In article , "T" wrote:
Good info, (you must be in europe, 220v!) I will test the UPS circuitry. I
think I need to find a 'sensor' to detect the charge build up on the boards.
Somehow if the supply voltage is not the nominal 120vac and is instead
180vac or so, somehow the regulators are developing tremendous voltages on
the circuit board to jump those gaps.


I don't think you need a sensor. That blue flash is good enough. Perhaps the exact
point is no known. Regardless, monitor the voltage feeding the supplies, and
the UPS input, and the voltages between ground and neutral.
greg


"Eric H" wrote in message
...

"T" wrote in message
...
A raised floor server room with 6 servers, fiber optic patch panels, large
control system cabinets and large UPS system, has been recently been
plagued by computer power supply failures. The supplies fail when
apparently many tens of thousands of volts jump from inside the PS around
the switching regulators to the chassis of the supply taking out all the
components. This has happened to 8 supplies so far. A visible blue flash
has been witnessed several times by pewrsons in the room.

The servers so far affected are all sitting on the raised computer center
floor. What ususally happens seems to be when a person enters the 12x20ft
room or gets near a server a large snap is heard and the server is on the
backup redundant PS, if it has not already failed.

We have been adding grounding bonding from all computer case to the bldg
steel, raised floor structure, etc to try to stop this. Its still
happening. The computers are various brands and varoious ages, from 1
week old to 5 years old. The probelm began occuring 2 months ago. The
data center was built 12 years ago.

We are at a loss to figure out how the charge is building up on the
inside of the power supplies. I am thinking about the common power
source, via the power cords Hot, Neutral and Ground conductor.

The green wire should be bonded to the PC case. Apparently the PS
regulator board floats above chassis potential. Apparently a large
potential differnce is building. It jumps a 1 inch gap to the chassis.
Burn marks from repeated arc overs are evindent (the arc that makes it
fail is not the 1st time it occurs!)

The Hot and Nuetral..how could a charge be coming in on these conductors
and getting past the MOVs to build a potential on the boards.


Any ideas would be appreciated.
TJS





I have had this problem before. The servers on site were blowing power
supplies intermittently and corrupting the data on the HDD. After lots of
parts swaps, softeware rebuilds etc the situation was getting extremely
serious. I was called in to investigate.
The site had a large UPS installed. Measuring the voltage with my digital
meter I found the supply varying from approx 198v to 224v over a period of
1 hour. The UPS was isolated but the readings were very similar.
I temporarily installed a mains analyser and when I returned a few days
later, after looking at the printout was amazed to find that the power was
fluctuating between 210v ~ 380v. It should be 220v.

I had a qualified electrician check out the site but he could find no
fault. I then called out the electricity board who discovered an
intermittent open circuit earth in the buildings 3 phase 415v power
supply.

This was traced to a connection outside the building underground. After
repair we had no more problems.