Group Actions
Engagement
  • Author Author: dougw
  • Date Created: Date Created
  • Views 2140 views
  • Likes 15 likes
  • Comments 12 comments
Related
Recommended

Nightmare Disk Failure & Recovery Attempt

dougw

A few months ago I purchased a 2.5" 2 TB USB hard disk drive mostly for backup, but I started using it for project work that had lots of video in it. A few days ago it wouldn't power up with my computer - it just made an intermittent beeping sound. After some research I decided this was probably a situation where the heads got stuck to the disk. I haven't seen any complete analysis of why or how heads get stuck to disks, but here is my speculation: Heads normally stay in their parked position beside the disk until the disk is up to speed. When the heads launch onto the surface the boundary layer of air at the disk surface, travelling with the disk, forces the heads to fly a few nanometers above the surface using something like wing-in-ground-effect. To counteract the lift force the heads must be pushing down due to a weak spring, because they can operate on their side. If the disk stops spinning while the heads are above its surface, the heads will contact the surface, because the lift force of the air disappears. There is about 1 nm of lubricant on the disk surface, which allows a perfect air-tight seal of the head to the surface, so air pressure clamps the head to the disk just like a suction cup. On a small disk like the one I have this air pressure creates enough friction to prevent the small motor from spinning the disk. There must be some smarts in the disk that senses an overload and intermittently tries to restart, hence the beeping sound.

image

Hopefully the heads got stuck just as the disk stopped rotating and didn't scratch a groove in the disk. The following video shows my attempt to dismantle the disk drive and un-stick the heads. I'm not recommending anyone else should or shouldn't do this, it is just a story of what I did.

 

This procedure worked perfectly for me. The disk fired right up and I immediately copied all files to another disk - about 1 TB of data.

I do not expect to use this disk again, but I will keep it around as the data on it represents some sort of last resort extra backup for the files that are already on it.

I did have one totally unexpected thing happen. After I copied all the files off the disk to another external hard drive and was shutting the system down, the computer went through its normal shutdown process and turned off. However, both the recovered drive and the new drive continued to spin. Both drives are powered via USB but these USB ports do not have power when the computer is off, so it is a mystery where the power was coming from to keep the drives spinning. I let them spin for a couple of minutes to see if they would shut off before yanking both USB cables. Of course this was quite worrisome as it could have caused another head crash on both drives. They definitely shut off when the cables were pulled. That is a new one for me.....anyone out there ever see something like that?

I also had never heard of stuck heads before this happened, but after researching the possible failure modes based on the symptoms, and watching some Youtube on the subject, I decided it was worth trying to fix myself. Although it isn't actually a fix, it is just a recovery process. I do not have a statistically valid sample to make a guess at the likelihood of success or failure because I have only tried it once and it worked 100% as a recovery process.

Parents

  • Stiction is not uncommon between ultra-smooth heads and the surface. If the drive loses power due to a poor connection and there isn't enough back-EMF from the spindle motor and stored rotational energy in the spindle to force the heads to park back onto the ramp (through an "uncontrolled" unload), the heads will stay on the disk as the disk comes to a stop which creates the risk of the heads sticking to the disk so well that the spindle motor cannot overcome the force to unstick them.

     

    This is something I saw several years back once ramp load-unload drives started hitting the market and were used in external cases. Sometimes people would let the drives partially spin-up and then unplug them - if the heads had been loaded onto the disk but the disk wasn't quite up to speed, often they could end up stuck. I have removed the top case and used a screwdriver to free the media from the head which was enough to get it going again for a friend. Of course, once violated, the clean atmosphere inside the disk is no longer guaranteed so the drive is not likely to remain reliable for long.

     

    Others who are resistant to opening the drive have found that a good firm "slap" or drop can free the heads just enough for the disk to get going. However, it's a delicate balance - too much of a hit and you'll either chip a head due to the "slapping" motion, or cause disk shift (where the disks move relative to the spindle) causing run-out error instead.

     

    The "beeping" noise is basically the spindle motor driver's "soft start" waveform into the motor coils, and its magnetostriction causing audible noise generation. Of course, with the monitored and current-limited condition of the motor driver IC, it will cut out and retry a fixed number of times before stopping until power is removed and cycled.

     

    In older contact-start-stop (CSS) disks where the heads rest on the platters when parked, the inner tracks designated as a landing zone have laser texturing applied to them so as to make them less sticky, however, they had a much more limited takeoff/landing cycle life as the friction during the spin-up process gradually wore away at the sliders causing their aerodynamics to be slowly altered until they couldn't "fly" in a stable configuration or height.

     

    - Gough

  • in reply to Gough Lui

    Thanks for the info . I find it amazing how much sticking force is applied just by having the tiny heads touch the disk surface.

    I guess air pressure would yield about 41 grams of vertical force. The coefficient of friction on an ultra smooth lubricated surface must be pretty small. I gather there is a very thin layer of carbon on the surface. The coefficient of friction for carbon is .16 for dry and .13 for lubricated, which implies the maximum stiction force might only be 7 grams. That is a pretty weak motor if its starting torque is less than 20 g-cm.

  • in reply to dougw

    - To me, it's not that great of a surprise as it might initially seem. I'm not sure if you've ever tried sliding two pieces of clean glass against each other - they really do "bind" and resist movement quite strongly. The smoothness of the surface being very uniform can increase the bonding between surfaces due to electrostatic and Van der Walls forces working together. That's how some "plastic film" adhesive whiteboards/protective plastic films work as well - try sliding these off rather than peeling them off and you'll appreciate just how much stiction can really get in the way.

     

    But more than this, hard drive manufacturers push the boundaries. For a 2.5" external drive, legally you might get about 2.5W - 5W of power to apply to the logic circuits and spindle motors. Where this traditionally on a 9.5mm drive would be expected to break the stiction on four heads (two platters) with heads resting in the landing zone (laser textured), newer 12.7mm storage drives are built with as many as eight heads (four platters) and no landing zone (heads are expected to rest on the ramp unloaded and never touch the disk). The increased mass of the disks alone with the limited power budget likely restricts the start-up impulse that can be developed that might "break" the heads free from the surface if they should (unfortunately) ever land on the smooth disk.

     

    Don't forget also that the developed "torque" also depends on where the head lays to rest on the disk - assuming equal adhesion, towards the outer edge and the mechanical advantage works against the spindle motor (if I'm thinking about this correctly) to require more torque than if the heads rested closer to the inside of the disk.

     

    As a result, a safer alternative would be to work out where the spindle motor is, and give it a good sudden rotational jolt - like "spinning" a bottle in your hand and stopping it suddenly. I've had some occasional success with this method compared to hitting/dropping the unit, although opening it up and moving the spindle using a screwdriver seems the sure-fire bet.

    - Gough

Comment
  • in reply to dougw

    - To me, it's not that great of a surprise as it might initially seem. I'm not sure if you've ever tried sliding two pieces of clean glass against each other - they really do "bind" and resist movement quite strongly. The smoothness of the surface being very uniform can increase the bonding between surfaces due to electrostatic and Van der Walls forces working together. That's how some "plastic film" adhesive whiteboards/protective plastic films work as well - try sliding these off rather than peeling them off and you'll appreciate just how much stiction can really get in the way.

     

    But more than this, hard drive manufacturers push the boundaries. For a 2.5" external drive, legally you might get about 2.5W - 5W of power to apply to the logic circuits and spindle motors. Where this traditionally on a 9.5mm drive would be expected to break the stiction on four heads (two platters) with heads resting in the landing zone (laser textured), newer 12.7mm storage drives are built with as many as eight heads (four platters) and no landing zone (heads are expected to rest on the ramp unloaded and never touch the disk). The increased mass of the disks alone with the limited power budget likely restricts the start-up impulse that can be developed that might "break" the heads free from the surface if they should (unfortunately) ever land on the smooth disk.

     

    Don't forget also that the developed "torque" also depends on where the head lays to rest on the disk - assuming equal adhesion, towards the outer edge and the mechanical advantage works against the spindle motor (if I'm thinking about this correctly) to require more torque than if the heads rested closer to the inside of the disk.

     

    As a result, a safer alternative would be to work out where the spindle motor is, and give it a good sudden rotational jolt - like "spinning" a bottle in your hand and stopping it suddenly. I've had some occasional success with this method compared to hitting/dropping the unit, although opening it up and moving the spindle using a screwdriver seems the sure-fire bet.

    - Gough

Children
  • in reply to Gough Lui

    Good points. It is the small size of the heads that make the force amazing. I can see how a torsional shock could start the disk moving, but unless there is power on the device, I don't think there is much force to move the heads over to the side. Maybe torsion shock combined with acceleration in the correct direction to move the heads over would work better .....

  • in reply to dougw

    The thing is that stiction doesn't really achieve its maximum adhesion "immediately". Very often, if it's a perfectly clean disk and the heads have only rested for a very short period, the drive may be able to get going again. But left (unknowingly) for a while, changes in humidity and temperature with the constant downward pressure of the head suspension would keep it in its place, slowly "welding" it to the surface and potentially damaging the diamond-like coating.

     

    Instead, torsional shock when the drive is attempting to spin up might give enough of a momentum boost to break the heads free. Then, resting on a "fresh" surface, it would not be as well adhered. Better yet, if the spindle motor is also giving a kick at the same time - now you need to overcome the coefficient of kinetic friction (assumed to be lower), rather than static friction thus the drive should get to speed enough for the heads to start flying on its own.

     

    Of course, slapping, shocking the drive etc is not intended to get the heads loaded back onto the ramp - merely break the temporary "welding" of the heads to the surface for just enough for the static friction to be overcome.

     

    - Gough