Magnetic Recording

In order to store information magnetically, we need a reliable way to translate data into a magnetic field and back again. The relationship between electricity and magnetism as described in the Introduction to Electromagnetism gives us just such a method.

Information Storage and Retrieval

Consider a horseshoe electromagnet, as shown in Figure 1. In the gap between the magnet's two poles, the magnetic field lines bend outward slightly, forming a fringe field. If a magnetic material passes against the horseshoe (or write head) beneath the gap, the fringe field will permeate the material and magnetize it in the direction of the field. An alternating current will thus write a pattern of magnetic bits on the material, where the domains are aligned first one way and then the other. The strength of the induced magnetism depends on the amount of current running through the electromagnet.

An inductive read head works in the opposite fashion. The tape is passed beneath the gap, and a change in direction of magnetization between domains induces a voltage in the coil. In the reverse operation of the write head, the strength of the current thus produced depends on the amount to which the magnetic bits were magnetized and on the rate at which the direction of magnetization changes.

Digital Recording

In a way, digital recording is simpler than audio recording because there are only two orientations of the magnetic domains. As a disk or tape passes under the write head, a direction reversal of the current (and resulting magnetic orientation) is a data "1," while no reversal is a data "0." Similarly, on inductive readback, a data "0" is corresponds to no change between magnetic bits, and no induced voltage (See Figure 2).

An alternative form of readback uses magneto-resistive (MR) materials: solids whose resistance changes when exposed to a changing magnetic field. In MR heads, a current is passed through a magneto-resistive film which is exposed to the storage medium. With a constant current supplied, the pulses of magnetic flux case a change in voltage across the film. This has led to a new, dual-head configuration: an inductive write-head and MR read-head.

Noise and interference are introduced into magnetic storage in a number of ways. First, even when a domain as a whole has been oriented a certain way, its component atomic magnets are still subject to a certain randomness. There can also be some residual magnetization left over from an earlier write, or leakage from adjacent tracks on the disk or tape. All of these factors make it difficult for the computer to judge whether or not a magnetic transition is present; whether it should translate a 0 or a 1.

And the picture becomes even more complicated. It was discovered that long strings of zeroes (no change in magnetic alignment) cause problems: the bit's "address" along the track (so that it can be accessed in the future) is obtained from a counter that counts the bits along with the data. If the counter does not receive fairly regular readback pulses (caused by magnetization transitions), the data and the bit address can become out of synch. This means that data cannot be written in its virgin form to the storage medium, but must first undergo channel encoding to eliminate long strings of 0's. During readback, the data must then be decoded.

Figure 3 shows one cycle of writing and reading data. The data from the computer first undergoes channel encoding as described above, and is then written to the medium. Once the data has been translated in a series of magnetized domains, it is subject to noise and interference; so the voltage in the read head is not as clear as the initial write current was. It is passed through a detector, which must decide whether a 1 or a 0 is present in each bit cell. Early detectors sampled a small window, and decided "1" if the voltage was above a certain threshhold, making them amplitude-sensitive. More modern detectors consider the derivative function of the voltage, which should cross 0 at a peak in voltage (maximum of the function). Today's detectors can also handle more complicated noise and interference effects, using advanced signal processing techniques. Finally, after a decision has been made by the detector, the data goes through channel decoding to be returned to its original form.