Hands-On Testing and Analysis

Hard Drive Evolution Part 3


Stepper Motors to Voice Coils

The first 5 ¼” hard drives like the ST-506, used a stepper motor to move the heads in and out. The head carriage was attached to a split steel band so each step of the motor in or out would define a track.

Stepper motors were cheap, but even by the measure of those washing machine drives painfully slow with average access times of 200ms or more. Even worse since there was no feedback loop that told the drive, or controller, where it was on the disk, tracks had to be far enough apart to allow for any slop in the mechanical system some of which was caused by the strap expanding as the drive heated up.

The Priam 14” disk drive that was our high-performance solution in the early ‘80s didn’t use a stepper motor but a series of magnets and voice coils much the same way the voice coil moves a speaker’s cone in and out to play “The Boogie Woogie Bugle Boy of Company B.” The voice coil was a lot faster than a stepper motor, delivering seek times of 40 vs. 200ms, in part because it exerted significantly more force.  We set one up on a conference table to demonstrate to a customer and when we started it seeking the whole table rocked back and forth with the head motion.

That old Priam used a linear voice coil that drove the heads directly in and out. Today’s drives use rotary voice coils that move the heads like a record player’s tone-arm.  If you young’uns don’t understand what a record player is, ask your parents, they’ll be happy you called.


Modern Rotary Voice Coil

The other big advantage of a voice coil positioner is the closed feedback loop created as the drive can constantly read the servo data, essentially a constant stream of “this is track x” recorded on the bottom surface of the disk stack. The servo signal specially recorded at the factory, so the signal is strongest in the center of the track. By reading the value and strength of the servo signal and making minute adjustments, the drive can pack tracks closer together.

Eventually, OK the ‘90s, track density reached the point where the flex in the head assembly between the top data head and the bottom servo head was enough that a dedicated servo surface couldn’t provide accurate enough position information and the drive vendors switched to embedding servo data in between the data sectors on every surface.

This change to embedded servo meant that the drive had to adjust the head’s position when switching from head to head as well as when moving from track to track. With embedded servo, the old model of drive cylinders no longer fit how disk drives actually worked.

Getting Smaller to Get Faster

Competition from upstarts like Seagate and Miniscribe wasn’t the only reason disk drives shrunk from 14” in diameter to 3.5”. After all larger platters present more surface area, so if the technology of the day would support x bits per square inch (b/in2) a 14” platter would hold many times as much data as a 3.5” one.

As vendors gave up on building SLEDs with multiple positioners like the IBM 3380 that had separate heads and positioners for the inner and outer tracks of each platter they needed to find some other way to create faster, as well as bigger disk drives.

Smaller disks made it easier to make faster drives in two ways. The first is simply that smaller disks take less energy to spin. The amount of power needed to spin a disk is a function of the square of the disk’s radius. A 5 ¼” disk takes less than an eighth the power of a 14” drive, a 2.5” drive about 1/20th.

No one should be surprised then that we’ve gone from the 4200 RPM of the old 14” Priam to the 2.5” 15K RPM disks that SSDs have now made obsolete.

As the disk shrinks, so does the “tone arm” that supports the heads and moves them from track to track. Smaller head/positioner assemblies have less mass which means they can accelerate, and decelerate at higher rates reducing latency.