DC fans dead after hot disconnect / connect.

[Doug Warner]

Three of these fans:
Nidec TA150CD C34957-58 40 x 28
http://www.nidec.com/ta150dc28/ta150dc28.htm

in a 1U rackmount server failed after a customer hot-unplugged and
re-plugged each one in an attempt to determine which one was noisy.
Now, none of them work even when conencted to a different system.

A 4th, thinner model of the same fan survived, and works if connected
to any of the failed fans' connectors. It continues to run even after
many hot connect cycles.

In the support center, we tried the same thing on a lab system, and
the same three fans died there as well. Also, the connectors are
keyed, so it would be extremely difficult to connect them backwards,
and it's unlikely that two different people would have made the same
mistake.

They are speed controlled, but I don't know if it's analog or PWM.
One theory is that the open circuit voltage floats above 12V, but the
highest voltage from the power supply is 12V, and I doubt they'd do
any step-up tricks just to run 12V fans. (I'll confirm when I get a
meter on the supply pins tomorrow.)

Any thoughts on this one?

To reply, please remove one letter from each side of "@"
Spammers are VERMIN. Please kill them all.

 

[legg]

On Wed, 20 Jul 2005 00:10:51 -0400, Doug Warner
<dwarner22@ccharter.net> wrote:

Three of these fans:
Nidec TA150CD C34957-58 40 x 28
http://www.nidec.com/ta150dc28/ta150dc28.htm

in a 1U rackmount server failed after a customer hot-unplugged and
re-plugged each one in an attempt to determine which one was noisy.
Now, none of them work even when conencted to a different system.

A 4th, thinner model of the same fan survived, and works if connected
to any of the failed fans' connectors. It continues to run even after
many hot connect cycles.

In the support center, we tried the same thing on a lab system, and
the same three fans died there as well. Also, the connectors are
keyed, so it would be extremely difficult to connect them backwards,
and it's unlikely that two different people would have made the same
mistake.

They are speed controlled, but I don't know if it's analog or PWM.
One theory is that the open circuit voltage floats above 12V, but the
highest voltage from the power supply is 12V, and I doubt they'd do
any step-up tricks just to run 12V fans. (I'll confirm when I get a
meter on the supply pins tomorrow.)

Any thoughts on this one?

Typical hot plug sequencing problems involve mating + and signal
first, before ground. This makes the fan see the signal lines driven
strongly negative, with respect to all its internal references.

Ground should mate first, break last, in a hot plug connector of any
supply polarity.

Diode-clamping fan signal lines to fan 0V, on the fan side of the
hot-swap connector might reduce damage, but there are no guarantees.

RL