Re: Disk errors ...
On Wed, 13 Jan 2021 18:14:16 -0500
Michael Stone <mstone@debian.org> wrote:
> On Wed, Jan 13, 2021 at 02:53:58PM -0800, David Christensen wrote:
> >I used to think "a cable is a cable, color does not matter", but I
> >have experienced many storage hardware issues over the years that
> >were caused by red SATA cables. Other readers on this list have had
> >similar experiences. The explanation is chemistry -- the red dye
> >slowly corrodes the conductors and/or contacts. Since replacing all
> >of my red SATA cables with black locking 6 Gbps SATA cables, all of
> >my storage hardware issues have been due to failed HDD's.
>
> And I've had a heck of a lot of hard drives connected with red SATA
> cables without issues, to the point that I'm more than confident to
> not jump on SATA cable color as the answer to any question as well as
> being reasonably confident that people pursuing this are going to end
> up interpreting coincidences as strong correlations.
>
> A more likely source of problems is the fact that the SATA connector
> is actually spec'd for a really low number of cycles. If someone
> frequently plugs and unplugs SATA cables there's a good chance
> they'll introduce connection issues, and if someone has old cables
> that have been used in a number of differnent systems there's a
> chance they're done. Regardless of color. Of course if you replace an
> old worn out red cable with a brand new not-red cable... Insertion
> cycles also might not be an issue at all; there's a difference
> between what the minimum requirement is and how robust a particular
> implementation is--but you can't really tell just by looking at it.
> This is, however, one of the reasons that a lot of people insist that
> "SATA III" cables are special even though the spec (and testing) says
> that "SATA I" cables perform the same--they simply tried a worn-out
> cable. Cables that were completely out of spec in the first place
> probably account for the rest, but again bogus junk can come in any
> color.
>
It is important to remember that all electrical signals, HDMI, SDI and
the rest of the alphabet are *analogue* signals. There is no such thing
as a digital signal. And that all of these modern 'digital' signals use
very high frequencies and are pushed down cables of a kind which were
never really intended to handle those frequencies. 4K television down a
single coax runs at 12Gb/s. Once, that would have had to travel down a
waveguide, and it still does if you actually need most of the input
power to reach the other end. 3Gb/s HD-SDI is routinely pushed through
cables that have a flat frequency response and reasonable group delay up
to 5MHz.
The neat, tidy square waves that the textbooks show are a long way from
reality, and receiver ICs have to cope with the real-world stuff. It
only takes a tiny shift of voltage with temperature to move detection
of 1s and 0s from 'rock solid' to 'most of the time'. It is also
impossible to properly terminate a cable at high frequencies without
using a lot of components and alignment equipment, so there will always
be small reflections added to the mix. The poorer the standard of the
cable, the more loosely its alleged characteristic impedance is
defined, and will anyway be messed up by a little corrosion between the
connectors. Gold plating helps but is not a complete answer, and the
thinner the plating, the sooner it gets rubbed away by repeated
connections and disconnections, as Michael said.
--
Joe
Reply to: