Intermittent And Environmental Faults

Every method in this volume so far has assumed one mercy: that the fault is present while you look for it. This chapter is about the faults that refuse — the device that crashes twice a week, the connection that drops only on cold mornings, the board that works flawlessly on the bench and fails in the field. Intermittents are the hardest problem in repair not because their mechanisms are exotic — a cracked joint, a marginal component, a drifting parameter — but because diagnosis needs evidence and an absent fault produces none. The chapter opens with why that is: the classes of intermittent fault, why the bench itself changes the conditions that provoke them, and the two ideas that turn the problem tractable — the failure window, the set of conditions under which the fault appears, and the reproduction recipe, the documented sequence that summons it on demand. Then it arms the campaign. Thermal provocation drives heat- and cold-dependent faults into the open with controlled warming and freeze spray. Mechanical provocation — flex, tap, and vibration — forces cracked joints, fractured pads, and marginal connectors to confess. Power and load provocation squeezes supply margins and switches loads to expose faults that live at the edges of tolerance. The long watch turns instruments into unattended sentries — logging meters, scope tripwires, and persistence accumulating evidence over hours when no provocation works. And the chapter closes where every intermittent repair must: turning a reproduction into a verified fix, because a repair of an intermittent is proven only when the recipe that once summoned the fault reliably fails to. By the end, the fault that only happens sometimes stops being a matter of luck and becomes a matter of method.

• 9.1Why Intermittents Are the Hardest FaultsEvery method in this volume so far has leaned on one quiet assumption: the fault is present while you look for it. A dead rail stays dead for the meter, a distorted signal holds its shape for the scope, and a short keeps heating for the thermal camera. The intermittent fault breaks that assumption — the device that crashes twice a week, the audio that cuts out only in a cold room, the board that fails in the field and passes every test on the bench. This opening section is about why these faults are the hardest problem in repair, and about the two ideas that make them tractable. The difficulty is not exotic mechanisms: an intermittent is usually an ordinary fault — a cracked solder joint, a marginal component, a connection at the edge of contact — that has not finished failing, so it fails only when some condition pushes it over the line. The difficulty is evidence: diagnosis runs on observation, and an absent fault produces nothing to observe, which is how devices earn the miserable verdict of no fault found and go home to fail again. Worse, the bench itself changes the conditions — different power, different temperature, different position, a probe's touch — so the very act of examining the patient can hide the disease. The way out is to stop hunting the fault and start hunting its conditions: mining the fault history for the failure window — when, where, and under what circumstances the fault appears — and then converting that window into a reproduction recipe, a documented sequence that summons the fault on demand. Once a fault can be summoned, it is no longer intermittent in any way that matters: it is a steady fault that happens to need a ritual first, every tool in this volume applies again, and — just as important — the repair can be proven, because a fix is verified only when the recipe that once summoned the fault reliably fails to. By the end of this section you can classify an intermittent by what provokes it, explain why it hides at the bench, write its failure window from the history, and draft the recipe the rest of this chapter will teach you to execute.IntermediateLow Risk20 min read
• 9.2Thermal Provocation — Forcing Heat- and Cold-Dependent FaultsThe thermal intermittent announces itself in its history: the amplifier that dies after an hour, the monitor that will not start on a cold morning, the console that recovers after a rest and fails again once it is warm. The last section taught you to read those histories into a failure window; this one teaches the campaign's first provocation — recreating the window's temperature on demand instead of waiting for the weather. The tools are not new: thermal diagnostics taught freeze spray and localized heating as active probes for isolating a fault that was already showing. What is new is the direction of use. A hidden intermittent is summoned globally first — a thermal soak that runs the device warm in its own case the way its owner does, or a controlled chill that recreates the cold morning — with an instrumented observable armed the whole time, because a fault that appears unwitnessed teaches nothing. Once the fault can be summoned at will, the provocation narrows: warm or chill zones of the board, then single components, letting the shrinking recipe do the localising until the freeze spray is landing on one suspect joint or one drifting part. Around the technique sits the discipline that keeps deliberate temperature work safe and honest: heating that stays inside component ratings and never dwells, chilling that respects condensation — the moisture film that a cold board pulls out of the air, which can invent leakage faults, corrode, and confuse everything — and the patience to let a board dry before conclusions or power. And the section closes where every intermittent repair must: the thermal recipe becomes the proof, because a repair of a warm-up fault is verified hot, at the temperature that used to summon it, not on a cool bench where the fault never lived. By the end you can take a thermal failure window and summon its fault in minutes, narrow it to a zone and then a component, and prove the repair across the temperatures that once broke it.IntermediateMedium Risk22 min read
• 9.3Mechanical Provocation — Flex, Tap, and VibrationThe mechanical intermittent is the one that follows motion: the laptop that blanks when its lid moves, the console that crashes when the desk is bumped, the car radio that cuts out on rough roads and behaves in the driveway. Behind it is almost always a connection at the edge of contact — a solder joint cracked by years of thermal cycling or one hard drop, a pad lifting from its laminate, a connector whose spring contacts have relaxed, a wire strand broken inside intact insulation — and it fails exactly when flex, tap, or vibration moves the crack. This section arms the campaign's second provocation: summoning that fault with deliberate, controlled motion instead of waiting for the next bump. The wiggle test from physical inspection grows into a method — the whole device handled and repositioned first, the way transport and use would move it, then zones of the board pressed through an insulated tool, then single joints and pins prodded one at a time, every round with an instrumented observable armed, because a dropout that nobody captured teaches nothing. Direction matters and the section teaches why: a crack opens under one bend and closes under the opposite one, so a joint that responds to pressure from the north and not the south is describing its own geometry. Around the technique sits the discipline that keeps force diagnostic instead of destructive: insulated tools that cannot short what they touch, graded pressure that starts featherweight, flex measured in millimetres — because ceramic capacitors crack under enthusiastic bending and the test must never create the faults it hunts — and respect for live boards under moving hands. And as with every provocation, the recipe closes the loop: the response is confirmed by repetition and reversal, the suspect is verified under magnification, and the repair is proven when the same flex, tap, and handling that once summoned the fault reliably summons nothing. By the end, the fault that follows motion performs on your schedule, names its own joint, and stays fixed.IntermediateMedium Risk22 min read
• 9.4Power and Load Provocation — Margins, Brownout, and Load StepsThe power-and-load intermittent lives at the edges of tolerance: the tablet that crashes only on battery, the router that reboots when the compressor kicks in, the console that fails during the loading screen and never during the menu. Behind it is a margin that has quietly eroded — a tired electrolytic whose reserve is gone, a regulator running at the edge of its dropout, a connector adding resistance it should not have — so the device works while conditions are generous and fails the moment its source weakens or its demand peaks. This section arms the campaign's third provocation: squeezing those margins on purpose. The bench has two levers. Voltage margining sweeps the input downward through the device's rated range and beyond toward brownout, finding the exact voltage at which the fault appears — a number, not an anecdote, and one that a known-good unit turns into a verdict. The load step drives demand the other way: dummy loads switched onto rails and the device's own heaviest function run on command, so the sag-and-recover transient that once needed the owner's timing happens under an armed tripwire. Around both sits the trap the bench must avoid: its own generosity. The stiff bench supply, the short fat leads, and the fresh battery are exactly what hide this class of fault, so source substitution — testing with the owner's actual adapter, the tired battery, the long thin cable — is often the whole summons. The discipline keeps the provocation inside absolute ratings and keeps its artifacts out of the evidence, because a current-limited bench supply sags on its own schedule and long test leads drop voltage the device never sees at home. And as always the recipe closes the loop: the fault's margin becomes its measure, the repair must move that number back where it belongs, and the proof is the same sweep and the same steps run against the fix. By the end, the fault that lives at the edge of tolerance gets dragged to that edge on your schedule, measured there, and repaired until the edge moves back where the designer put it.IntermediateMedium Risk22 min read
• 9.5The Long Watch — Monitoring and Logging Over TimeSome failure windows no lever can compress. The fault tied to weather, to the third day of operation, to a neighbour's welder, or to nothing anyone has spotted yet will not answer heat, motion, or a squeezed supply — its window is time and circumstance, and the only way through is to be watching when it finally happens. This section turns that waiting from a vigil into an instrument problem. The bench already owns the sentries: a multimeter's min/max recording holds the extremes a rail visited while nobody looked; a data logger turns hours into a voltage-versus-time trend; the scope's tripwire and persistence — the standing watches of the capture chapter — hold the precise, fast view; and an event counter tallies how often the abnormal recurs. The craft is in the rigging: sentries matched to the fault's expected shape, layered so a coarse trend and a fine trigger cover each other's blind spots, the device running its real workload rather than idling politely, and an environment logger recording temperature and humidity alongside — because the correlation is often the diagnosis. Which is the section's second skill: timestamp correlation. Clocks synchronized across instruments and a diary of real-world events — the compressor, the rain, the cleaning crew — turn a captured fault's timestamp into a refined window: not just what happened, but what else was happening. And the watch has honest bookkeeping: a silent week means one thing against a daily fault and nothing against a monthly one, artifacts like probe bumps and bench mains events must be recognised before they become findings, and when the fault will not visit the bench at all, the watch goes home with the device. By the end, the faults that outwait patience stop outwaiting the instruments — and every watch, silent or successful, leaves the window smaller than it found it.IntermediateMedium Risk22 min read
• 9.6From Reproduction to Verified RepairThe campaign has done its work: the history became a window, the window became a recipe, the provocations and the long watch summoned the fault and named its mechanism. What remains is the part that decides whether the device ever comes back — turning that evidence into the right repair, proving the repair with arithmetic instead of optimism, and writing the record that makes the next case of the family start ahead. This closing section assembles the chapter's endgame. The repair is chosen by the evidence: a confirmed mechanism dictates its own fix — the named joint is reflowed, the eroded margin is recapped, the relaxed contact is retensioned or replaced — and the temptation the evidence must hold back is shotgunning, the resolder-everything reflex that sometimes works, never teaches, and cannot be verified. The proof is designed, not assumed: the verification count is chosen to beat the fault's own reproduction rate, the minimal recipe that summoned the fault is the recipe that must now stay silent, and the runs sweep the window's whole band — hot and cold, loaded and idle, flexed and still — because a repair can move a window instead of closing it. The campaign's failures get honest endings too: a no-fault-found issued only over a documented, explored window; a preventive repair offered as a labelled wager when evidence is partial and reproduction is out of reach; and the return path — a field watch, the next season — kept open in the record. And the record is the second product: window, recipe, captures, margins, repair, and verification count filed under the device family, growing the recipe library that lets the next twin of this fault skip the weeks this one cost. By the end, the chapter's promise is kept end to end: the fault that only happened sometimes was summoned, named, fixed, proven fixed, and filed — and the bench is better armed for every device that follows.IntermediateMedium Risk22 min read
• Chapter Quiz42questions · 80% required to continue