[tw]

Background: Using temperature to find failures tended to associate cold with mechanical failures and heat with semiconductor problems. One exception to this was marginal timing. Worked just fine on those Perkin Elmer and 68000 microprocessor board at 70 degrees F but would fail both hot and (less often) cold. And then there was this weird computer called Gould (same as the pump company) - that would fail on any temperature above 68F. I had to intentionally cause crashes before I would believe it. NASA's solution was to keep that computer room below 50 degrees at all times. Such timing problems detectable by using temperature suggests never buy their product. Temperature should only detect unique manufacturing defects - its design confirmed by temperature extremes long ago.