Diagnosing An Intermittent Fault Is A Problem Of Reproducing It

An intermittent fault is not a harder version of a permanent one. It is a different problem, because every diagnostic technique requires the failure to be happening at the moment of measurement.

Testing only sees the present state

Measuring a circuit, a joint or a mechanism reports its condition at that instant, and a fault that has temporarily gone away leaves nothing to detect.

This is why a device frequently works perfectly on a repair bench, and why the report of the fault from the owner ends up being more useful evidence than any measurement.

The consequence is that the first real task is not finding the fault but finding the condition under which it appears, since without that there is nothing to work on.

Most intermittents follow one of four variables

Temperature is the commonest, because expansion opens and closes cracks, and semiconductors and connections change their behaviour as they warm through their operating range.

Movement is the next, covering vibration, flexing of a cable or board, and the position a product happens to be resting in when it is used.

Load and humidity account for most of the remainder, with faults appearing only when a circuit is drawing significant current or when moisture has had time to condense.

Provocation is a legitimate technique

Deliberately applying the suspected variable turns an intermittent fault into a controllable one, which is the whole basis of tapping a board, flexing a cable or warming a section.

The value lies in the specificity: if pressing one corner reproduces the fault reliably, the search area has shrunk from the entire product to a few square centimetres.

The corresponding risk is that heavy-handed provocation creates new damage, so the technique works best applied gently and progressively rather than as a single decisive test.

Logging replaces presence

Where a fault cannot be provoked, the alternative is to leave a record running so that the conditions at the moment of failure are captured without anyone watching.

Even a crude approach helps, such as noting the time and circumstances of each occurrence, because a pattern in when it happens narrows the cause considerably.

Many products keep internal error logs for exactly this reason, and their contents are frequently more informative than any measurement taken after the event.

Confirming a repair needs a different standard

A permanent fault is proven fixed the moment the product works. An intermittent one is not, because it worked before the repair as well.

The only reliable confirmation is that the provoking condition no longer produces the fault, which requires having found that condition in the first place.

Where the fault was never reproduced, a repair is a hypothesis rather than a fix, and it is worth treating it that way when deciding how much further money to spend.