Was this article helpful?
Intermittent faults: catching a problem that will not stay broken
The fault happens on the way to work and never at the workshop. Logging turns that into a solvable problem, because the recording is the diagnosis.
Every workshop knows the conversation. The customer describes a fault vividly, the car is tested for an hour, and everything is perfect. The car goes home and misbehaves on the first roundabout.
Intermittent faults are not mysterious. They are faults with entry conditions, and the job is to find out what those conditions are.
Start with what the car already recorded
Before buying anything or driving anywhere, harvest what is stored.
Freeze frame. When a code sets, the ECU snapshots the conditions: engine speed, load, coolant temperature, vehicle speed, fuel trims. That snapshot describes the moment the fault occurred, which is exactly the information you are missing. See freeze frame data explained.
Coolant temperature in the freeze frame alone splits the problem: 20 °C means a cold fault, 90 °C means a warm one. That single number halves the search.
Pending codes. A fault on its first strike sets a pending code with no light. On an intermittent problem, pending codes are often the only evidence — pending vs stored codes.
Mode 06 test results. Generic Mode 06 exposes each monitor’s measured value alongside its pass limit. A value sitting at 95% of the limit is a fault that has not quite happened yet. This is the closest thing to a crystal ball that OBD-II offers.
Find the pattern
Intermittent means conditional. Keep a note of every occurrence with these five columns:
| Column | Why it matters |
|---|---|
| Time and date | Reveals cold-morning patterns |
| Ambient temperature | Splits thermal faults from mechanical ones |
| Weather | Damp is a huge factor in connector faults |
| Journey stage | Cold start, warmed up, or after an hour |
| What you were doing | Turning, braking, accelerating, over a bump |
Four or five entries usually make the pattern obvious.
Cold only — points at plugs, coils breaking down cold, thick oil, or a sensor reading out of range on a cold start. See P0316 misfire on startup and rough idle when cold.
Warm only — points at thermal expansion opening a marginal connection, a component that fails once hot, or a coil that breaks down under heat.
Damp only — connector corrosion, cracked coil insulation, water reaching a module. Related: damp carpet and electrical faults.
Over bumps only — a chafed wire or a loose connector. This is the easiest category to confirm, because you can reproduce it deliberately.
After refuelling only — fuel quality, or an EVAP purge valve stuck open, covered in P0496.
Log the whole journey
This is where a phone adapter beats a workshop visit outright: you can leave it recording for a week.
Set up a log with the values relevant to the symptom, then drive normally. When the fault occurs, note the time. Afterwards, look at what every value was doing in the thirty seconds before that timestamp.
Useful sets by symptom:
- Stalling or hesitation — RPM, MAF, fuel trims, throttle position, rail pressure.
- Electrical gremlins — battery voltage, and whether it dips at the moment of the fault.
- Overheating — coolant temperature, fan status, vehicle speed.
- Transmission — gear, turbine speed, RPM, fluid temperature.
Which PIDs are worth watching generally: best OBD PIDs to log on a daily driver and reading live data as graphs.
The wiggle test still wins
For anything that smells electrical, the highest-yield test costs nothing.
Engine idling, live data on screen, gently move each connector and harness section in turn. A value that jumps, drops out or goes to zero has just told you precisely where the fault is.
Do it on a warm engine and again on a cold one — some faults only exist at one temperature.
Voltage is the usual culprit
A weak battery or failing alternator produces symptoms across the whole car that look like several unrelated faults. Modules brown out, drop off the network and store codes that mean nothing.
If several unrelated systems misbehave at once, test the battery under load before diagnosing any of them. This is the reasoning behind P0606 and the U-code family such as U0100.
When the code is real but the cause is not
A stored code names a circuit, not a culprit. Ghost codes — plausible codes with no matching fault — are usually the product of low voltage, a bad ground, or a module resetting. Full treatment: when the scanner lies about ghost codes.
Do not clear until you have the data
Every clear wipes the freeze frame and resets the monitors. On an intermittent fault, where evidence arrives once a fortnight, that is genuinely costly. Record first, always. Related: why the check engine light comes back.
What to hand a workshop
If you do end up paying someone, arrive with:
- The exact code and its freeze frame.
- Your pattern notes.
- A log covering an actual occurrence.
That turns “it does it sometimes” into a specific, testable claim, and it is the difference between a diagnostic fee and a diagnosis. See how to read a workshop quote.
Try it on the free tier of the app — pair an ELM327 adapter and leave it logging on your commute. The fault only exists for a few seconds a week; the recording is the only way anyone gets to look at those seconds.
Read next
Comments …
Loading…