PSU ?

philo wrote:
On 7/1/19 9:59 AM, John McGaw wrote:
On 6/30/2019 7:00 PM, philo wrote:
On 6/30/19 9:29 AM, John McGaw wrote:
On 6/26/2019 9:27 AM, philo wrote:
I've repaired thousands of machines and once in a while have seen a
failed power supply take out a motherboard.

Only once though have I seen the hard drives taken out.

It was with a good quality supply and I thought it was just a very
odd fluke.


Last night I talked to someone who had also had it happen ,....so
my question is:


Does any manufacturer have a supply with fully surge-protected OUTPUT?


I assumed they all did.

I, too, have had one PSU misbehave and take out a hard drive but
that was a very long time ago and with a no-name PSU. I learned my
lesson after that and now use nothing but Seasonic supplies, at
least in situations where they make a supply that will fit. Right
now the only computers I own that don't have Seasonic are my Shuttle
cubes and they are tied to the form-specific equipment. Luckily the
Shuttle PSUs have proven to be faultless also.

I don't know of any maker that installs crowbar protection on their
supplies although I believe it would take little more than a
high-current SCR a Zener diode and a few other passive components to
work the magic.



All I know is that for industrial use MOV's were everywhere and they
worked.

That said, If I try my own modifications I'll do it on a non-critical
machine first.


Will also look into Seasonic PSU's, thanks

Oh, I've use MOVs for various things but the problem is that they
aren't all that precise. Great at absorbing and dissipating lots of
power but the voltage at which they operate are vague at best. I've
used bushels of them protecting broadcast and computer equipment from
line transients but I'd be hard pressed to set them up to protect a
12V circuit from going above 13.5V where damage to a hard drive might
be expected to begin. A glance at some spec sheets suggests that using
a 12V MOV the clamping voltage might be 40+. But if you decide to give
it a shot and make it work well, please let up know.




Well, if I install MOV's and never have a HD burn out again it still
does not prove they did the trick.


I have a few other ideas though.


The person I know who lost two drives had a RAID 1 (mirrored) so he
thought he was safe.

Now I see that one possible way to build a reliable machine would be to
have each drive powered by a completely separate supply and probably
from a different manufacturer.


That said, it's been my experience in life that no matter what you plan
for, something else is going to happen!

Maybe I'll just continue doing backups because to reinstall the OS is
not a big deal...all that's important it the data

Even the RAID controller itself can fail and damage an array.
We lost two RAID5 arrays at work due to that. The first failure,
there were no "lessons learned" - the staff just restored and
pretended nothing happened. The second failure prevented hundreds of
employees from working at 2PM in the afternoon. And *that* failure,
finally got some executive oversight.

One poster on USENET, recounts how a SIL3112 failed in an interesting way.
One day, one of the two drives in RAID1 failed. Fine, the thing
continued to work with the remaining drive (as intended). However,
the files on the second drive *were three months old*. In other
words, the (software) RAID stopped mirroring *three months* before
one of the drives failed. There were no warning signs or error
messages.

It's not normal for array products like this, to carry out any sort
of audit and warn you of that sort of problem (drive mis-match
while mirroring).

All that RAID really does, is introduce more failure modes.

So even if you power the drives separately, put the drives
in different computers so the SATA ports are separated,
there is still the danger of a firmware or software failure
failing to meet functional requirements.

Paul