[Bug-ddrescue] How freezing helps?

[Jarkko Lavinen]

How freezing a hard drive is supposed to help? Does it make some
bad blocks working again? Or does it make some bad blocks less
bad, susceptible for retries? Or does it bring dead drives back
to life like a magic?

Trying to understand and speculate on how freezing could help:

- Cold air is denser than warm air and could lower the head
  flying height and thus make reading easier. Air density is
  1.3943 at -20 C and 1.1455 at +35 C. 
  Source: http://en.wikipedia.org/wiki/Density_of_air

- If the drive uses thermal fly-height control, cold reduces the
  heater capability to bring the head closer the surface. If the
  cold exceeds the heating power, the head flies higher than was
  designed.
  
http://www.hgst.com/tech/techlib.nsf/techdocs/98EE13311A54CAC886257171005E0F16/$file/TFC_whitepaper041807.pdf
  http://maeresearch.ucsd.edu/callafon/publications/2011/UweIEEETonM.pdf

- Cold could make preamplifiers more sensitive by reducing noise.
  I wonder if -20C is cold enough to have effect on hard drive
  head preamplifiers? For example in some radio telescopes the
  preamplifier are cooled with liquid nitrogen.

  I speculate that some bad blocks are just too weak and noisy and
  require either more retries than drive can withstand or better
  signal to noise ratio ie. more correctly read bits.

- Cold changes mechanics. Warped platters become less
  warped. Lubricants become stiffer.

I have not yet tried freezing but probably will, once the drive I
am currently trying to rescue reaches the point where I cannot
fetch any more data at room temperature.

After 2 weeks of running >99.9% is rescued and hoping to reach
>99.99% before the freezing attempt.  The splitting is over
100000 times slower than the initial reading of working areas.

I am planning to use a plastic bag filled with silica gel grains
and cold packs.  The idea is to keep the drive cold as long as
possible and at the same time as dry as possible. Cat litter is
the cheapest source of silica gel and the idea came from this
discussion at Slashdot:
http://hardware.slashdot.org/story/08/05/12/2330200/a-walk-through-the-hard-drive-recovery-process

Jarkko Lavinen