Hard Drive Technology: Lots of Storage in a Tiny Space Imagine taking a tiny section of a flat disk, say 10 nanometers

or smaller, heating it up to 650 degrees centigrade for a nanosecond or so, and then magnetically charging the location. It sounds pretty incredible, but it's likely what will be happening inside the hard drives we'll all be using in the next few years. While many of us who follow technology pay a lot of attention to semiconductor technology, such as the transition to 32 or 22nm process technology, we often take hard drive technology for granted. But there's a lot going on here. I've recently talked to the three major remaining hard drive makers--Seagate, Toshiba, and Western Digital--to get a feel for where they think the technology is heading. (Currently, only five companies make actual hard drives, though many more package them. Seagate is in the process of acquiring Samsung's hard drive business; and Western Digital is doing the same with Hitachi's; both deals are still awaiting regulatory approval.) On average, hard drive capacity has been doubling every two years for the past 50-plus years. Today, a 2.5-inch drive stores up to 320GB per platter, which is moving to 500GB, and most drives have two platters. But over the next few years, keeping the technology moving forward will likely require some very new technology. As Bill Cain, the vice president of technology for Western Digital, explains it, the growth in capacity per area of the disk platter (known as areal density) hasn't been as steady as with processors or memory. Rather, he said, it was better characterized as "technology S-curves," in which improvements start out slow, then get very fast, and then slow down again. Mark Re, Seagate's senior vice president for recording media R&D and operations, noted that the industry made a transition from longitudinal to "perpendicular recording" about seven years ago. Cain noted that for the last few years, we've seen about a 35 percent annual growth, but it's beginning to slow down. Right now, the high volume production is of disk platters that store 500Gb per square inch, Cain said, but by the end of the year will be seeing drives with 600Gb to 700Gb per square inch. That means that in a 2.5-inch drive, you could get up to 500GB per platter and in a 3.5-inch drive, you can get up to 1TB per platter (at an aerial density of 640Gb per square inch). Re noted that enterprise drives always lagged consumer drives in capacity, saying that while a typical notebook drive could be up to 500Gb per sq. inch, density is only about 300Gb per square inch for a typical enterprise drive (which spins faster and has lower error rates). Seagate's Re said that the general industry consensus is that about 1 terabit per square inch is the limit of that technology, marking a limit at about 2TB 2.5-inch drives. Toshiba's Joel Hagberg was a bit more pessimistic, thinking the next technology would be needed at 750Gb per square inch to 1 terabit, while WD's Cain said the industry's "track record in predicting when a technology expires is pretty poor" and said the industry would do what it could with the existing technology.

Going forward, Cain said, all the new technology use perpendicular recording (with the bits stored normal to the disk surface), but will continue to make the bit size smaller. Cain indicated that WD's next step was likely to be "shingled magnetic recording (SMR), as it uses existing drive head and media technologies. Today, the data on a hard drive is "reasonably isolated" with a "guard band" between adjacent tracks. In SMR, instead of a guard band, the data is overlapped, like shingles on a roof, Cain said. The result is tracks that are narrower than the heads that wrote it. You then can't arbitrarily update a single sector, but instead have to manage the data in blocks, just like in an SSD today. But in most applications, a user wouldn't notice the difference. Cain said WD was working within the International Disk Drive Equipment and Materials Association (IDEMA) on this technology and expects that you'll see drives with this technology in the next couple of years, following a couple of more generations with conventional perpendicular technology. All of the vendors then talked about various methods of energy-assisted recording, often focusing on a technology called heat-assisted magnetic recording (HAMR). Cain explained that in today's discs, media is recorded in clusters of physical "grains" of magnetic materials, and each cluster contains a single bit of data. According to Cain, ideally, you make the grains smaller, which keeps signal-to-noise similar as the density increases. However, he said as they get smaller, drive makers run into three issues: the intrinsic magnetic properties of the media need to be adjusted to compensate for thermal excitation; the signal-tonoise ratio needs to stay similar and relative to the size; and the amount of energy it takes to write to the media. Over the past 50 years, he said, the drive industry has changed the property of the media, by making the grains smaller, and changed the heads, so they have more power. But now, Cain said the heads have reached the limit to how much magnetic field they can create, and the material can't have a higher magnetic moment. A new approach is to have additional energy to help the magnetic fields. This is generally known as energy-assisted recording; one form is heat-assisted magnetic recording (HAMR) and another is microwave-assisted magnetic recording (MAMR). In the first case, the media is heated to several hundred degrees centigrade, so the magnetic field in the head can write to it, but then as it cools, it becomes thermally stable. In the microwave case, it's similar, but using a microwave tuned to the resonance of the media. Cain said these methods are actively being developed, and the industry is looking to see deployments over the next several years. Seagate's Re was also very bullish on HAMR as the next step. As he explained it, with HAMR, the actual material gets harder to switch, so it can be more dense, but then it becomes harder to write to. So instead, thermal energy heats up the spot where you want to write. He talked about using semiconductor lasers in the recording heat and near-field optics.

Re discussed using commercially available red or near-infrared lasers to head up a very small bit of media on the scale of ten nanometers to 650 degrees centigrade for a nanosecond. You could then write to the media, which would then cool off very quickly, leaving the media stable. "To get that kind of thermal gradient over that area in such a small time scale is pretty amazing," Re said, noting that since much of the basic technology is already around, it would be faster to deploy than some of the other options. Toshiba's Hagberg even talked about potentially seeing HAMR platters at 750GB to 1TB, again suggesting that it would like appear in consumer drives first and enterprise drives later. A third solution is patterned media, in which the median has little "islands" physically patterned onto the media. These are larger than grains today, but smaller than bits. The challenge, Cain said, is manufacturing technology, as the bits are smaller than the cell size in semiconductors today. You would need to store trillions of little bits on the media surface--the equivalent of manufacturing semiconductors at a single-digit nanometer process technology. Cain said the concept is clean, but the technology and equipment have yet to be developed. Things like nano-imprint lithography or e-beam lithography might work, but these aren't available in the near term, according to Cain. "HDDs are really nanotechnology in action," he said, explaining that bits on a hard drive are smaller than the smallest cell in flash memory. He noted that the grains that hold bits are equivalent to about 9 nanometers, and the heads write tracks that are 80 to 90 nanometers wide with a 5 nanometer accuracy. Creating new technology involves lots of different engineering disciplines, including software, electronics, aeronautics, mechanical, and material science. Overall, WD's Cain said he thinks that the first energy assisted drives are three to five years out, with patterned media several years away. Re said that Seagate thinks HAMR comes first because it can do that with the equipment it already has for heads and disks, while bit-patterned media would require nano-imprint lithography and new equipment. Re expects the first HAMR drives out by the middle of the decade. And Toshiba's Hagberg wouldn't give a timeline, but seemed to be leaning in the same general direction. While a lot of people have talked about flash memory-based solid state disks (SSDs), hard drives remain much less expensive on a cost-per-gigabyte basis, and are likely to stay that way for a very long time. Today, a typical 2.5-inch 500GB drive costs a system vendor under $40 wholesale, while a 32GB SSD based on NAND Flash Memory might cost slightly more. Unless users decide they are happy enough just streaming content from the Internet, it seems like hard drive storage is going to be a part of most PCs for a long time.