Intermittent Fault Escapes Bench Test
A board intermittently fails to boot, roughly one attempt in twenty, at a customer site but never on the bench. The board uses a 3.3 V regulator, a crystal oscillator, and a reset supervisor. Bench testing at 22 °C with a lab supply passes 200 boot cycles. At the site the ambient is 4 °C in the morning and the supply is a 24 V industrial rail that carries a 1 V peak-to-peak ripple at 100 Hz. The technician wants to replace the crystal first because crystals are a common cause of boot failures.
1. Evaluate the crystal-first hypothesis and design a test that could reproduce the fault.
Read the worked answer
Crystal failure is possible but the evidence points elsewhere: the fault is site-specific, temperature-dependent, and correlated with a rippled supply, none of which the bench reproduced. Replacing the crystal first is a parts-swap guess that consumes time and may mask the real cause. A better test is to reproduce the site conditions on the bench: drive the board from a 24 V rail with a superimposed 1 V 100 Hz ripple, and run boot cycles in a chamber at 4 °C and at the high end of the site range. Log the 3.3 V rail, the reset line, and the oscillator startup on a scope with sufficient bandwidth, triggering on failed boots. That converts an intermittent into an observable event.
2. Name two plausible root causes consistent with the evidence and state how each would be confirmed.
Read the worked answer
- The reset supervisor trips on supply ripple: the 3.3 V rail or the supervisor's threshold dips during ripple troughs at low temperature, holding or releasing reset at the wrong time. Confirm by scoping the reset line and 3.3 V rail during a failed boot and checking the supervisor threshold against the datasheet over temperature. 2) Oscillator startup margin degrades at low temperature, so the MCU misses its startup window. Confirm by measuring oscillator startup time and amplitude at 4 °C and comparing with the MCU's specified startup requirement. Either cause would explain why 22 °C on a clean supply never fails.