectors

Each platter surface is radially divided into concentric tracks, thousands of them per platter. The tracks can be visualized as rings of data spanning from the inner diameter to the outer edge of each circular platter.

The tracks are further broken down into sectors. Sectors represent distinct storage blocks that serve as the smallest individually addressable unit for data storage and retrieval. A sector typically holds 512 bytes of user data.

Thousands of sectors make up a track. Multiplying all the tracks and sectors together determines the overall storage capacity of the drive. So data storage density increases with more, narrower tracks and smaller sectors.

Cylinders

Another organizational concept is the cylinder. This represents the collection of tracks vertically aligned across the stack of platters. If you think of the platters as a stack of CDs, the tracks would align vertically like the walls of a cylinder. The read/write heads are physically connected together to move in unison from cylinder to cylinder.

Data is written sequentially by cylinder rather than by individual platter. This enables concurrent writes or reads across the platters with a single seek motion instead of individual head movements, improving throughput.

Low-Level vs High-Level Formatting

Before an HDD’s platters and sectors can efficiently store files and folders as we normally think of computer data storage, some formatting needs to take place. Formatting prepares the disk surfaces by creating tracks and sectors.

Low-level formatting handles the division of platters into tracks and sectors. It sets up all the tracking information needed to delineate tracks and sectors for operating system use. High-level formatting then writes the file system structures needed to organize user data.

File allocation tables (FATs) keep track of which sectors belong to which clusters and files. The sectors themselves also receive metadata tags identifying their order and location details. Directory structures are also written to define file paths and hierarchy.

Recording Methods

We’ve looked at the foundational magnetic properties as well as the logical disk organization. Now let’s examine some of the approaches employed to actually record the data onto the platter surfaces, which have progressed over HDD history as recording densities increased.

Longitudinal Recording

The first HDD recording method was longitudinal or vertical recording, used from the 1950s through around 2005. As its name implies, it involved orienting the magnetization along the horizontal plane of the platter.

A perpendicular write head polarizes small regions of the magnetic material to have their magnetic field point either left or right horizontally. These left or right-pointing fields represent 1s and 0s for storing data. The narrower each magnetic region can be made, the higher the areal density of data storage.

Perpendicular Recording

But longitudinal recording reaches a density limit as the regions become too small. Too much magnetic instability and interference occurs between closely packed horizontal magnetic fields. So in 2005, the HDD industry shifted to perpendicular or vertical recording.

Instead of aligning magnetization horizontally along the platter surface, perpendicular recording orients the magnetic fields vertically through the thickness of the magnetic film. This allows for closer packing of vertical regions with less cross-talk effects. Storage densities of over 1 Tb/in2 are possible using perpendicular recording.

Shingled Magnetic Recording (SMR)

A newer approach called shingled magnetic recording (SMR) takes this vertical storage a step further for even greater densities. It partially overlaps tracks in an overlapping shingle-like pattern to squeeze more tracks onto the platter surface.

But this overlap comes with a drawback. To modify a shingled track requires also rewriting the overlapping adjacent tracks, increasing write complexity. SMR drives are thus better suited for sequential write workloads like archival data storage rather than random rewrites.

Two-Dimensional Magnetic Recording (TDMR)

On the horizon is two-dimensional magnetic recording (TDMR). As its name indicates, it aims to push magnetic storage densities by recording data in two dimensions instead of just vertically. This will allow narrowing of both track width and reader element dimensions at the same time for density multiplication.

TDMR leverages advances in signal processing algorithms along with nanoscale head engineering. Proof-of-concept products are in development but widespread implementation is still years away.

The Head Disk Assembly

Now that we’ve looked at the platter recording methods, let’s zoom out and see how the platters fit together with the read/write heads in the complete head disk assembly (HDA). This is the core mechanical hardware stack inside the HDD housing.

HDA Components

The HDA refers to both the rigid base structure that holds all the internal components as well as the components themselves as an integrated unit. Key elements include:

  • Spindle motor – Rotates the disk stack
  • Rotational actuator – Holds and moves the read/write head arm
  • Arm positioning system – Controls head motion
  • Platters – Stacked disks that store data
  • Read/write heads – Electromagnets for data recording

These primary HDA parts all work together to enable data access from anywhere across the platters. The actuator rapidly moves the head arm from track to track while the spindle motor spins the platters below.

Head Actuation

There are a couple different mechanisms used over the years to position the read/write heads.

Stepper motor actuation – Early HDDs moved heads using a stepper motor, which incrementally rotates a set angle per step. Control firmware tells the motor how many steps to move to transition between tracks.

Voice coil actuation – Most modern drives use a voice coil actuator. This works like a speaker coil interacting with a fixed magnet. A feedback loop enables precise positioning down to the nanometer scale.

Head Assembly

The read/write heads themselves are integrated into a singular head assembly which is attached to the end of the movable actuator arm. The assembly consists of:

  • Slider – Aerodynamic base that glides over the platter surface
  • Suspension – Holds the slider and applies pressure to maintain minimal head-media spacing
  • Read/write elements – Embedded electromagnets for data recording

Advanced engineering enables the head assembly to fly incredibly close over the high speed spinning platter – often just nanometers apart.

Data Encoding Schemes

The last major aspect of HDD data storage to cover is how the binary 1s and 0s are encoded into magnetic signals. Different schemes have been used over the decades.

Frequency Modulation (FM)

One of the earliest methods was frequency modulation (FM) encoding introduced in the 1970s. It represented 1s and 0s by magnetizing sectors with high and low frequencies.

A higher frequency translated to more magnetic flux changes per inch. But this scheme was susceptible to external magnetic interference.

Modified Frequency Modulation (MFM)

Modified frequency modulation (MFM) improved upon FM encoding. It marked sector boundaries with missing clock transitions to differentiate from data signals. MFM also inserted gaps between sector data regions.

These changes provided more distinct 1 and 0 MFM waveforms that were less prone to inter-symbol interference.

Run Length Limited (RLL) Encoding

As HDD densities increased into the 1980s, more efficient schemes were needed. Run length limited (RLL) encoding became popular. RLL formats data into coded sequences with constraints on the minimum and maximum number of consecutive 0s between 1s.

Different RLL schemes are denoted like RLL (d,k) with the d and k numbers referring to the constraints – for example RLL 2,7. RLL encoding provides superior data density compared to FM or MFM.

Advanced High-Speed Coding Schemes

Higher-speed serial ATA HDD interfaces drove the development of more complex encoding algorithms in the 2000s. Low density parity check (LDPC) and iterative coding schemes are now used, such as two-dimensional multilevel coding (TDM).

These modern techniques approach the Shannon limit for the media while enabling serial transfer speeds. They build in redundancy to counter inter-symbol interference effects at high densities.

Conclusion

From the magnetic properties of domain orientations to platters’ logical formatting and head actuation designs, we’ve covered the key facets of how HDDs read and write data. Advancements in recording methods, components, and signal processing continue pushing hard disk storage densities higher to satisfy the world’s exponentially growing data storage demands.