DR. J. G. WOODWARD - RCA Laboratories, Princeton, N. J.

DR. J. G. WOODWARD received the BA from North Central College in 1936, the MS in Physics from Michigan State College in 1938, and the PhD in Physics from the Ohio State University in 1942. He held teaching assistantships in Physics from 1936 to 1942 while in graduate school at Michigan State and Ohio State. In March 1942 he joined the research department of RCA in Camden, New Jersey, and later that year moved to the newly-formed RCA Laboratories in Princeton, New Jersey.

His research has been in a variety of fields including rhe study of vehicular radio noise, underwater sound, ferroelectricity in barium tita-nate, electromechanical feedback devices, Theological measurements at audio frequencies, musical acoustics, sound-reinforcement systems, stereophonic sound reproduction, magnetic-tape recording and disk-phonograph recording. He currently holds the position of Head, Audio Recording Group in the Acoustical and Electromechanical Laboratory. Dr. Woodward is a Fellow of the Aoouslital Society of America, the Audio Engineering Society, and the American Association for the Advancement of Science. He is a member of Sigma Xi.

The performance of magnetic tape recording systems is limited by various parameters of the tape coating and the read-write heads. This paper presents a qualitative description of the mechanisms involved in these performance limitations, and outlines the direction future developments must take to ensure further improvement in the performance of digital-data magnetic-tape recording systems.

The hysteretic magnetic behavior of ferromagnetic materials is well-known. This behavior is described by the familiar diagram of Fig. 1. If a ferromagnetic material is subjected to a magnetic field H, which is sufficiently strong to drive the material into magnetic saturation, the material retains a residual magnetism called the saturation remanent induction B, after the magnetizing field is removed.

Similarly, when a field -H, of the opposite polarity is applied and removed, the material retains a residual magnetism -B. Thus, a ferromagnetic material possesses two distinct and reproducible states of remanent induction. This property makes these materials natural candidates for use in digital data-storage systems, where the +B state may represent a 0 and the -B state a 1. Indeed, magnetic storage devices have become indispensable elements in digital systems. These storage devices have taken the form of a variety of types of core memories and also of several types of moving-medium recording systems such as magnetic drums, disks, and tapes.

Each of these magnetic storage devices has its particular set of advantages and shortcomings and, hence, performs most effectively in a certain type of application in a digital system.

Magnetic-tape recording offers the advantage of providing an almost-unrivalled density of stored information when measured in terms of the number of bits per unit volume of the storage medium. This capability, combined with the reliability and convenience of use which are now possible, and with the fact that the data recorded on a magnetic tape can be erased completely or changed bit by bit as desired explains the popularity of this form of storage.

Despite these advantages, a continued search is underway for improved materials, components and techniques for increasing the recorded density of information which can be reliably handled by a magnetic tape-recording system. A number of more or less sophisticated circuits and techniques have been devised for use in conjunction with the basic tape-recording elements to increase significantly the capabilities of the system.

Nevertheless, certain fundamental limitations in performance remain which are inherent in the magnetic and mechanical relationship between the tape and the recording and playback heads which write and read out the signal.

It is the purpose of this discussion to outline the basic head-tape relationships in order to show qualitatively the types of limitation existing in present-day magnetic-recording equipment and to indicate the direction which developments must take to further increase the density of recorded information.

In modern commercial digital magnetic-tape recording systems the recording medium consists of a smooth, flexible plastic tape usually between lO^-3 and 2 x 10^-3 inch thick, on one surface of which a coating of magnetic material has been deposited.

The thickness of the coating is commonly between 0.1 x 10^-3 and 0.5 x 10^-3 inch. A high-precision mechanical transport pulls the tape at a nearly constant speed over a recording head, with the magnetic coating of the tape close to or in contact with the pole faces of the head.

The head consists of a magnetic core on which is wound a coil carrying a magnetizing current. A fringing magnetic field extends outside the core in the region of a non-magnetic gap in the core, and the tape coating becomes magnetized as it passes through this fringing field.

The recording situation is shown diagrammatically in a sectional view in Fig. 2. The polarity of the magnetization in the tape depends, of course, on the polarity of the currents in the coil of the recording head, and the tape may be left in either of its two remanent states corresponding to forward or reverse directions of current flow. If the direction of current flow is reversed at various times as the tapes move past the head, a corresponding pattern of remanent magnetization will be left in the tape coating, as sketched in Fig. 3. If the tape travels a considerable distance between the times of current reversal, the boundaries between regions of opposite polarity in the tape are clearly defined when measured in terms of the distance of tape travel per unit time.

However, if the switching rate is increased without a corresponding increase in tape speed the boundary between adjacent regions of opposite polarity becomes less and less definite until, ultimately, a small region surrounded by oppositely-magnetized regions can no longer be resolved. At the same time phase shifts may occur which will introduce timing errors in pulses reproduced from the tape in playback. In order to describe these effects we will now consider in more detail the magnetic field in the neighborhood of the recording gap and the way in which the field configuration influences the recording process.

The magnetic field lines in the fringing field associated with the non-magnetic gap in the recording head follow almost semi-circular paths between the portions of the core on opposite sides of the gap.

Only in the region very close to the gap do the field lines deviate greatly from semi-circular paths. Such a fringing field is sketched in a cross-sectional view in Fig. 4. In general, at any point above the pole pieces the fringing field will have both a longitudinal component (parallel to the tape motion) and a perpendicular component (perpendicular to the surface of the tape).

The longitudinal component is almost entirely responsible for the magnetization on the tape, while the vertical component usually produces only second-order effects. This is the case because the longitudinal component is the stronger of the two in the region of the fringing field through which most of the tape coating passes, and also because the coating is given a preferred magnetic axis in the longitudinal direction during manufacture of the tape. Consequently, only the longitudinal components of the field will be considered here.

The results of a calculation of the longitudinal field component around the gap are plotted in Fig. 5. The ordinate gives values of the field strength in terms of H/Ho, where Ho is the field strength inside the gap where the field is uniform.

The abscissa is the distance in the direction of tape travel away from the central plane of the gap, measured in units of x/l where "l" is the gap length. The various curves plotted correspond to various distances above the surface of the pole faces, measured in units of y/l.

The plots of Fig. 5 show the field in air. The configuration of the field is altered only slightly by the presence of magnetic tape on the head, since the permeability of the tape is quite small (u ~ 4.

Gap lengths between 0.1 x10^-3 and 10^-3 inch are used in commercial
equipment. Gap lengths as short as 0.02 x 10^-3 inch are sometimes used in experimental equipment.

Since the thickness of the tape coating is comparable to the gap length, there will be a considerable difference in the mag-netizing-field strength at various depths in the coating as it passes over the recording gap.

The separation between the head and the tape coating is also a factor in determining the configuration of the field through which the tape passes. In some applications the head is intentionally spaced from the moving coating by about 0.1 x 10^-3 inch to reduce wear. Even when "in-contact" operation is intended, an effective separation exists due to roughness of the tape surface. This effective separation has been estimated to range between 0.02 x 10^-3 and 0.1 x 10^-3 inch, depending on the condition of the tape surface.

With this background, let us now consider the magnetization process in the tape as it passes the recording gap.

Another way of depicting the magnetic hysteresis of the tape coating will be useful in this discussion. As was shown in Fig. 1, when the material is saturated by a sufficiently-strong applied field and the field is then reduced to zero, the material is left in a remanent state, Bx. If the material is saturated, and subsequently a field of the opposite polarity and of any magnitude is applied and reduced to zero, the material will be left in some other remanent state. If the remanent induction is measured for each of a number of values of applied field, a hysteresis loop such as is shown in Fig. 6 results, in which H is the value of the applied field and By is the remanent induction resulting from each value of H.

For qualitative purposes the loop of Fig. 6 may be thought of in terms of the parallelogram completed by the dashed-line extensions at the corners of the loop. Thus, Hs is defined, as before, as the field which results in the maximum remanence Bs. A minimum field Ha is required before any change in remanence can be observed.

The magnetic characteristics described by Fig. 6 can now be combined with the gap field described by Fig. 5 to predict the pattern of magnetization impressed on the tape when the current is reversed in the recording head.

As an illustrative example let us take a case in which the gap and tape dimensions are such that the front surface of the tape coating will be spaced by 0.1 l and the back surface by 0.75 l from the recording-head surface.

In order to relate the H/Ho ordinate of Fig. 5 to the H abscissa of Fig. 6 we must take into account the necessity of saturating the tape coating throughout its entire thickness as it passes the recording gap. This is an important requirement for digital recording systems in which the signs of individual bits must be reversed. If a saturating field did not extend completely through the tape, variations in head-to-tape spacing or in instantaneous recording current might cause an incomplete removal of previously-recorded information, with the result that the signal-to-noise ratio would be degraded, and the probability of an error in read-out would be increased. Therefore, as a minimum requirement in the present example, the field strength above the center of the gap at a distance of 0.75 l above the head should be Hs. The value of Ha, below which the magnetization in the tape is not affected, will be determined by Fig. 6 and can be located on the ordinate of Fig. 5. Horizontal dashed lines indicating the levels of Hs/H0 and H/H0 have been drawn in Fig. 5. The intersections of these lines with the family of curves determines the region in which recording takes place at the various distances above the head.

When a steady current of the required magnitude flows in the head, all of the tape which has passed the gap will be saturated in one polarity. If the direction of the current is reversed, all portions of the tape in regions where the field is greater than Hs at that instant or which subsequently move through these regions will be saturated in reverse polarity. All portions of the tape which do not encounter fields greater than Ha will retain the original polarity of saturation. At the instant of current reversal some portions of the tape will be in regions where the field has values between Ha and Hs. These portions will be left with values of remanent induction between Bs and -Bs, in accordance with the characteristic shown in Fig. 6. The resulting pattern of magnetization on the tape is depicted in Fig. 7. Instead of a sharp plane of demarcation between regions of positive and negative saturation in the tape there is a sizeable region of transition, indicated by the shaded area in Fig. 7. Within this region the remanent induction decreases gradually from Bg to zero to - Bs. The locus of the points of zero field is indicated by the dotted line. The exact shape of the transition region is strongly dependent on the absolute field strength, the gap and tape dimensions, and the magnetic characteristics of the tape. The important fact to note is that a transition region exists having dimensions comparable to the tape thickness and the gap length, and that the region is shortest at the front surface of the coating and may become very extensive at the more remote surface. The transition region lies, for the most part, beyond the trailing edge of the recording gap.

The manner in which the transition region is a factor limiting the recording resolution (or the density of recorded information) becomes apparent when the qualitative analysis just made for a step function in the recording current is extended to the case of a short-duration pulse.

In this case, after a short interval following a current reversal, a second reversal returns the current to its initial polarity, and two transition regions are left in the tape as sketched in Fig. 8.

Since the transition regions overlap, a portion of the first region is overlaid and reversed in polarity by the second region. Consequently, the volume of the coating left saturated by the first reversal is smaller than would be the case if the transition regions were of infinitesimal extent, and the signal representing the recorded pulse will be reduced in magnitude in playback.

If the pulse duration is made so short that the reduced magnitude of the signal in playback is close to the noise level, a readout error may occur. Although it is not immediately evident, and the mechanism will not be considered further here, it can be shown that the alteration of the magnetization pattern in the tape due to recording short-duration, closely-spaced pulses results in timing errors in readout as well as the errors due to the reduced magnitude.

In some applications the timing errors may be more serious than the errors arising from the reduced magnitude of the pulses. On the basis of the preceding discussion, one can suggest certain directions in which developments should move to reduce the limitations on the attainable information density in the recording process. The obvious moves are to make the gap length shorter, to make the tape-coating surface smoother and thereby provide a more intimate tape-head contact, to make the tape coating thinner so all portions of the coating will traverse the recording field at nearly the same distance from the head surface, and to provide a recording medium having a BR-\s-H characteristic which has steeper sides (a square-loop characteristic).

When the coating is made thinner, a material having a higher Bs should be used to maintain a sufficiently high output with the reduced volume of material. Also, with the thinner coating a more wear-resistant material will be required. Recent studies2 have suggested that while a medium having a square-loop characteristic may be desirable for reducing errors due to a reduced magnitude of pulses in playback such a characteristic may not be best for reducing timing errors, and that in some applications a compromise may be required. This point deserves further experimental and analytical investigation.

In addition to the limitations imposed by the recording process, other, and comparable, limitations are imposed by the head-tape relationships existing in the playback process.

In order to isolate the playback from the recording effects we will suppose the magnetic recording medium to contain an ideal pattern of magnetization, i.e., to be uniformly magnetized through the thickness of the coating and to have regions of opposite polarity separated by infinitesimally-short transition regions in planes perpendicular to the tape surface and to the tape motion.

This is the situation depicted in Fig. 3. The limitations inherent in playback are primarily geometric in their origin and are not critically affected by the magnetic characteristics of the recording medium.

Three such resolution-limiting factors exist. Their effects have been

termed scanning loss,
separation loss, and
azimuth loss.

Since these factors have been known for many years and have been described in many places in the technical literature, they will be discussed only briefly here.

By way of background, the basic playback process will first be described with reference to Fig. 9, which depicts an ideally-magnetized tape moving over the pole pieces and the gap in the playback head. Field lines leaving the tape in the vicinity of the gap enter the core on one side of the gap and follow paths through the core to re-enter the tape on the other side of the gap.

As the tape moves past the head, the varying pattern of magnetization in the tape causes the number of field lines passing through the core to change in magnitude and polarity with time with the result that a time-varying "emf" is generated in the coils wound on core. The "emf" is proportional to the time-rate-of-change of the magnetic field, so that ideally the voltage at the terminals of the playback head is proportional to the derivative of the remanent magnetic induction in the tape as it passes the head.

Hence, a rectangular pulse recorded on the tape produces a positive-going spike and a negative-going spike in the readout signal. If the recorded pulses have transition regions between segments of opposite polarity the readout pulses, instead of being ideal spikes, take the form of positive and negative pulses of considerable duration and of smaller magnitude. A corresponding loss in resolution occurs. This, however, is a manifestation of the limitation in the recording process already discussed. The first playback limitation to be considered will be the scanning loss.

Scanning losses occur when the gap length of the readout head is comparable to the length of the pulses recorded on the tape. Clearly, if a train of pulses has been recorded such that the gap length is equal in extent to two adjacent pulses of opposite polarity the net number of field lines threading the core is zero for any position of the tape along the head, and no electrical signal is developed.

Even before this extreme condition is reached the magnitude of the readout signal is reduced by an amount depending on the ratio of the pulse length to the gap length. The mathematical analysis of scanning losses has been carried out for recorded sine waves with the result shown in Fig. 10. Here the abscissa is the number of wavelengths per inch on the tape and the ordinate is the scanning loss measured in decibels.

Curves for several lengths of playback gap are shown. The curves of Fig. 10 represent the playback response to a recorded sinusoidal magnetization pattern. They can be related to recorded rectangular pulses by considering the number of Fourier components required to give an adequate synthesis of the pulses. Between five and ten Fourier components usually must be passed without serious attenuation if signal deterioration due to scanning loss is to be negligible. This, clearly, requires a short gap length in a high-resolution recording system.

In the playback process, as in the recording process, a reduction in resolution occurs if the magnetic coating is not in intimate contact with the head. Theoretical values of the separation loss as a function of the number of wavelengths per inch of tape, with separation as a parameter, are plotted in Fig. 11 for recorded sine waves.

As in the case of scanning loss, these data can be applied to the case of recorded pulses in terms of the number of Fourier components which must be passed without attenuation to adequately represent the pulses. The separation loss is a function of the ratio of the head-to-tape spacing and the wavelength and amounts to approximately 55db loss per wavelength of separation.

Since the thickness of the tape coating may not be negligible in comparison to the recorded pulse lengths, the separation loss will be different for distances from the head surface into the coating, and short-wavelength components will suffer more attenuation than the long-wavelength components.

If the separation loss in playback is to be minimized, the coating surface should be very smooth to permit an intimate head-tape contact, and the coating should be thin to avoid serious relative loss of the short-wavelength components.

A recorded track on a tape always has an appreciable width in order to provide a reasonable signal-to-noise ratio and to make the head-track alignment feasible mechanically. Because of the finite track width it is highly desirable that the azimuth angle, i.e., the angle which the gap edges make with the direction of tape travel at each instant be precisely controlled.

If the tape behaved in exactly the same manner in every passage over the recording and playback heads, azimuth angle variations would not present a problem.

However, even in the best tape transports some variability in azimuth angle inevitably occurs. This is due to tolerances in the dimensions of the tape and of the tape-guidance system, and to lateral curvature or skewing of the tape.

Consequently, the azimuth angle in playback often will be somewhat different from the angle at the instant of recording. Hence, the gap in the head will be reading different phases of the recorded signal from one edge of the track across to the other edge, with the result that a partial cancellation of the signal occurs in the head. This leads to losses of the type plotted in Fig. 12 where the loss in decibels is shown on the ordinate and the number of recorded sinusoidal wavelengths per inch along the tape is shown on the abscissa.

The parameter for the four curves is the product of track width and the difference in azimuth angle in recording and playback. These curves can be applied to pulse recording by again considering the Fourier components. For high-density recording systems the tape-guidance system must be made to control the tape motion very precisely, indeed.

Azimuth angle variability also produces another type of error, since in digital-recording systems a number of parallel tracks are recorded on the tape. When the azimuth angle is different in recording and playback, timing errors occur between tracks. That is, some tracks may be advanced or delayed relative to other tracks as the azimuth angle varies, and signals from the various tracks which should be read out simultaneously are read at different instants. Attempts to overcome this problem have led to the development of a number of sophisticated electronic techniques to recognize and correct for timing errors.

In this paper we have attempted to present a qualitative description of the recording and playback processes in digital tape-recording systems with emphasis on those factors which offer basic limitations to the density of recorded information on the tape.

Only the most important features could be considered. Such factors as the perpendicular component of the recording field, self-demagnetization of the tape, the particulate structure of the magnetic coatings, the finite permeability of the recording medium and of the recording and playback heads also influence the system performance but usually only in second-order phenomena.

The principal factors controlling the recording process are the magnetic characteristics of the recording medium, the thickness of the recording medium, the head-to-tape separation, and the gap length of the recording head. The principal factors controlling the playback process are the gap length of the playback head, the head-to-tape separation, and the thickness of the recording medium. In addition, variability of the azimuth angle results in reduced performance capability of a system.

Control of the azimuth angle is accessary in both recording and playback, although electronic correction techniques usually are applied only to the playback signal.

Knowledge of these elementary factors has led to improved tape and tape-recording systems in RCA as well as elsewhere. The direction of development is toward thinner, smoother, more wear-resistant tape coatings having higher remanent inductions and generally square-loop magnetic characteristics.

Also, shorter gap lengths are used in both the recording and playback heads, and more precise transports have been devised. Some of the limitations due to the head-tape relationships can be alleviated to some degree by means of sophisticated electronic techniques. However, many of these techniques can be used even more effectively if the basic limitations are further reduced. During recent years the tape stations in each generation of computer have offered significantly increased capabilities because of the developments listed above. And the end of such developments is still not in sight.

1. E. Delia Torre, "The Influence of Magnetic Tape on the Field of a Recording Head", RCA Review, March 1960.
2. G. Bate, H. S. Templeton and J. W. Wenner, "An Experiment on the Effect of Particle Orientation on Peak Shift in Magnetic Tapes", IBM Journal of Research and Development, July 1962.
3. W. K. Westmijze, "Studies on Magnetic Recording", Philips Research Reports, April, June 1953.
4. T. G. Woodward, "Approaches to Wideband, High-Resolution Magnetic Recording", Journal of the Audio Engineering Society, Jan. 1962.


Fig. 1 -The magnetization hysteresis loop of a ferromagnetic material.
Fig. 2 - Magnetic tape recording process.
Fig. 3-Pattern of magnetization in a magnetic tape coating.
Fig. 4 - The magnetic field in the vicinity of a recording gap.
Fig. 5 - The longitudinal component of the field at a recording gap. The y/l parameter refers to distance above the surface of the recording head.
Fig. 6 - Remanent magnetization in a tape coating as a function of applied field.
Fig. 7 - Pattern of magnetization in a tape coating following a step-function reversal of the recording field.
Fig. 8 - Pattern of magnetization in a tape coating for a short-duration reversal of the recording field.
Fig. 9-The magnetic tape playback process.
Fig. 10 - Playback scanning loss for recorded sine waves.
Fig. 11 -Playback separation loss for recorded sine waves.
Fig. 12 - Azimuth loss for recorded sine waves.