Why it is difficult
An intermittent failure appears and disappears: a contact that opens for a few microseconds, a fractured solder joint that conducts or not depending on the deformation, a connection that depends on temperature. An inspection at the end of the test or a periodic check almost always arrives when the product is working again. The only reliable way to see it is to observe without interruption while the environment acts.
What is monitored
A signal representing the critical function is watched: continuity of a circuit, an output voltage, a communication link, a logic state. What matters is that the monitoring is faster than the duration of the failure. A discontinuity event detector is designed for exactly this: it triggers when the resistance of a loop exceeds a threshold for a minimum time and counts each occurrence.
What is recorded for each event
- Exact instant, on the same time reference as the chamber profile.
- Event duration and cumulative count.
- Temperature, humidity or vibration level at that instant.
- Affected channel or path, to locate the weak point.
Synchronization with the environmental profile
The value of the record lies in the correlation. If events cluster in the temperature ramps, the issue is differential expansion; if they occur at the vibration peak, resonance or a loose connector; if they appear when moisture condenses, surface leakage. Without a common time scale between the monitoring system and the chamber control, that reading is impossible.
Wiring and mounting
The monitoring wiring itself can produce intermittent faults. Instrumentation cables and connectors must withstand the environment, be secured so that movement does not inject noise, and be tested before the test with a stable reference. If the record fails on a control channel with no product connected, the cause is the setup and not the sample.
How to interpret the result
A single short event is already information: a product that fails intermittently under test may fail more seriously in service. The acceptance criterion must therefore be defined before starting, stating whether any loss of function is allowed and what is done with events that recover on their own. After the test, the sample is kept for root cause analysis, since the absence of a failure in a later inspection does not prove that it did not exist.