Towards Terabytes

Published Date
01 - Oct - 2005
| Last Updated
01 - Oct - 2005
Towards Terabytes
Here's an in-depth look at the technologies that will let you own a terabyte hard disk for your PC as early as next year

Imagine a Boeing 747 flying at 600 mph four feet above the ground, trying to count blades of grass as it flies by. That's pretty much what your hard drive's head does-be it something as simple as opening a file in Notepad, or something as disk-intensive as file indexing! The blades of grass are, of course, magnetised areas on the disk platter. And these magnetised areas (bits) are being packed more closely together by disk manufacturers. That's why we keep seeing increased capacity hard disks.

In 1973, IBM shipped the model 3340 Winchester sealed hard disk drive, the predecessor of all current hard disk drives. The 3340 had two spindles, each with a capacity of 30 MB. The Winchester got its name because it was invented at IBM's Winchester, NY, laboratory. What we use today are basically Winchester disks.

The recording method that's used with hard disks today dates back to the Winchester and beyond, and is called longitudinal recording (LR). And, as it turns out, LR is running out of steam. There is a method, though-namely, using AFC (AntiFerromagnetically-coupled) media-which promises to extend its life a little, but beyond that, we're on to other methods.

HAMR looks very promising to us. Our view is that heat-assisted technology will be needed sooner rather than later"
Dr Mark Kryder, Director of Seagate Research and CTO of Seagate Technologies

Who Needs Higher Data Densities?
The world does. A recent UC Berkeley study reports that the world produces between one and two exabytes (an exabyte is one billion gigabytes!) of information each year, comprising data on all media-magnetic, paper, film, and optical. Magnetic media is the towering king of all these-at 1 KB a page of written matter, and at 100 pages a typical office file folder, your 80 GB disk can hold eight lakh file folders!

We need denser and denser hard disks to store all the new information being created in the world.

But what about Flash memory and optical (holographic) storage and such? Will these not overtake hard drive storage at some point of time? At some point of time, yes, but not right now. Seagate claims that magnetic recording will remain the preferred form of mass storage. "With technologies (like holographic optical storage), what you've got is something that's much more expensive for the storage density you get," Seagate spokesperson John Paulsen said in 2002. "Magnetic technologies are mature, and they've been on a trajectory."

As for Flash memory, Jim Handy, director of non-volatile memory services at Semico Research writes, "Flash will be the leading choice in portable applications where a limited number of small files are used because it will offer the lowest overall system cost. Other limited-capacity applications will also gravitate towards Flash. In applications where file size or the total number of files to be stored is of more concern than total system cost, HDDs will prevail." There are further conclusions that Handy derives, but what we are to suppose is that there is still a long time before the hard disk goes the way of the floppy.

However, like we hinted at earlier, there will soon be a major shift in the way data is recorded on hard disk platters. Here, we'll be talking about AFC media, which could extend the life of LR a little while; perpendicular recording, or perpendicular magnetic recording (PMR); heat-assisted magnetic recording (HAMR, pronounced 'hammer'); and patterned media. But before all this, we need to talk a little about how a hard disk works.

Inside The Black Box
Hard disks have hard platters that hold the magnetic medium, as opposed to the flexible plastic film found in floppies. That's why they're called 'hard' disks. Each hard disk may have one or several platters. A platter is typically made either of aluminium alloy or a mixture of glass and ceramic. Both sides of the platters are coated with a magnetic medium.

The head-similar in function to the head in a tape recorder or VCR-flies over the disk, a fraction of a millimetre above it. Overall, the disk is a sealed box with controller electronics attached to one side. The electronics control the read/write mechanism as well as the motor that spins the platters. The electronics also assemble the magnetic domains on the drive into bytes for reading, and turn bytes into magnetic domains for writing. These electronics are contained on a small board that can be detached from the rest of the drive.

There's an arm that holds the read/write heads. It can move the heads from the hub to the edge of the disk. The arm and its movement mechanism are extremely light, fast, and precise.

Now comes the interesting part: how are the bits stored on the magnetic medium? The magnetisation is done on clusters of magnetic grains. About a hundred grains form a cluster. In LR, the magnetisation happens in the plane of the platter. Think of the clusters as little bar magnets placed in the same plane as the platter.

According to Dr Mark Kryder, director of Seagate Research and CTO of Seagate Technologies, LR still has some time left before it is phased out altogether-it will take us beyond 100 gigabits psi. That's a very high areal density, but the world needs even higher densities. Areal density refers to how closely packed the magnetic clusters are. As areal densities increase, the superparamagnetic effect kicks in at some point.
The superparamagnetic effect, like we said, kicks in when the clusters that hold the bits get too closely packed together. To increase density, apart from packing the clusters closer together, they have to be made smaller, too. This is what is happening now in the hard disk world: we're just two product generations away from reaching the limit that superparamagnetism imposes on areal densities. You're unlikely to see a 2,000 GB hard disk manufactured using LR. (AFC media, like we said, could delay the death of LR for a while.)
So what exactly is the superparamagnetic effect? Superparamagnetism occurs when the magnetic grains on the disk become so tiny that random thermal vibrations at room temperature cause them to lose their ability to hold their magnetic orientations. What results are 'flipped bits'-bits whose magnetic north and south poles spontaneously reverse. This, naturally, corrupts the data, rendering it and the storage device unreliable and ultimately unusable.

If the problem lies in the bits getting demagnetised, why not just use stronger magnetisation? It turns out that the fields that can be applied are limited by the magnetic materials from which the head is made, and these limits are being approached, too!

Superparamagnetism is forcing the industry to slow the historically rapid pace of growth in drive capacity-a pace that, at its peak over the past decade, doubled capacity every 12 months. The way around superparamagnetism-or rather, the way to forestall its effects-is to use AFC media and/or perpendicular recording, scientists believe. This could create opportunities for continued growth in areal densities at a rate of about 40 per cent each year.

The exact areal density at which the superparamagnetic effect occurs has been subject to a lot of debate. Scientists in the 1970s predicted that the limit would be reached when data densities reached 25 megabits per square inch (psi)! But through constant innovation, such limits have been pushed forward by orders of magnitude.

In our discussion of how the limitations imposed by the superparamagnetic effect are being tackled, first up is AFC media.

AFC Media
AFC Media was designed to extend the life of LR. The idea is simple, but there are quite a few details under the hood. Regular media uses one layer of magnetic material; AFC media uses two layers, the top one thicker than the bottom one, with a three-atom-thick layer of ruthenium (a non-magnetic material) sandwiched in between. (IBM calls the ruthenium layer 'pixie dust'.) This arrangement allows for magnetic domains (or bits) to be more closely packed.

The highest areal density with LR has crossed 100 gigabits psi. And researchers now believe that 120 gigabits psi will be 'too much', and that at that point, PMR will become dominant

So how does it work? First off, recall that the superparamagnetic effect originates from the shrinking volume of the magnetic grains that compose the storage properties of the media. The magnetic grains represent the bits that are stored as alternating magnetic orientations. To increase densities while maintaining acceptable performance, designers have shrunk the media's grain diameters and decreased the thickness of the media. The smaller grain volume makes them increasingly susceptible to thermal fluctuations, which decreases the signal sensed by the drive's head. And if the signal reduction is great enough, data could be lost over time to the superparamagnetic effect.

Now, the precise thickness of the ruthenium in AFC media causes the magnetisation in each of the magnetic layers to be coupled in opposite, or anti-parallel, directions, which constitutes 'antiferromagnetic coupling'. ('Antiferromagnetically-coupled media' is AFC.) Refer to the diagram above.

When reading data as it flies over the rotating disk, the head senses the magnetic 'transitions' in the magnetic media. The strength of this signal is proportional to the media's 'magnetic thickness'-the product of the media's remnant magnetic moment density Mr (think of this as simply how strongly the media has been magnetised) and its physical thickness, t. As the data density increases, the magnetic thickness-Mr x t, or Mrt-must be decreased so that the closely-packed transitions will be sharp enough to be read clearly. This, in turn, is because the 'transitions' will not be recorded nearly as well if the signal amplitude is too high.

A Magnetic Head Reading AFC Media

The effective Mrt equals Mrt(Top) - Mrt(Bottom). The ruthenium layer is sandwiched between the two layers. 'CoPtCrB' is cobalt-platinum-chromium-boron. The point here is that the effective Mrt needs to be reduced if the bits are packed more closely together and if the magnetic transitions are still to be read accurately   

At this point it's better explained with equations. It should be obvious that if Mrt is to be reduced, you can either reduce Mr or you can reduce t. Reducing Mr is not an option because that means lower magnetisation, so t has to be reduced. And when t (the thickness of the media) is reduced too much, as would be done traditionally, the superparamagnetic effect would come in. If it weren't for the ruthenium layer, that is.

With the ruthenium layer in place, it turns out that the effective Mrt, or Mrt(eff), is given by Mrt (eff) = Mrt (top) - Mrt (bottom). This is the killer equation: this property of AFC media permits the overall Mrt to be reduced, and its data density increased, independent of its overall physical thickness. Thus for a given areal density, the Mrt of the top magnetic layer of AFC media can be relatively large compared with single-layer media, permitting larger grain volumes-which, of course, are inherently more thermally stable.

Two additional advantages of AFC media are that it can be made using existing production equipment at little or no additional cost, and that its writing and read-back characteristics are similar to conventional longitudinal media.

Mass production of the first hard drives with AFC media began in May 2001 with Hitachi GST's 2.5-inch Travelstar 15GN and 30GN, which featured 15 GB per platter, and an areal density of 25.7 gigabits psi. And within three years, AFC media enabled the industry's first 400 GB 3.5-inch drive, the Deskstar 7K400.

AFC media isn't the dominant way forward, however, and the immediate future lies in PMR.
Perpendicular Magnetic Recording (PMR)
The highest areal density with LR has just recently crossed 100 gigabits psi. And researchers now believe that 120 gigabits psi will be 'too much', and that at that point, PMR will become dominant.

So what is PMR? Simple. The clusters that hold bits of data are not magnetised along the plane of the platter, but rather, 'vertically', that is, perpendicular to the plane of the platter. Think of it as high-rise buildings or skyscrapers in the place of sprawling condominiums. Or, looking at a platter held on the top of your palm, longitudinal recording creates horizontal bar magnets 'on' the platter; whereas PMR creates 'vertical' bar magnets.

Longitudinal Recording

In longitudinal recording, the magnetisation happens in the plane of the platter. Think of the bits of data as little bar magnets lying along the platter

Perpendicular Recording

In perpendicular recording, the magnetisation happens in the plane vertical to the platter. Think of the data bits as little bar magnets standing on end

Coercivity (the amount of magnetic field required to write data) is enhanced in PMR-meaning the data will be more stable against superparamagnetism. And the geometry lends itself to denser packing together of magnetic clusters. According to a 2005 Hitachi GST (Global Storage Technologies) report, "the geometry and coercivity advantages of PMR led scientists to believe in potential areal densities that are up to 10 times greater than the maximum possible with longitudinal recording. Given current estimates, that would suggest an areal density using perpendicular recording as great as one terabit psi-making possible, in two to three years, a 3.5-inch disk drive capable of storing a terabyte of data."

Of course, it's not all rosy, and challenges remain. As an indicator, the same Hitachi GST report goes on to say, "Even though PMR is technically akin to the current generation of longitudinal devices, a number of technical challenges remain. For example, engineers are engaged in research to invent new kinds of read/write heads; to experiment with new materials that have enhanced magnetic properties and improved surface finishes; to maintain signal-to-noise ratios as the magnetic bits and signals become smaller; and to detect and interpret the magnetic signals using ever more advanced algorithms."

IDC, a global provider of market intelligence for the IT and telecom industries, said in a recent (2005) report that PMR technology will be key to the hard drive industry, and that its adoption cannot be delayed. "The transition to PMR and to the new heads and media required for this technology will begin late in 2005 and will be broadly adopted in products by the end of 2007." IDC predicts that PMR will be ubiquitous in the 630 million hard drives that will be shipped in 2009.

And yes, it's not all in the future. As far back as November 2002, Seagate Technology announced that its scientists had broken new ground by demonstrating areal densities of over 100 gigabits psi using PMR. That's already a tenth of what is predicted as being possible. And in July 2005, Seagate announced that virtually all its hard disk production would use PMR by the end of next year. "Over the coming year, the vast majority of products shipping (from Seagate) will be using perpendicular technology," said Charles Pope, Seagate's Chief Financial Officer. This is notable because no other company has made such bold claims about such a rapid transition of production to PMR.

Seagate's first hard disk to use platters with perpendicular recording capability will be the 2.5-inch Momentus 5400.3 with capacity of 160 GB, and will ship in the winter of 2005.

The transition to PMR and to the new heads and media required for this technology will begin late in 2005 and will be broadly adopted in products by the end of 2007

Heat-Assisted Magnetic Recording (HAMR)
Another recording technology that has been proposed and demonstrated as a way out of the superparamagnetism problem is HAMR. HAMR promises densities of 50 terabits psi-huge compared to the 1 terabit psi that's promised by PMR! The basic idea is quite simple: use a laser to heat the spot at which the recording is to be done-the heating makes the medium easier to magnetise.

To elaborate a little, heat-assisted recording involves producing a hot spot (usually with a laser) on the media, while data is written magnetically at the same time. The net effect is that when the media is heated, the coercivity, or field, required to write on the media is reduced. This makes it possible to write on high-coercivity media, which have higher stability against superparamagnetism. This writing can be done despite the magnetic write heads having limited fields.

Explained in a different way, HAMR entails using films that hold very tightly onto the magnetic orientation of each memory cell-so tightly that a laser must heat up each cell to be able to write data. Because the materials hold magnetic charges so well, memory cells can be packed together far more tightly than with less robust, conventional materials.

Seagate spokesperson John Paulsen said in 2002 that HAMR would be incorporated in products between 2007 and 2012. But guesses and predictions about this date vary, and it's not entirely clear when HAMR will be commercially mainstream.

Reality check: where does HAMR stand today? Seagate has already demonstrated the technology with prototype HAMR devices, but hasn't actually made a working, out-of-factory HAMR hard disk.

Heat-Assisted Magnetic Recording

Isotherms are lines indicating regions of equal temperature. In HAMR, the laser heats up a spot on the medium, which is simultaneously magnetised by a ring head. The heating is required because higher-coercivity media is used
Patterned Media

Here, a single magnetic grain can hold a bit. Along the recorded track, a change in grain magnetisation means a 1, and no change means a 0

Patterned Media

Patterned media is looking considerably into the future. However, we mention it because it is an important technology being looked into, and might well be the distant future of hard disks. Seagate's Mark Kryder believes PMR might take areal densities to as much as 1 terabit psi, but doesn't expect it to go any further than that. On the other hand, he points out that work is continuing on other technologies that can be ready when the industry needs them. HAMR and patterned media recording are the two most favoured alternatives.

In patterned media, we need just about one grain-sized volume per bit. Density can therefore be increased by about two orders of magnitude

John Best, CTO of Hitachi GST, has said that PMR "in its first two generations probably provides a fairly modest areal density increase, on the order of a factor of two or three." Bigger advances will come when PMR is linked to a new generation of so-called patterned media, Best said.

The idea of patterned media is this: like we said before, a few hundred magnetic grains are required to store a bit. Densities could increase drastically if we could find a way to store one bit per grain. This can happen only when the grains are ordered in a particular way-that's why it's called patterned media. Of course, this assumes that drive makers can find economical ways to read and write the patterns.

"I'm a bit of a patterned-media enthusiast myself, but I couldn't predict the time frame it might happen," Best said. "It might come before, after or at the same time as work in HAMR."

So here's how patterned media works: look at the figure alongside. A '0' is indicated when there is no transition in magnetisation, and a '1' is indicated when there is a transition. The boundaries between regions of opposite magnetisation must occur along the boundaries between the grains-this makes the boundaries meandering, approximating the ideal straight boundary. Remember that the straighter the boundary is, the easier it is to detect (read).

If the grains are small enough, the magnetic transitions are straight enough, so it's easy to detect which bit cells contain a boundary and which do not. However, if the density is increased (meaning the bit cells are shrunk) without shrinking the grains, the magnetic transitions become noisier, eventually preventing the head from accurately reading the data. To keep the noise associated with grain boundaries small enough for reliable data detection is why fifty to hundred grains are needed per bit cell.

Now, the solution to achieve higher densities is obvious: make smaller grains. As of now, however, grain sizes have gotten so small that further shrinkage would bring in the superparamagnetic effect. But with patterned media, each bit is stored in a single, deliberately-formed magnetic volume. This may be one grain or several coupled grains, rather than a collection of random decoupled grains. The magnetic transitions no longer meander between random grains, but form perfectly distinct boundaries between precisely located islands. This means that we need just about one grain-sized volume per bit. Density can therefore be increased by about two orders of magnitude.

What To Look Out For
AFC media will be here for a while, and PMR will take off in a big way this year. On the whole, the picture looks perfect enough. This is one industry that has everything going for it. There's a fantastically huge market, which will only grow; so far, the roadblocks have been cleared, and research is already underway to clear the hurdles ahead. Prepare for 1,000 GB hard disk within 12 months from now-and prepare for a 2,000 GB hard disk in the year after that!

The only change that you will probably see is hard disks being used in more and more devices. Expect to see hard disks in places you haven't seen them before. Also, as hard disks become smaller and smaller and more affordable, they will come more directly in competition with the other storage media, such as Flash and MRAM.

Some aver that holographic storage is the future of storage-refer the March 2004 issue of Digit. It may be, but like we said earlier, practical, everyday holographic storage is still very much in the future, and the hard disk industry need not worry about it yet.

The immediate future, beginning this year, is PMR-just remember that you read about it here first!

Team DigitTeam Digit

All of us are better than one of us.