Since the end of winter, there's been an epidemic of PC workstation failures. The failed components are mostly power supplies and motherboards.

The power monitor doesn't show any anomalies that could account for this. However, from the repair records, you find:

  • It's not affecting any particular brand or model.
  • It's affecting only PCs in the administrative offices, not any PCs on the factory floor.
Fortunately, the company keeps data on servers and not on individual workstations. Therefore, there hasn't been a catastrophic data loss. Nevertheless, the work interruptions and repair costs now have upper management's attention, and they want this fixed. What should you do?

The timing with the end of winter doesn't necessarily indicate a thermal problem. It could be a hint that the air-conditioning system's dehumidifier is a bit too aggressive. Take some humidity readings around the office and/or use an inexpensive humidity gage that records highs and lows. Ask people if they're getting "carpet shocks." If they aren't, then humidity probably isn't the issue.

The timing coincides with an annual ritual at many companies. It's the "budget release" that ushers in buying of deferred purchases in the first quarter. Department managers finally can purchase that new production scanner or wide format printer.

Have your accounting department generate a report showing new equipment purchases made (or, preferably, delivered) since the first of the year. Carefully examine each installation for bonding errors, especially with the communication cabling. If your DMM shows more than zero volts between that new printer's network cable shield and the ground terminal of its power receptacle, then you have a bonding problem.

All metallic objects must be at the same potential, or a dangerous voltage can build up on any one (or more) of those objects. This can result in "flash over" between objects.

If network cables are at varying potentials, this explains the motherboard failures. Today's PCs provide Ethernet from the motherboard rather than from a network card. If the cables vary in potential with each other, then they vary in potential with the PC power supply — thus power supplies also fail.