With new laser, media, and signal-processing technologies, magneto-optical storage is primed for capacity growth rates rivaling that of magnetic disks
David K. Campbell and Kraig Proehl
Magneto-optical storage, or MO, is a rewritable storage technology that offers far better cost per megabyte than does magnetic storage. This is primarily because MO disks are removable and interchangeable with one another. However, because the standard storage capacity associated with MO technology remained steady at 650 MB (per 5 1/4-inch platter) for the first four years after its commercial introduction, most people viewed MO technology as slow moving, especially when they compared it to magnetic disk technology, which experiences major capacity gains at least once every year. In fact, it was only last year that 1.3-GB MO drives
hit the market.
Today, MO technology is poised to shake off its "slow-moving" image. It has taken time for the researchers and developers behind MO technology to gain a complete understanding of the technology, its tolerances, and how to push the technology in a way that allowed for commercialization. This learning curve has been absorbed, which means you will see dramatic capacity increases for MO technology in the next four years.
In addition, the companies selling MO technology are attacking a similar learning curve. They are providing more drivers and applications that exploit the high density and removability that MO drives and media provide. The removability aspect of MO technology has spawned an entire new market for auto-changers and robotic library systems, which will only require more and more storage in the future. MO technology is poised for explosive capacity increases, which will satisfy those needs.
In fact, today's 1.3-GB capacity for 5 1/4-inch MO technology is expected t
o double before 1995, again before 1996, and once more before 1998. The upshot is that in a period of four years, MO storage capacity will probably increase from today's 1.3 GB per 5 1/4-inch platter to 10.4 GB. That's what absorbing a learning curve will do for a technology.
Short-Term Improvements
The next logical step in the evolution of MO technology is to double the capacity of the current drives and media to 2.6 GB of storage. The exciting aspect of this jump is that it can be accomplished fairly easily with current technologies and methods.
The key factor for the initial capacity jump from 650 MB to 1.3 GB was the implementation of ZBR (zone-bit recording) or ZCAV (zone constant angular velocity). ZBR is a method in which more sectors are recorded at the outer radius of the disk to maximize bit density. This method has been used extensively in magnetic recording over the last three to four years. MO devices will continue to use this method for the 2.6-GB drives and beyond and will leve
rage the programmable equalizers and frequency synthesizers already developed for this method.
The major advance that will allow MO recording to make the next capacity jump to 2.6 GB is PWM (pulse-width modulation) recording. Current 650-MB and 1.3-GB products use a recording method known as PPM (pulse-position modulation). In PPM, the information (i.e., data) is contained in the time between the positive peaks of the readback signal. The positive peaks correspond to the center of the written domain. With PWM, the information is contained in the time between the transitions (edges) of the readback signals (see the figure "Dots vs. Edges"). It is the length of the domains and the time between the domains that are important here.
In optical PWM recording, the big challenge is the precise control of the domain edges. Writing an MO disk is primarily a thermal process with the laser serving as the heat source. Care must be taken to control the laser power accurately to write consistent, well-defined
domains. The change to a PWM system alone will double disk capacity.
Other changes, however, will likely steal some of the capacity. The biggest change will probably be from the current RLL 2,7 encoding scheme to an RLL 1,7 encoding scheme. The change is necessary to buy back some bit-window margin for the read channel systems. The new encoding will sacrifice about 12 percent of the capacity improvements gained with PWM.
Fortunately, there are other means to return this lost capacity. The first method is to reduce the track pitch on the media. Track pitch is the distance between adjacent tracks on the disk. By reducing this distance, more tracks (and thus, more data) can be contained on the media. First-generation media (650 MB) had a track pitch of 1.6 microns. Second-generation media (1.3 GB) reduced this number to 1.39 microns, and the next generation of media (2.6 GB) will most likely go to 1.15 microns.
The Wavelength Issue
The final evolutionary changes to the MO system that are
needed to double capacity to 2.6 GB involves the optics of the drive. The bit density you can achieve on a disk is directly related to the spot size of the focused laser beam on the disk. Although elementary physics explains that coherent light passing through a lens can be focused to an infinitesimally small point, there is a physical limit to how small the spot can be, based on the diffraction properties of the optics. The size of the spot is directly related to the laser wavelength and the NA (numerical aperture) of the objective lens. (NA can be easily defined as the inverse of the "f-number" that is on cameras and telescopes.) Spot size is proportional to the wavelength divided by NA, so by decreasing the laser wavelength and increasing the NA, the spot size--and therefore, the bit size--can be significantly decreased. Current 1.3-GB drives typically use a laser with a wavelength of 780 nanometers and an NA of 0.55. Future drives will incorporate 670-nm lasers and objective lenses with an NA approaching
0.60, which will allow the density to reach 2.6 GB with room to spare.
Because of the removability of MO media, interchangeability standards are an important part of bringing higher-capacity products to market. Early in the life of MO technology, the standards process was excruciatingly slow and inefficient. As the market has expanded and the knowledge base has increased, drive, media, and component manufacturers have worked together to eliminate much of the cycle time in defining standards.
Beyond 2.6 GB
Numerous methods are already under development for even further increases in disk capacity. These methods will allow capacities of 5.2 GB per 5 1/4-inch platter; that's eight times the capacity of the pre-1993 650-MB drives. Some of the methods such as optical and magnetic superresolution are unique and revolutionary. Others will be leveraged from new methods being applied to magnetic disk drives.
As described earlier, decreasing the size of the laser spot focused on the MO media allo
ws for increased capacity. This spot, rather than being infinitesimally small, is limited by optical diffraction. Extrapolating from current trends, we expect laser wavelengths may drop to 640 nm, and NAs may increase to 0.65 by 1998. This, unfortunately, will give only a 12 percent reduction in spot size and only a 30 percent potential capacity increase from the 2.6-GB generation.
Further advances are needed, such as the shorter wavelength blue and green laser diodes just now operating in research labs and not likely to be commercialized until 1997 or 1998. Other techniques expected to be important for 5.2-GB products are optical superresolution and magnetic superresolution, which can provide capacity increases.
Optical superresolution modifies the optical power distribution of the beam focused on the optical disk. Power is shifted away from the center of the beam toward the edges, producing--through diffraction--a smaller focused spot. This smaller spot can resolve smaller features on the disk
, allowing increased capacity without increases in NA or decreases in wavelength.
In many ways, optical superresolution is analogous to an electronic filter that boosts high frequencies and suppresses low frequencies. Unlike the electronic filter, however, optical superresolution boosts the high spatial-frequency amplitude response but does so with no change in phase response. Some demonstrations of optical superresolution have given 20 percent to 30 percent in track density increases. (See references 1 and 2.)
Submicron Dots
Magnetic superresolution, or MSR, is an even more promising, superresolving technique for MO drives. Rather than changing the optical power distribution in the focused laser beam, MSR produces a submicron-size aperture, or window, at the focused spot. The size of this window is what determines the resolution of the MO drive, rather than the size of the focused laser spot. While the focused laser spot may have a diameter of 1 micron, this magnetic window can have a size o
f much less than a micron, on the order of 0.3 microns in size.
This amazing discovery was first demonstrated by researchers at Sony, who created the MSR window by adding an extra MO layer to the standard MO disk (see the figure "Magnetic Superresolution"). With MSR media, data bits are stored in a buried memory layer. During readout, the focused laser beam heats the memory layer and a readout layer. Through magnetic-exchange coupling, the bits are copied from the memory layer to the readout layer and are visible to the laser beam. Since only a submicron area is heated, only data in a submicron area is seen and resolved. Once the data is read and passes from beneath the laser beam, the films cool and pass through a magnetic field, erasing the domains in the readout layer but leaving the data intact in the memory layer. (See reference 3.)
Another promising technology is the use of PRML (partial-response maximum-likelihood) channels that employ digital filtering techniques to allow the readback sy
stem to adapt to the many variables that contribute to ISI (intersymbol interference) among the recorded bits. ISI is the tendency of one bit to interfere with, or distort, adjacent bits. It imposes a practical limit to how closely bits can be spaced on the media. Adaptive PRML channels allow the bits to be placed closer together because the filtering technique is better able to find the data in the midst of all the ISI (see "Digital Hard Drives" on page 91). In addition, PRML channels incorporate a sampled detection scheme, rather than a traditional continuous-time, zero-crossing detection. In this technique, the channel samples the filtered analog readback signal once per bit time. It then looks at each sample and the adjacent samples and makes a decision whether the sample is a 1 or a 0. It is this "intelligence" and its ability to adapt to different heads, media, and environmental conditions that give the PRML channel so much power.
Reductions in track pitch typically contribute to increased capaci
ty. Unfortunately, as tracks are pushed closer together, interference, or cross talk, from adjacent tracks can become significant and result in unacceptable loss of margin. This is a problem for optical and magnetic drives. Fortunately, under development for MO technology are techniques that eliminate cross talk and will allow track spacing of 0.6 microns or less.
The first technique for eliminating cross talk is the previously mentioned MSR. With MSR, only bits from the desired track are copied to the readout layer; all adjacent tracks are invisible to the readout beam and thus contribute no signal or cross talk to the desired signal.
A second technique called cross-talk cancellation requires no changes in the thin-film structure for MO media but takes advantage of the parallelism of optics (i.e., the ability to read or write multiple tracks at one time through the same optical system). Cross-talk cancellation incorporates a diffraction grating and additional photo detector elements into an MO
drive so that three tracks can be read at one time. The desired data track is read, as well as both neighboring tracks.
Then, in a custom cross-talk cancellation IC, the unwanted signal from the neighboring tracks is subtracted electronically from the desired signal. Tracks can be placed closer together without the deleterious effects of adjacent track cross talk causing data corruption. (See reference 4.)
Media Changes
Semiconductor diode lasers, LSI circuits, and low-cost, environmentally stable MO media are the technologies that enable and, for the most part, pace the development of MO drives. Of these, it is the media--that seemingly simple combination of thin-film coated polycarbonate housed in a plastic cartridge--that is unique to MO technology. R&D in MO media fuels future MO drive developments. Today's media research is focusing on four key areas: lower noise media for higher density, short wavelength-response MO materials, magnetic superresolution constructions, and media that will al
low single-pass direct overwrite of data.
The control of noise sources in MO drives is critical to higher-density storage, and central to this is understanding and minimizing media noise. Media noise comes from two sources: writing noise and readback noise. Writing noise is evident in irregularities in the written magnetic domains. For example, even with perfect control of the laser-writing pulse, the resultant domain has submicron irregularities in the domain shape. The causes of these irregularities are micromagnetic variations in properties of the MO RETM (rare earth transition metal) films, such as domain wall motion, domain anisotropism, and thin-film defects. Improved materials and thin-film deposition processes are under development to reduce these effects.
Readback noise includes all the effects that contribute to noise, or jitter, in the detection of the domain edges from the disk. Electronic noise and laser noise are the dominant drive sources. Media readback noise results from interfe
rence sources such as polycarbonate substrate birefringence, tracking groove irregularities, thin-film reflectivity variations, and thin-film variations in MO Kerr effect. (The MO Kerr effect causes the polarization of the readback laser beam to be rotated in either a positive or negative direction, depending on the polarization of the domains on the disk. The direction of the rotation determines whether the bit is 1 or 0.) Techniques are being developed to separate these different effects, and substrate and thin-film materials and processes are being improved to minimize these readback noise sources. (See reference 5.)
Development of media for use with short wavelength lasers is a critical research topic for increased capacity MO storage. Modification of today's (RETM) films and transparent polycarbonate substrates will suffice through 5.2-GB capacity drives. However, 10.4-GB capacity drives--16 times the original MO capacity--will likely use either blue- or green-wavelength laser diodes, and media mu
st be matched with those wavelengths.
Unfortunately, today's RETM films have little, if any, MO response at blue or green wavelengths. For this reason, research into new MO materials, such as cobalt-platinum multilayers, neodymium-iron-cobalt films, and bismuth doped garnets, is active. IBM researchers recently demonstrated read/write operations on a newly developed MO media at 2.5 Gb per square inch using a blue wavelength, 428-nm laser. (See reference 6.)
Writing Without Erasing
DOW (direct overwrite) MO media will let the drive write data using a single pass of the focused laser beam, rather than the two-pass erase-then-write process used today. Two forms of DOW media and drives are being researched and developed. The first uses fairly conventional MO media but coats the back side of the disk to allow a small magnetic head to either contact or fly above the MO films, as shown in "Magnetic Superresolution." This magnetic head is small enough to allow it to switch its magnetic field at the d
ata rate of the recording process. This process is being used in Sony's MO Mini Disc for digital-audio recording and will likely be adopted into 3 1/2-inch and
2 1/2-inch MO drives.
DOW technology for 5 1/4-inch MO drives will likely use exchange-coupled MO media. This media is similar to MSR media in that it incorporates additional MO layers into the disk construction. One MO layer acts as the memory layer, storing the data; another acts as a magnetic reset layer, eliminating the need for a switching magnetic field. Exchange-coupled DOW MO media is written by driving the laser diode among three power levels, a low-power P subscript read for reading the data, a P subscript lo for writing 0s, and the highest-power P subscript hi for writing 1s (see reference 7). Because the temperature dependent coercivity of the memory layer is different from that of the bias layer, these varying power levels allow direct overwrite of MO data. The advantage of exchange-coupled DOW media and the reason it will likel
y appear in 5 1/4-inch drives is that it doesn't preclude the use of double-sided disks. Also, the data rate of this type of recording is limited only by the laser-diode modulation frequency, rather than by a magnetic coil modulation frequency, as is the case for magnetic-field modulation DOW.
The Changing Laser
One common thread for much of the higher-capacity technology is the need for improved power control of the diode laser during the writing and reading process. PWM recording, PRML channels, magnetic superresolution, the control of disk writing noise and direct overwrite media require more precise laser-power control. Fortunately, the advent of digital signal processor-controlled servo systems in MO drives gives this higher degree of control and, more important, provides adaptive control of the laser power. DSP-adaptive control adjusts the laser-power levels for reading and writing to adapt to variations in drive and media temperature, media sensitivity, and data-writing patterns. MO drives fr
om Hewlett-Packard and others are being shipped with full DSP-based servos, which will become the norm for all future drives.
Reductions in laser wavelengths have and will be an important pacing item in MO drive capacity increases. One exciting aspect of shorter wavelength lasers is that the time from research samples at a particular power and wavelength to full commercial production has been decreasing. With green and blue laser diodes being produced in research labs today, it is quite likely that these devices will become commercial before the end of the decade.
Two techniques for producing short wavelength lasers are being explored. The first takes the output from a high-power 830-nm laser diode and injects it into a nonlinear crystalline material to produce frequency-doubled radiation at a 415-nm wavelength. Power outputs as high as 54 mW have been achieved. Commercial development of a compact, frequency-doubled laser diode is under way at Coherent in Palo Alto, California. Other companies a
re also developing these frequency-doubled lasers.
An even more exciting but riskier prospect is the development of direct-output blue and green diode lasers using Group II-VI materials such as zinc selenide. These lasers will likely be less expensive and smaller and require less power than frequency-doubled lasers. 3M researchers have demonstrated room-temperature pulsed operation of these devices but only with lifetimes of several hours. Researchers are optimistic that lifetimes and output powers will be improved, and these lasers will be commercialized by the end of the decade. (See reference 8.)
The Need for MO
As mainframes give way to distributed-computing environments, organizations need a cost-effective and dependable way to handle enormous amounts of secondary storage. Furthermore, the advancements in MO storage capacity mean that the technology will keep pace with customers' storage requirements, as well as an even lower per-megabyte cost.
The bottom line is MO technology has
made the transition from an "interesting technology" to a reliable solution that's part of mainstream computing. For storage-hungry applications (e.g., image management, network data management, on-line archives, and unattended backups) MO delivers an ideal solution. Given the capacity advances on the horizon, the uses of MO technology will only expand with time.
MO Applications
ARCHIVAL STORAGE
DOCUMENT MANAGEMENT
DOCUMENT IMAGE PROCESSING
LARGE-CAPACITY DATABASES
TEXT INFORMATION REFINING
Figure: MO Capacities Over Time
After a slow beginning, the capacity of standard MO drives is expected to top 10 GB by the end of the decade. Some companies--such as Hitachi, with its 2-GB drive--are already exceeding the ISO standard 1.3-GB drives, although it's not known whether the market will support a lot of different formats.
Figure: Dots vs. Edges
With pulse-position modulation, each 1 bit has a corresponding spot written to the disk. With p
ulse-width modulation, it isn't the spot that is significant but the leading and trailing edge of spots. Each transition corresponds to a 1 bit, with the time between corresponding to a 0 bit.
Figure: Magnetic Superresolution
The read layer heats only a submicron area to a temperature high enough to copy data from the memory layer to the readout layer, letting you use spots smaller than the laser beam diameter.
Figure: Halving Write Times
Media designed for magnetic-field modulation DOW includes a protection layer that permits a small magnet to contact or fly just above the media surface. The magnet is small enough so that it can change its orientation at the data rate of the drive. Its small size--and thus low power--make necessary its close proximity to the disk.
David K. Campbell is an optical engineer and Kraig Proehl is an R&D engineer at Hewlett-Packard's Greeley Storage Division (Greeley, CO). You can reach them on the Internet at
david_campbell@hp5800.desk.hp.com
and
kraigp@hpgrla.gr.hp.com
, respectively, or on BIX c/o "editors."
Optical Advances
