December 09, 1996, Issue: 1036
Section: Computers & Multimedia
MR Not The Only Drive Issue
By Alex Naqvi
The hard-drive industry has been able to deliver products with annual
60% increases in capacity and quarterly 12% decreases in cost per
megabyte.
As EBN pointed out in its Sept. 30 article "Clamor For MR Heads
Threatens Drive Supply," MR heads are an important factor in increasing
storage density. An equally important factor, however, is read-channel
technology.
For more than 20 years, hard drives have used peak-detect read-channel
technology. Peak-detect technology indicates peak magnetic energy
passing below the read head as it rotates near the rotating disk surface.
As drive makers push for higher bit densities, the separation between bit
peaks grows smaller, thereby increasing the likelihood of peak-detection
errors. Though MR heads are capable of directing lower-energy peaks,
the spatial separation problem still confounds peak-detect read channels.
That's why today virtually all disk-drive manufacturers are designing
products with a different read-channel technology. Called partial
response, maximum likelihood (PRML), the technology promises to
sustain increased densities and speeds well beyond the limits of peak
detection.
Using sophisticated predictive algorithms, PRML technology identifies the
most likely bit sequence in a captured read-signal sequence. Close
spacing of peaks has a much smaller effect on the accuracy of PRML
results. Thus, by switching to PRML technology, disk-drive makers
ensure being able to meet the market's capacity and performance
expectations.
New disk-drive prices are a different matter. All of today's PRML ICs
are mixed-signal devices. Those whose functions are primarily designed
using analog technology are referred to as analog PRML devices.
Similarly, those whose functions are primarily digital are called digital
PRML devices.
Analog PRML devices have been around longer and have typically been
faster and lower in cost. Digital PRML ICs were developed later and
offer much more functional richness. The early digital PRML trade-off,
however, was slower speeds and higher costs. That disparity no longer
exists.
By designing digital PRML solutions using mainstream, digital CMOS
technology, the price gap between analog and digital PRML ICs has
narrowed considerably. The fastest read-channel speeds are now claimed
(and verified) by digital PRML ICs.
Channel integration is the time and cost required for matching disk, heads,
and electronics in a disk drive. Along with the quest for higher capacities
and speeds have come tighter tolerances and higher channel-integration
costs. But by using a digital PRML IC solution, it is possible to design a
self-calibrating drive.
The remaining challenge is pre-amp technology. While MR heads can
detect lower-level peaks, these electrical signals must be amplified before
they are processed.
With the disks rotating faster and with the bits closer to one another, the
period of the signal impulses becomes shorter and shorter. Since every
amplifier has some rise and fall time delay, it is possible for the period to
be of the same order of magnitude as the rise and fall characteristics.
Thus, pre-amp technology will have to be designed in such a way that rise
and fall characteristics remain just a fraction of signal pulse periods.
All of these challenges are currently being met, and the outlook is bullish
that disk-drive makers will be able to continue to deliver increasing
capacity and lower per-Mbyte prices.
Furthermore, thin-film heads in conjunction with PRML read channels can
continue to support these increased capacities, speeds, and performance
without undue reliance on MR heads. This will allow MR head technology
and production to proceed at its own pace without necessarily threatening
the supply of next-generation drives.
-Alex Naqvi is vice president of marketing for Marvell Semiconductor
Inc., Sunnyvale, Calif.
Copyright r 1996 CMP Media Inc. |
