Diagnosing engine noises: what each one is telling you

A noise carries four pieces of information, and people reach for them in almost exactly the wrong order.

1. What it sounds like — tick, knock, rattle, whine, squeal, puff.
2. When it happens — cold, hot, idle, load, overrun, one moment in a thousand.
3. What it changes with — engine speed, road speed, load, clutch, temperature.
4. Where it comes from — which is the one everybody starts with, and the least reliable, because an engine bay is a metal box that conducts sound everywhere at once.

The third question is the one that does the work, and one part of it splits the car in half before you have opened the bonnet:

Does the noise keep time with the engine, or with the wheels? Put the clutch in and coast. A noise that carries on unchanged is in the engine. A noise that falls away with road speed is not.

The original of this article is a careful walk through the bottom end — crankshaft, bearings, con rods, pistons, thrust washers — organised by part. That is the wrong way round for a reader who has a noise and no idea what is making it, so this version is organised by noise. Its content is inside, and its best observation is one almost nobody writes down.

Reading it by ear

Two entries there deserve more than a row.

Detonation, which is not a mechanical noise at all

A sharp metallic rattle under load — pulling up a hill in too high a gear is the classic — is not a loose part. It is the fuel-air mixture igniting spontaneously somewhere in the chamber, ahead of the flame front from the plug, and the two pressure waves colliding. The rattle is those waves ringing the block.

The original never mentions it, and it is probably the noise most drivers actually hear.

It matters because it is both the easiest to fix and, ignored, one of the most destructive. Common causes, cheapest first: low-octane or poor fuel, carbon deposits raising the effective compression, an engine running hot (see why an engine overheats), ignition timing over-advanced, too lean a mixture, and a spark plug of the wrong heat range — which spark plugs covers, because a plug that runs too hot becomes an ignition source in its own right.

Modern engines have a knock sensor listening for exactly this and retarding the timing when it hears it, which is why a car can develop the underlying condition and stay quiet — until the sensor or its circuit fails. That is P0325 and its neighbours, and the practical consequence is that a knock sensor fault is not really a sensor problem — it is the loss of the thing that was keeping the underlying condition quiet. Be precise about what is lost, though: most modules do not simply carry on regardless, they fall back to a conservative fixed timing, which is why the corpus rates that code moderate and says to expect slightly reduced performance and economy rather than imminent damage. What has gone is the adaptive protection — the engine is safe enough on good fuel, and has nothing watching for the day it gets bad fuel.

A persistent rattle under load should be treated as urgent. It puts pressure spikes into pistons and bearings that they are not designed for, and the failure it eventually causes is a hole in a piston crown.

Bottom-end knock, and the instruction the original leaves out

The original describes the crankshaft and the bearings in two separate sections with identical symptoms and an identical cause — heavy knocking on acceleration, then a seized engine, caused by oil not reaching them. That is not two faults described twice. It is one fault, and the duplication is what stitched sources look like.

But the underlying point is right and is the most useful thing in the piece: a bottom-end noise is an oil-supply story, not a metallurgy story. Bearings are soft on purpose; they survive on a film of pressurised oil a few microns thick, and they fail when that film fails — low level, wrong grade, a diluted or emulsified oil, a worn pump, a blocked pickup, or oil that has simply been in there too long. Low engine oil level is the mundane commonest one, and P0521 and P0522 are the codes for the pressure side of it.

And here is what the original does not say anywhere. If an engine develops a deep knock that rises with load, switch it off. Not at the next junction — as soon as it is safe to stop. A knocking big end is metal running on metal, and the distance between a bearing shell replacement and a con rod through the side of the block is measured in minutes of driving. If the oil pressure light is on as well, it is already too late to be debating it.

The thrust washer, which is worth knowing and almost never written down

The original's best contribution: a very slight knock heard only when the clutch is pressed or released.

The crankshaft is located fore and aft by thrust washers — semicircular shims sitting against one main bearing cap. Pressing the clutch pushes the crankshaft forward through the release bearing, and if those washers have worn, the crank moves further than it should and taps. Nothing else in the car produces a knock synchronised to the clutch pedal rather than to engine speed.

It matters because it is an early warning of something that ends badly. Left long enough, endfloat wears the crank journal itself, and then the repair stops being a set of shims.

The tick that is not a tappet

An exhaust manifold leak ticks in a way that is genuinely hard to tell from valve clearance, and it is misdiagnosed constantly in both directions. The distinguishing feature is temperature, and it runs the opposite way to intuition:

• A manifold leak is loudest cold and fades as the engine warms, because the metal expands and closes the crack or reseats the flange.
• Valve clearance is loudest cold too — but it does not disappear, it just gets quieter.
• A manifold leak is more of a rhythmic puff or hiss than a sharp tap, and it comes from the side or underneath rather than from the cam cover.

The cheap confirmation is to feel and listen around the manifold flange with the engine cold and running — with a length of hose, not a hand — and to look for the sooty grey trace that escaping exhaust gas leaves at the joint.

Finding where it comes from

Only after the questions above, because localising a noise is the least reliable step and it will happily confirm whatever you already believe.

A mechanic's stethoscope is the proper tool and it is cheap. Touch the probe to the cam cover, the block at different heights, each accessory bracket, the pump housings. A long screwdriver with your ear to the handle does the same job, and so does a length of hose held to one ear.

Take the belt off — briefly, on a cold engine, and not on a car whose water pump the belt drives. If the noise goes with it, it is an accessory bearing rather than the engine, and you have eliminated most of the expensive answers in thirty seconds.

Cut one cylinder at a time with the scan tool's cylinder balance function. A knock that quietens noticeably when one cylinder stops firing has just named its cylinder. Do this with the tool rather than by pulling leads or plug caps — ignition leads explains why that method destroys coils and modules.

Once you have a candidate

A noise on its own rarely proves a diagnosis. It tells you which test to run.

• Top-end tick → check the clearances, and remember that too tight makes no noise at all. Valve clearance.
• Rattle under load with a lack of power → look at fuelling, timing, and temperature before anything mechanical.
• Any suspicion of rings, valves or a head gasket → compression testing, then leakdown, which says where the leak goes.
• Noise with blue or white smoke → head gasket failure and oil consumption.
• Whine or growl that follows engine speed with the belt on and vanishes with it off → an accessory, and on the charging side the alternator has the bearing and rectifier symptoms.

One note on vibration

The third piece folded in here is a Persian-dubbed Toyota course on noise, vibration and harshness, and its framing is worth borrowing even though its text is only a caption: noise and vibration are the same measurement at different frequencies, and the diagnostic question is the same one this article opens with — does it track engine speed, or road speed?

A vibration that appears at a particular road speed and is unaffected by gear or engine revs is a wheel, a tyre or a driveshaft, and it is not an engine problem however much it feels like one from the driver's seat. A vibration at a particular engine speed that is there in neutral is the engine or its mounts.

Video guides

Video, for the parts of this that are easier watched than read.