C&Ped from the WAVX Board. I didn't recall seeing it here previously but then July 99 seems so long ago that perhaps I just don't remember.
To: Scott Rafe who started this subject
From: Klingerg
Sunday, Jan 2, 2000 9:16 AM ET
Respond to Post # 9642 of 9644
Hard Disk Technology Article found by Nelzoni on RB which I haven't seen so I
thought I'd post it here for all. The article is from 7/99. NOTE!!!! the links listed at the
bottom of the article.
nikkeibp.com
GMR Heads Expedites
Sub-$100 HDDs for Home Use
The hard disk drive market is entering a new era. Products such as the sub-US$1,000
personal computer for the home are pioneering new drive applications where price
takes the highest priority.
Twenty gigabytes of storage for only US$229 ? this 3.5-inch hard disk drive (HDD),
released by Quantum Corp of the US in April 1999, would have been hard to imagine
even a few years ago. Approximately US$100 cheaper than similar drives from
competitors, the price is about the same as existing 13.6-Gbyte HDDs. In fact, Quantum
itself released a 12.7-Gbyte 3.5-inch drive at the end of 1998 for the same price. Even
this product will probably be out of the running in another few months, though, because
price competition for 3.5-inch HDDs is intensifying rapidly (Fig 1).
Prices are being forced down through the growth of two markets. The first is the
desktop personal computer (PC) market, in which the drives find their major
application. The other is the emerging consumer electronics market. Both are highly
attractive markets in terms of growth potential and, as a result, manufacturers are eager
to make their presence felt; releasing an inexpensive drive before the competition could
mean life or death for a HDD manufacturer.
Low-End PCs Catalyst
The price war was kicked off by a change in the PC market. There has been an
enormous growth in the market for desktop PCs in the sub-US$1,000 price bracket,
especially in the US. This has forced PC manufacturers to seek out low-priced HDDs,
which were supplied as low-end models by HDD manufacturers. This approach cannot
last for much longer, however, because the sub-US$1,000, PC is becoming the most
common model.
According to a survey by PC Data Inc of the US, the sales volume of PC systems by
retailers in the US in February 1999 rose only 1% over the prior year, but the
percentage of sales for sub-US$1,000 PCs soared to over 60% of the total. Growth was
especially high in the under US$600 price class, which accounts for over 20% of sales
volume.
For HDD manufacturers, assuring product shipment volumes in the face of a saturated
market requires aggressive product development and marketing for low-priced PCs.
Shipping a low-priced HDD faster than the competitors to assure profit is an urgent
priority for manufacturers.
Consumer Electronics
While hurrying to develop low-priced HDDs, manufacturers have more than just
low-end PCs in their sights. There is an emerging market looming on the horizon ?
consumer electronics. Products for this market are finally becoming available and
include such products as image recorders with internal HDDs; these are being shipped
by Replay Networks Inc and TiVo Inc, both of the US.
HDD manufacturers have high hopes for the consumer electronics market, and expect
that it will eventually grow to rival the PC market. HDD manufacturer Western Digital
Corp of the US believes that internal HDDs will be used in consumer electronics
including television program recorders, Internet-connected home appliances and
set-top boxes. ?We expect that the market for HDD for products like these will be 10 to
20 million units a year through 2002, growing to 50 million units annually by 2005,?
said a spokesperson.
Home appliance manufacturers are more demanding about HDD prices than PC
manufacturers. Believing that they are headed for competition with existing home
appliances, they consider that the market will not develop without low prices. ?We
don?t think it will be possible to achieve widespread sales of HDD video recorders
unless the price is dropped to between US$299 and US$399. And to make that price,
we have to drop the OEM (original equipment manufacturing) price of HDDs to
between US$50 and US$70,? said a Replay spokesperson. This is confirmed by a
major Japanese home appliance manufacturer, who believes that releasing a HDD
recorder at the same price as existing video television recorder (VTR) machines,
namely ¾20,000 to ¾30,000, would require the purchase of HDDs at ¾1,000 or ¾2,000
apiece. In the US, the OEM price for a low-end 3.5-inch HDD is between US$100 and
US$200 ? still too great by a factor of ten.
Use of Latest Technology
As the market shifts towards lower prices, the methods used to achieve them are also
evolving. Until now, low-priced HDDs have used inexpensive components based on
outdated technology to achieve the lower cost needed, being unable to use expensive
cutting-edge components. The magneto-resistive (MR) heads that were state-of-the-art
a few years ago, for example, cost about 1.5-times as much as the thin-film heads they
replaced. Low-priced 3.5-inch HDDs began using MR heads in the second half of
1996.
MR heads were adopted in the 2.5-inch HDD field well in advance of 3.5-inch drives.
Compared to thin-film heads, MR heads offered increased recording density, and so
were clearly the right choice for 2.5-inch drives where small size and large capacity
were more important than price.
As a result, a major difference in recording density appeared between low-priced 2.5-
and 3.5-inch drives, which in terms of advancement was represented by a time lag of
one year. This situation is changing fast. Manufacturers are adopting the latest
technologies for the components used in even 3.5-inch HDDs, and recording density is
increasing to match pace with that of 2.5-inch drives (Fig 2). One head manufacturer
estimates that the time lag between the advancement of levels of recording densities
attained in 2.5- and 3.5-inch drives is now only about three months. The peak areal
recording density of the 20-Gbyte 3.5-inch drive released by Quantum was 5.2
Gbits/inch2, which is essentially the same as the 6 Gbits/inch2 of the latest 2.5-inch
models.
The key factor behind this was the appearance of the giant magneto-resistive (GMR)
head. Although it significantly boosts recording density, the GMR head costs about the
same as the MR head. The price difference between MR and GMR heads is less than
US$1, and the next generation of heads have been estimated to have a price difference
of under 10%. The major manufacturers who were slow to switch to MR heads in their
low-priced drives are now jumping on the GMR bandwagon. New GMR products have
already been announced this year by Maxtor Corp, Western Digital (both of the US),
and Quantum.
When the latest GMR heads are used, they enable a change in the methods used to
lower costs. Because recording density is boosted significantly compared with that
offered by MR heads, per-platter capacity rises. This means that fewer components are
needed to make a HDD with a given capacity, thus reducing cost.
In other words, achieving lower production cost in advance of the competition is
coupled with the benefit of offering the highest recording density in the industry. This is
a technique that was unthinkable back in the days when the MR head offered a new high
in recording density, but it is being put into practice by Quantum. Areal recording
density has been boosted about 35% over existing products, making it possible to
provide 20 Gbytes of capacity on three platters instead of four, and a major decrease in
cost.
Now that the appearance of the GMR head has made it possible to manufacture cheaper
products, there is no need to use previous technologies. From now on, it is the most
advanced technologies which will win in the arena of price competition.
Storage Doubles Annually
HDD manufacturers hope to maintain an annual growth rate in HDD storage capacity of
about 100%. Forecasts based on the current state of the industry would mean densities
of 10-Gbits/inch2 by mid-2000 and 20-Gbits/inch2 by mid-2001. These do not seem to
be unreasonable targets, and in fact even 40 Gbits/inch2 can be realized. At the same
time, manufacturers will continue to refine their cost-cutting measures. Component cost
slashing and manufacturing process reviews are already well under way (Fig 3).
?Improvements to existing GMR heads will easily enable them to handle densities of
20 Gbits/inch2 . We are still brainstorming for 40,? said a spokesperson for TDK Corp
of Japan. Fuji Electric Co, Ltd of Japan is at a similar stage, ?We have established
technology for 20-Gbits/inch2 recording media at the volume-production level, and the
technological barriers to the development of a 40-Gbit/inch2 technology have been
cleared.? The realization of 10- to 20-Gbits/inch2 is now just a matter of time.
HDD manufacturers are still pushing development as they always have. Prototype
heads and platters are used first to confirm that recording and replay are possible at the
target density. Based on that initial verification, details like the head positioning
mechanism and the integrated circuit (IC) for the required data transfer rate are settled,
and the product hits the streets several years later.
The 20-Gbit/inch2 drive is following the same path, and at the present stage
record/replay performance is being verified in the lab. In February 1998, Seagate
Technology Inc of the US demonstrated a drive that was able to record and replay at
16.3 Gbits/inch2. US media manufacturer HMT Technology Corp plans to release
results of record/replay tests at 20 Gbits/inch2 and above in the near future. The
technologies needed to implement 20-Gbit/inch2 drives are already available (Table
1). Thus, for example, the positioning mechanism will have to follow a track pitch of
0.5 to 0.6æm, but this seems to be quite possible by embedding microactuators in
sliders and head arms.
40-Gbit Longitudinal Limits
In recognition of changing circumstances, R&D targets are already shifting to the 20- to
40-Gbit/inch2 range. Many in the industry seem to agree that 40-Gbits/inch2 is possible
through an extension of existing technology, and that the biggest remaining hurdle is
overcoming thermal fluctuation after-effects. Even so, more and more engineers are
now saying attaining 40-Gbit/inch2 density will not present a serious problem.
The thermal fluctuation after-effect is a phenomenon whereby the magnetization on the
hard disk is reversed by heat. When this occurs at room temperature, it leads to
recorded data being lost. The longitudinal recording technology used in existing HDDs
is thought to have an upper limit of about 40 Gbits/inch2, imposed by thermal
fluctuation. This problem is one of the reasons why perpendicular magnetic recording,
and magneto-optical recording methods have received so much attention in recent
years; they are less susceptible to the phenomenon.
The increased priority placed on track density in recent years has led many in the field
to believe that the problem of thermal fluctuation can be averted (Fig 4). An increase in
linear recording density would magnify the effects of thermal fluctuation, but there is
little problem when track density is increased. The cause of the problem is the fact that
as the crystal grains in the disk?s magnetic layer grow smaller, they are able to hold
less and less magnetic energy. The smaller the grains, the less heat it takes to cause a
magnetic reversal. An increase in linear recording density requires that the grains in the
magnetic layer be made smaller, which makes them more susceptible to the problem,
but the grains do not necessarily have to be made smaller to boost track density.
Increasing the track density does, however, require higher replay head sensitivity. The
technology required is already established in this area as well, because improvements
to GMR heads address sensitivity. GMR head technology is already well on the way to
40-Gbit/inch2 technology. Toshiba Corp of Japan, for example, is developing a film
that they call the spin filter in hopes of achieving densities of 40 to 50 Gbits/inch2.
The key factor in improving the GMR head lies in increasing the sensitivity of the spin
valve film built into it. In principle, all that is needed is to make the free layer of the
spin valve film a little thinner, which will make it easier to change magnetic direction
with low-strength magnetic fields from the disk. Because this would mean even a weak
field could change the resistance of the spin valve film, it would provide better head
sensitivity.
Additional Cost Cutting
In parallel with efforts to boost recording densities, manufacturers are also working to
reduce costs through conventional methods, such as reducing the drive price by cutting
the cost of its components. If a disk has only one platter and one head, an increase in
areal recording density provides no price reduction.
Firms that manufacture drive components are hard at work reducing their own costs and
prices, and the device?s circuitry in particular seems a likely target for their attention.
The cost of the chips and boards together runs to above about US$20, but it should be
possible to drop this to about US$15 by around 2001 (Fig 5). The key method here
would be the integration of the functions that currently are implemented on multiple
chips, onto a single chip. Mechanical components like spindle motors can also be
manufactured at lower prices, albeit at a loss in performance.
There do not seem to be any obvious price-cutting methods available for heads and
platters, however. Each costs about US$8 at present, and what is key here to successful
competition is whether or not these costs can be cut by one dollar.
The trump card in reducing circuitry cost is single-chipping (Fig 6). Some HDD and IC
manufacturers have already begun the process. Seagate, for example, used an IC that
integrates the hard disk controller (HDC), microcontroller and buffer memory, in a
3.5-inch drive that was released in March 1999. Buffer capacity was from 256 to 512
Kbytes. IC manufacturer Cirrus Logic Inc of the US has announced a chip that integrates
a HDC, microcontroller and read/write signal processor. According to the company,
the chip has so far been adopted by two firms and should begin appearing in products
by the end of the year. Full-scale adoption will probably be in 2000.
IC manufacturers other than Cirrus, including Lucent Technologies Inc, Marvell
Semiconductor Inc, and Texas Instruments Inc (TI), all of the US, seem to be aiming for
about the end of 1999 to achieve the same. Cirrus used a 0.35æm complementary
metal-oxide semiconductor (CMOS) process, while the others plan to work with rules
from 0.18 to 0.16æm. ?Of course we can make them with our existing process
technology, but dissipation is just too great,? explains one IC manufacturer.
They are all, however, aiming to produce the circuits that Cirrus has single-chipped.
Proposals have been made to integrate buffer memory as well, but the enlarged chip
footprint is a problem. In many cases, it would be cheaper just to buy commercial
memory.
There is an alternate way to reduce circuitry cost; this being to remove the controller
from the drive entirely. If the HDC can be integrated into the PC chipset, then obviously
HDD component costs will drop even more. Intel Corp and Compaq Computer Corp,
both of the US are considering this possibility now.
At present, however, this approach does not seem very practical, because it would
require a total overhaul of the PC-HDD interface. The HDD?s internal processing of
the downstream of the HDC varies widely between manufacturers, and it is no easy
matter to define an interface. Even if a new interface could be defined, it would
probably be quite difficult to assure multivendor compatibility for personal computers
and HDDs, at least in the initial stage.
Cheaper Mechanical Parts
Mechanical components like spindle motors and voice coil motors can probably be
made cheaper if performance is sacrificed somewhat. Hard disk drives designed for
use in inexpensive PCs and home appliances will not require state-ofthe-art
performance, and so this approach becomes a practical option.
Seagate, for example, announced one product in March 1999 for low-end PCs.
Compared to a drive released at the same time for mid-range PCs, the per-platter
capacity and disk speed were the same, but he ?performance was different.? In
translation, this means that it featured average seek times as well as an average number
of platters. By lowering performance standards, the firm was also able to use cheaper
motors and other components.
Most products for low-end PCs have only one or two platters. With these limitations,
the price of the spindle motor can be dropped. The spindle motor developed by Sankyo
Seiki Mfg Co Ltd of Japan for use in low-end PC drives was designed with this
assumption, and uses an integrated hub-yoke unit that is considerably cheaper than
assemblies of separate hub and yoke parts. While a conventional spindle motor sells
for US$5 or US$6, the Sankyo motor is probably only about US$4.
Simplifying Inspection
According to industry experts, the only way to reduce the cost of heads and disks is to
simplify and improve the inspection process. For example, instead of inspecting every
unit, random sampling could be used. As quality needs to be maintained the
manufacturing process has to be stable enough to prevent any variation in quality.
According to Mitsubishi Chemical Corp of Japan, inspection accounts for 20% to 30%
of the cost of a hard disk, increasing the areal recording density causes a proportional
increase in inspection time, which means higher cost. In the future it will probably be
necessary to either scan the defects optically, or only inspect some of the tracks. Some
manufacturers have already implemented this method. HMT Technology of the US, for
example, checks disks optically for defects in the initial manufacturing stages, solving
problems upstream in the process.
Plastic Substrates
The use of plastic substrates for platters has also been proposed. A Mitsubishi
Chemical spokesperson said, ?If the industry demands platters at US$5 apiece, there
will be no choice but to switch to plastic. It will still be difficult to meet that price.?
However, practical application of the technology is still quite far off.
Glass substrates are not only cheap, they also have the advantage in that the same
high-precision servo information used in optical disks can be adopted. Sony is
developing this technology now, although as yet no manufacturers have adopted it.
The problem with plastic substrates is that they require considerable expertise to use. It
is difficult to achieve flat surfaces with plastic, which makes it difficult to reduce the
flying height. This is a major difference compared with existing substrates, and as one
media manufacturer said, ?American HDD manufacturers considered plastic a few
years ago, but gave up because of the many differences it has compared with existing
media.?
Another area in which this technology differs is that the magnetic film must be formed
at low temperatures, and it is difficult to attain the requisite magnetic characteristics.
These problems have yet to be resolved.
by Hiroki Eda, Takuji Imai
Websites:
Intel
Microsoft
RSA
Sony
Wave Systems |
