Connect with us

harddrive read-head arm

Discussion in 'Electronic Design' started by Jamie Morken, May 2, 2004.

Scroll to continue with content
  1. Jamie Morken

    Jamie Morken Guest

    Hi all,

    How does a harddrive readhead arm move so precisely to be able to read a
    microscopic track of data on a harddrive? I took apart a harddrive and the
    read head has a single coil of wire on it, and there is a NS magnet pair on
    each side of the coil of wire. The read head arm goes back and forth
    between the magnets when the read head coil is energized, the direction
    dependant on the polarity of the battery. I was thinking that the way it
    must work is with an H-bridge driving the coil in locked antiphase, or maybe
    the coil in controlled with an opamp? Any ideas on this?

    Jamie Morken

    read head arm:
    < hardrive read head actu

    read head arm and two magnets:
    < hardrive read head actu

    read head arm between magnets:
    < hardrive read head actu
  2. Ian Stirling

    Ian Stirling Guest

    The drive reads the signal coming off the head for tracking information.
    Back when the largest drives were 60Mb, there were a number of ways.
    From one surface in the drive containing track information, to an optical
    encoder reading it out, to stepper motors that moved in discrete steps.
  3. Jamie Morken

    Jamie Morken Guest

    I assume for the new larger drives that the current track is read by the
    read head magnetically. What I don't understand is what type of signal is
    being sent to the coil that moves the read head arm back and forth so

    Jamie Morken
  4. Garrett Mace

    Garrett Mace Guest

    I assume for the new larger drives that the current track is read by the

    It's just a precisely-controlled current. The force between the coil and
    magnet is proportional to the current flowing through the coil. I haven't
    taken apart a hard drive any larger than 1 gig, but I believe it used a
    multipole magnet with the coil balanced between poles. Supposedly there are
    types that balance the magnetic force against a spring, but I don't know if
    that would be as fast as the multipole configuration.

    Here's a datasheet for an old voice coil driver, it should give you some
    more hints on what kind of signal control a hard drive voice coil requires.
  5. Tim Wescott

    Tim Wescott Guest

    The position of the head is read off of the "factory formatting" of the
    drive; the voice coil is driven in pure current mode so the arm
    accelerates proportional to the drive current. Then the whole thing is
    wrapped with a PD or a PID controller to bring the arm to the correct
  6. Mike

    Mike Guest

    No, the read head is that little tiny thing at the end of the arm. These
    days, the heads are magneto-resistive, and have no coils for reading. The
    voice-coil, which I think you're describing, is what moves the heads back
    and forth. The arm is part of what used to be the head stack, but these
    days most drives are only one or two heads, so I'm not sure if they still
    call it a stack.

    The servo information used to position the heads is written in between the
    data sectors. As the read head moves across the tracks, it can detect the
    servo information, which contains information about which track the head is
    currently flying over. When it gets to the right track, the servo stops
    moving the head across tracks, but at this point the head is still not
    positioned precisely. To complete the precise placement, the servo
    information is written in a way that part of the information is off-track
    in one direction, and part of it is off-track in the other direction. The
    servo algorithms have become quite sophisticated, but the general idea is
    that if the head is off-track in one direction, one of the off-track servo
    bursts will have larger amplitude when it's read than the other one. So, as
    the head moves across the servo bursts, the off-track amplitudes are
    detected and subtracted from each other to create an error signal. If the
    head is positioned precisely in the center of the track, the error signal
    will be zero. The error signal is used to drive the voice coil
    (electrically, at least) to a high degree of accuracy.

    There are mechanical issues as well - probably even larger than the
    electrical issues - but I've never been involved with the mechanics, so I
    can't say anything knowledgable about that.

    -- Mike --
  7. Ian Stirling

    Ian Stirling Guest

    My hard drive has 160G, and does 50Mb/s.
    Divide by 120 (7200RPM) gives 416K/track.
    Assuming that it's got one platter, then that's 832K/cylinder, or about
    144000 tracks.

    200nm track spacing.
    If the track has an average length of 8cm, then that's 100K/cm, or
    bit length of 100nm.


    If you could get a printer that did this resolution, then you could
    print perfect holograms.

    Speed goes up as the square root of capacity (if the drives are at the same
    RPM, and the tracks density increases at the same rate as the bit density)
    this leads to being able to read a 60M disk in a couple of minutes to a
    160G drive taking the better part of an hour.
  8. Tim Wescott

    Tim Wescott Guest

    True, but does that help the OP?

    It doesn't take much more than a PID loop and good mechanical design to
    get the kind of performance that they get with disk drives -- do you
    know what additional measures are taken? Do they just add the
    obligatory 1st notch filter and a lead-lag, or do they get wild and have
    fully adaptive controllers?
  9. Tim Wescott

    Tim Wescott Guest

    Hey Mike:

    Once they know the position, what sort of algorithms do they use to
    locate the arm? I've closed a lot of loops with speaker coils and
    torquer motors, I'm curious what one does when one can justify lots of
    engineering time with high production numbers.
  10. Mike

    Mike Guest

    Dunno. Servo is divided into two parts. The first part takes place in the
    channel, where the data is read from the disk and the amplitude is
    extracted. Simple algorithms were okay until around 1995 or so, when the
    track density reached the point that the errors and noise associated with
    simple sample and hold peak detectors got to be too large (that's the last
    time I worked on a servo detector). The problem with peak detectors is that
    noise is biased - one polarity adds to the peak, while the opposite
    polarity is ignored. You can filter the noise before it gets to the peak
    detector, but sooner or later more sophisticated detectors are required.

    The second part is the voice coil driver that moves the heads back and
    forth. My recollection is that the actual servo system isn't that
    complicated, but I could be mistaken. The complexity comes when the head
    has to move rapidly across multiple tracks during a seek (non-linear), then
    slow down and finish the acquisition (linear), all in the shortest time.
    Additional complexity comes in the head-parking circuits. When the drive
    powers down, the back-emf in the motor coils is used to drive the servo
    long enough to move the heads onto the landing ramps.

    -- Mike --
  11. Mike

    Mike Guest

    The data rate on the outer tracks is significantly larger (close to 3X,
    IIRC) the data rate on the inner tracks. The limiting factor is transition
    spacing - on the outer tracks the speed of the disk is faster, so data can
    be written faster for the same lineal transition density.

    Here are some specs from a current Seagate Cheetah 15k drive:

    Bytes Per Sector 512
    Track Density (TPI) 64,000 tracks/inch
    Recording Density (BPI, max) 533,000 bits/inch
    Bytes/Track (avg) 377,770

    They have a press release about a new storage record they set as well:

    Areal Density: 101 Gb/in2
    Track Density: 149k TPI
    Bit Density: 680k BPI
    Data Rate: 256 Mb/s
    Bit Aspect Ratio: 4.6
    On track raw error rate: 5 X 10-5
    Raw error rate @ 5% squeeze and 10% OTC: 1 X 10-4

    Check out the raw error rate. On average, a modern drive will contain
    roughly 1 in 10^4 bits, or about one error per sector. It gets corrected
    back to acceptable levels using coding.

    -- Mike --
  12. Terry Given

    Terry Given Guest

    I think if you look closely you'll find that the position control is now a
    lot more sophisticated than "mere" PID control. Precision is amazing, and
    bandwidth is pretty darn high too (If only I could recall the numbers.....).

    Ultimately it is just the "magic" of extremely well designed closed-loop
    control (ie negative feedback)
  13. Terry Given

    Terry Given Guest

    Alas, I cant remember much about the papers I have read on this subject re.
    advanced control. And of course just because people write papers on things
    doesnt mean that it gets done that way in practice.

    Hopefully someone in the know will elucidate us - Nice work though Ian. The
    numbers really are impressive.

Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day