On Aug 11, 3:43 pm, (Al Dykes) wrote:
We wound up getting a ferroresonant constant voltage transformer so
large it had to be moved in with a forklift.

Why did the "other mainframes" not have problems?

A story from Datamation magazine. One computer would crash
constantly. The tech would run diagnostics all night - no failure.
But when the tech went for coffee, then diagnostics would crash every
time. Got so bad that the tech would loudly announce he was going for
coffee, stomp out of the room, then sneak back to peek around a door
and watch the machine. No failures.

Problem was traced to how the computer was grounded. It was
grounded to the elevator because a human did not learn his job.
Everytime the tech took an elevator down for coffee, that elevator
would crash the computer.

There are no ghosts. But there are many humans who cannot bother to
first learn even basic grounding; then want to cure problems with
magic boxes because they never learned even simple concepts of
grounding.

Con Ed capacitors should never cause problems for computers. But
some early computers had marginal designs. For example, specs for PCs
are quite blunt. Power must be completely lost for 17 milliseconds
and computer must maintain output voltages. Computers that put fewer
loads on their power supply can operate even longer without power.
UPSes that switch over to battery backup may interrupt power for up to
10 milliseconds. If computer was upset by switching in reactive
loading, then computer would also crash when a UPS switched to battery
backup.

Described were some simple procedures to even identify bad
grounding. Of course, if the only failing device is the router, well,
is that router design too marginal? So we buy a $500 UPS to fix a $50
defective router?