Low level SATA failure?

Do you have a question? Post it now! No Registration Necessary.  Now with pictures!

Threaded View
Lets examine the low-level physics of a SATA hard-drive.

Surely the C/H/S is fixed, like the ManufctrID, and is read,
even if the device fails to turn?

Is it true that a HD, which is moved while running, experiences
centrifugal forces [gyroscopic] which when forced, by eg. the
humans hand, is likely to damage the device?

To move the device, it should first stop and park.
Then the gyoscopic forces have stopped.

Or can it tolerate movement?


Re: Low level SATA failure?

On 4/5/2015 9:04 AM, noSpam@gmail.com wrote:
Quoted text here. Click to load it

Most hard disks, with the claimed exception of Toshiba, store their  
firmware on the platters. The portion of permanent memory on the circuit  
board is enough to read the on-disk firmware. i.e. a boot process. How the  
drive responds is dependent on how far the boot goes and may be something  
or nothing.

Yes, in theory a spinning drive can fail from gyroscopic forces. The faster  
the rate of rotation, the faster the spin, the larger the radius, the lower  
the flight height the worse it can be. A huge capacity high-RPM drive being  
rotated forcefully perpendicular to the plane of rotation seems to be the  
most vulnerable. Of course rotating the drive around the spin axis doesn't  
cause any noticeable effect but that would be pretty rare to have a pure  
rotation like that. Personally I've never suffered a failure even with 3tB  
drives and I seem to forget to wait for spindown as often as not.

Re: Low level SATA failure?

Quoted text here. Click to load it

But C/H/S is only 3 bytes or so.
And how can it START to "read the on-disk firmware",
if C/H/S is unknown ?

Re: Low level SATA failure?

noSpam@gmail.com wrote:
Quoted text here. Click to load it

CHS is known as "fake geometry" for a reason.

Modern drives, the real transactions are based on LBA
(logical block address).


CHS is used for addressing purposes, by old OSes. Perhaps
your DOS boot floppy, is using CHS at some point. I wouldn't
know, as I don't have a full copy of DOS here.

Modern fake geometry is:  <some_large_number> 255 63.

The 255 is "heads", and there are not 255 heads on the drive.
(My hard drive actually has 8 heads right now.)

Neither it is true that there are 63 sectors on the track.
There are a *lot* more sectors. The sector layout isn't
even uniform - it's "zoned". So CHS is such a joke (of
no physical significance at all), that the topic isn't
even funny. Only the drive controller knows "the map of
the city", and given the LBA property tax number,
it knows what street and plot of land that is.

The switch from 28 bit to 48 bit LBA, made a big
difference to what you could address. This doc is
the original proposal to the committee, on extending
the addressing for modern drives (2003 era). While the
register names make imaginative reference to CHS,
the number loaded into it seems to be a continuous
logical number instead. Sector is an 8 bit number,
Cylinder is a 16 bit number, Head is a 4 bit number,
giving the 28 bit (137GB) disk drive size limit. The
move to 48 bit LBA in 2003 or so, extends the addressing
for the future. The best part of this, is the proposal
is so clever - the double pumping of the registers,
it's genius.



The only enemy of hard drives is

    Head crash!

Any manipulation which causes the flying head to contact
the platter, "dings" the surface, and causes garbage
to be thrown up. You can use your physics class
notes, to figure out which axis might be the worst
for shocking the setup, and making the head contact
the platter.


Site Timeline