Magnetic Recording Principles
Part of Data Storage
The physics of hysteresis, coercivity, and remanence that make it possible to encode data in a magnetized surface.
Why This Matters
Magnetic recording is the technology that made large-scale computing possible and practical. From the first magnetic drums of the 1940s through the hard drives of the 2010s, magnetic storage has been the dominant medium for bulk data storage for over half a century. Understanding the underlying physics lets you evaluate the feasibility of building or repairing magnetic storage devices, make informed choices about materials, and diagnose failure modes.
The principles are not exotic. Magnetism is a basic property of matter. Ferromagnetic materials — iron, nickel, cobalt, and their compounds — can be magnetized and retain that magnetization. This retention of magnetization is the basis of all magnetic recording. You write a bit by briefly magnetizing a tiny region of a ferromagnetic surface in one direction; you read the bit by detecting which direction that region is magnetized.
The challenges are engineering, not physics: making the magnetized regions small enough and stable enough, and building heads sensitive enough to read and fast enough to write.
Ferromagnetism and Magnetic Domains
Most materials are essentially non-magnetic: their atomic magnetic moments point in random directions and cancel out. Ferromagnetic materials are special: quantum mechanical exchange interactions cause neighboring atoms to align their magnetic moments, creating magnetic domains — regions where all atoms point the same way. A typical domain contains millions of atoms and is a few micrometers across.
In an unmagnetized ferromagnetic material, the domains point in different directions and their fields cancel. Apply an external magnetic field strong enough, and the domains rotate to align with the field. Remove the external field, and most domains stay roughly aligned — the material is now magnetized. This is called remanent magnetization or remanence.
For magnetic recording, you want a material with:
- High enough coercivity that random magnetic fields (nearby magnets, electrical noise, the earth’s own field) will not accidentally flip the recorded bits.
- Low enough coercivity that your write head — which produces only modest fields — can still reliably write new data.
- High remanence so that the stored magnetic field is large enough for the read head to detect reliably.
This balance is expressed in the B-H hysteresis curve: a graph of magnetic flux density (B) versus applied magnetic field strength (H). The area of the hysteresis loop indicates how much energy is dissipated per write cycle; the height of the loop (maximum B) indicates signal strength; the width (coercive field Hc) indicates how hard bits are to flip accidentally.
Magnetic Recording Media
Early magnetic recording media were particles of ferric oxide (Fe₂O₃, rust) mixed with a binder and coated onto a substrate. Ferric oxide particles have good coercivity and can be manufactured to controlled sizes. The substrate can be mylar film (for tape), aluminium alloy (for disks), or phenolic resin (for drums).
Particle size is critical. Each recorded bit must be a well-defined magnetic region. If the particles are too large, the minimum bit length is too long and storage density is low. If the particles are too small, thermal energy can randomly flip the magnetization of individual particles — a phenomenon called superparamagnetism that sets a fundamental limit on how small particles (and thus how dense recording) can be.
For practical rebuild purposes, oxide-coated magnetic recording tape can be salvaged from audio cassettes, video tapes, and computer backup tapes. The coating technology is similar across all of these — ferric oxide or chromium dioxide (CrO₂) particles in a polyurethane binder. A severely worn tape may have degraded signal; otherwise salvaged tape is fully functional for digital recording.
Cobalt-modified iron oxide and metal particle (MP) formulations offer higher coercivity and signal strength than plain ferric oxide, enabling higher recording density. These are found in higher-quality audio and video tapes.
The Read/Write Head
Data is written by passing the recording medium past an electromagnet (the write head). The head creates a strong, localized magnetic field that magnetizes a tiny region of the medium as it passes. By reversing the current in the head coil, you write a domain pointing in the opposite direction — this transition between magnetization directions represents a 1-bit in most recording codes.
The write head is a C-shaped core of soft ferromagnetic material with a coil of wire around it. The gap in the C forces the magnetic field to fringe outward into the recording medium. Gap width determines the minimum bit length; narrower gaps allow higher recording density but are harder to manufacture and align.
Reading is the reverse process. As magnetized domains on the medium pass under the read head, the changing flux induces a voltage in the read coil (Faraday’s law of induction). The transitions between opposite-direction domains produce voltage pulses; the absence of transitions produces little or no signal.
Modern drives use separate read and write heads optimized for their respective tasks. The write head must produce a large, sharp-edged field to reliably switch domains. The read head must be sensitive to tiny fields from very small magnetized regions. Magnetoresistive (MR) read heads — whose electrical resistance changes with magnetic field — are far more sensitive than inductive read heads and enabled the dramatic density increases of the 1990s and 2000s.
Recording Codes
A simple approach to magnetic recording would be: 1-bit = magnetized north, 0-bit = magnetized south. This works but has a severe problem: a long string of zeros or ones produces no signal transitions, making it impossible to maintain timing synchronization.
Non-return-to-zero (NRZ) recording maintains the current magnetization direction for 0-bits and reverses it for 1-bits. A long sequence of identical bits produces a constant field with no transitions — the read circuit gradually drifts in timing.
Frequency modulation (FM) recording addresses this by adding a clock transition at the beginning of every bit period. A 1-bit has a clock transition plus a data transition; a 0-bit has only a clock transition. This guarantees at least one transition per bit period, enabling continuous synchronization. The cost is that every bit uses two potential transition slots, halving maximum density compared to NRZ.
Modified FM (MFM) improves on FM by omitting the clock transition between two consecutive 1-bits (since those bits’ data transitions serve the synchronization purpose). This roughly doubles density compared to FM and was the standard for floppy disks and early hard disks.
Run-length limited (RLL) codes constrain how many consecutive bit cells can have no transition, and were used in high-density disk recording from the 1980s onward.
For a scratch-built magnetic recording system, FM or MFM recording codes are practical to implement in discrete logic. RLL requires more complex encoding/decoding circuitry.
Noise, Errors, and Margins
Magnetic recording is analog at the physical level — the head reads a continuous voltage signal — but it is interpreted digitally. The read channel must decide, for each bit period, whether the signal indicates a transition (1) or no transition (0). Noise in the signal can cause wrong decisions.
Sources of noise include:
- Medium noise: non-uniformity in the coating, particle clumping, scratches
- Electronic noise: thermal noise in the read head and amplifier
- External interference: vibration, temperature variations, nearby magnets
Signal-to-noise ratio (SNR) determines the reliability of the reading. Higher SNR allows faster reading, narrower tracks, or smaller bit cells — all of which increase storage density. Lower SNR requires wider margins, reducing achievable density.
Error rates in well-designed systems are extremely low — one error per billion to trillion bits. But even this low rate matters over large volumes of data. A 1 TB drive reading at 100 MB/s reads its entire capacity in about 2.5 hours; even a one-per-trillion error rate produces one error during a full-drive read. This is why error correction codes are always applied on top of the magnetic recording layer.
Constructing a Functional Head
For experimenters and rebuilders, constructing a functional magnetic recording head from scratch is feasible but demanding. The requirements:
Core material: Soft ferrite (nickel-zinc ferrite) or laminated silicon steel. The material must have high permeability (for good field concentration) and low coercivity (to avoid the core retaining fields from previous writes, which would corrupt subsequent reads).
Gap width: For low-density recording (thousands of bits per centimeter rather than millions), a gap of 0.1–1 mm is workable. This can be achieved by grinding and lapping core halves to precise flatness and inserting a non-magnetic spacer (glass, aluminium foil) of the desired thickness.
Coil: 50–500 turns of fine insulated wire wound around the core. More turns increase both write field strength and read sensitivity but add inductance, limiting high-frequency response.
Impedance matching: The read head coil produces very small voltages (microvolts to millivolts). An operational amplifier or transformer preamplifier close to the head is essential to amplify the signal before it is affected by cable noise.
A functional test setup for verifying a head: attach the coil leads to an oscilloscope, pass a magnetized piece of audio tape under the gap, and observe the voltage pulses. If you can see clean pulses as the tape transitions pass the gap, the head geometry is correct.