The crankshaft position sensor: how the ECU knows where the engine is

Almost every sensor on an engine can fail without stopping it. A dead coolant sensor gets replaced with a default value and the engine runs rich. A dead airflow meter gets substituted from throttle angle and manifold pressure. The module has a fallback for nearly everything, because most of what it measures it can estimate.

There is no plausible default for "where is the crankshaft right now". That is why the crankshaft position sensor is different in kind rather than in degree: lose it and the engine does not run badly, it stops — instantly, completely, and with no warning. P0339 is the code for losing it intermittently, and the corpus is direct about why that is the dangerous one: at motorway speed the engine cuts, and the power steering and brake assist go with it.

Everything else follows from that. This article is a translation of a Persian original with 377,000 views, and the thing it explains best — the toothed wheel and the missing teeth — it explains wrongly, in a way that reads like expertise.

What it is looking at

A toothed steel wheel turning with the crankshaft, and a sensor a few millimetres from its edge. On many engines the wheel is a second row of teeth machined into the flywheel — one row for the starter to engage, a finer one for the sensor — and on others it is a separate reluctor behind the crank pulley at the front. The original describes both positions correctly.

The near-universal pattern is called 60-2: sixty tooth positions six degrees apart, with two of them left out, so the wheel carries fifty-eight teeth and one gap. Every tooth edge that passes the sensor is a pulse. Count the pulses and you have crank angle to six degrees; time them and you have engine speed; watch them vary within a single revolution and you have misfire detection, because a cylinder that does not fire lets the crank slow measurably before the next one catches it.

So "engine speed sensor", the name the original leads with, undersells it. Speed is a by-product. Position is the product, and the whole of ignition and injection is scheduled against it.

The missing teeth are a reference mark, not a timing setting

This is where the original goes wrong, and it is worth taking apart because the mistake is a common one.

It says: when the two missing teeth arrive at the sensor, pistons 1 and 4 are at top dead centre, and the module takes the absence of a pulse as its cue to fire. It then derives a number from the gap — two teeth out of sixty is twelve degrees, therefore the module sees the signal twelve degrees before top dead centre, therefore that is the engine's static advance; had one tooth been removed instead, the static advance would have been six degrees.

Both halves of that are wrong, and the second half describes a relationship that does not exist.

The gap is an index mark. It tells the module one thing — this is the same place on the crankshaft as last revolution — and everything else is counted from it. Its position is chosen deliberately, and it is chosen to be nowhere near top dead centre. On a four-cylinder the gap typically passes the sensor somewhere between sixty and a hundred and twenty degrees before cylinder one reaches TDC, for two reasons that both push the same way. The module needs the reference to arrive before the event it is scheduling, with enough crank rotation left to compute a dwell time and start charging a coil. And the crankshaft decelerates hard on the compression stroke near TDC, which is exactly the condition in which a detector looking for "a longer than usual gap between pulses" is worst at telling a real gap from a slow tooth. Putting the index at TDC would be the single worst place to put it.

As for static advance: twelve degrees is the width of the gap, not an ignition setting. Ignition advance is not a property of the wheel at all. It is a figure the module computes afresh for every firing, from load, engine speed, coolant temperature, intake temperature, knock feedback and what it has learned about the fuel, and then schedules by counting teeth from the index. Re-cut the wheel as 36-1 and the advance would be identical, because nothing about the timing lives in the tooth pattern. Cars with a distributor had a static advance you set with a timing light, and that is almost certainly where the idea came from — but the whole point of the arrangement described here is that it replaced it.

Position is not phase, and that is why there are two sensors

The original mentions the camshaft sensor twice in passing and never says what it is for. It is worth stating plainly, because it explains a whole family of codes.

The crankshaft turns twice for every complete engine cycle. The gap therefore passes the sensor twice per cycle, and the two passes are indistinguishable — one has cylinder one coming up on compression, the other has it coming up on exhaust. From the crank sensor alone the module knows the angle and not the stroke.

The camshaft turns once per cycle, so the camshaft sensor resolves the ambiguity. Together they give phase: which cylinder is actually about to fire. That is what sequential injection needs, what variable valve timing needs, and what P0016 is checking when it reports the two signals no longer agreeing with each other — usually a stretched timing chain or a cam phaser, rarely a sensor. P0340 covers the camshaft side.

The original notes that on a car with a camshaft sensor, a failed crank sensor may still start but will not rev past about four thousand. That is a real limp-home strategy on some systems, and it is a strategy rather than a rule — plenty of engines simply will not start. Take it as "this is possible", not "this is what will happen".

Three types, and the original's list is a decade out of date

It lists optical, Hall-effect and inductive, and says the inductive type is the common one. Two of those three need correcting.

Inductive (variable reluctance) is what the original describes: a permanent magnet with a coil wound round it, generating its own AC voltage as teeth disturb the field. Two wires, no power supply, and beautifully simple. It has one property that matters enormously and the original does not mention: its output amplitude rises with speed, and at zero speed it produces nothing at all. An inductive sensor cannot tell the module where the engine is until the engine is already turning, and the signal while cranking is the weakest it will ever be — which is why marginal ones fail to start and then run fine.

Hall effect needs a supply and a ground and returns a square wave. Three wires, a clean digital edge, and it works down to very low speeds.

Magnetoresistive is the type the original does not have, and it is now the common one. Its output is digital, its amplitude is independent of speed, it tolerates a larger air gap, and it reads correctly at zero speed. That last property is not a refinement — it is what a stop-start system requires, because to restart an engine by firing a cylinder rather than by cranking it, the module has to know which cylinder the engine stopped on.

Optical belongs in a footnote rather than at the head of the list. Optical pickups were real, notably inside some Japanese distributors, and they are not what is bolted to a modern crankshaft.

Testing it, and why the test everybody does proves almost nothing

The original gives a resistance test, with per-model figures: 300–400 ohms on the 405, Pars and Samand, 320–400 on the Pride, 200–400 on the 206. Those figures are correct for those cars and the test is nearly useless, because of what it cannot see.

A resistance check confirms that the winding is neither open nor dead short, at room temperature, at DC. It does not detect the way these sensors actually fail. The classic failure is thermal: insulation inside the sensor breaks down as it warms, output collapses, the engine cuts out, and half an hour later the sensor has cooled and measures perfectly and starts perfectly. Anyone who has chased this fault has usually replaced a fuel pump first, because from the driver's seat the two are identical.

That is also the point on which the original is not merely incomplete but wrong. It says the inductive type, being just a magnet, an iron core and a coil, "usually does not fail". It fails constantly, and it fails in the most expensive way a component can — intermittently, hot, and while somebody is driving. The corpus says so at P0335 and P0339, which were written before this translation and disagree with it flatly.

So the useful tests are these.

• Look for output, not resistance. A meter on AC volts across an inductive sensor while somebody cranks will show a clear reading if it is generating and nothing if it is not. The presence of a signal is what you are proving; the exact voltage is not worth arguing about on a hand-held meter.
• Test it at the temperature it fails at. Warm the sensor with a heat gun while the engine idles. A sensor that cuts the engine when hot and recovers when cool has answered the question, and no bench measurement will.
• Capture it during a real cut-out. Recording live data through the fault is what actually closes this diagnosis — see reading live data, and the engine cut out and will not restart for working through a no-start in order rather than by guessing.
• Check the air gap and the wheel, not only the sensor. A cracked or bent reluctor tooth, a wheel that has moved on the crank, or a sensor that has been fitted with a spacer missing produces a signal the module rejects as implausible. That is P0336, and replacing the sensor will not fix it.

The one piece of maintenance that is free

The original's most practical observation is a good one and worth keeping: an inductive sensor sitting over a flywheel collects iron swarf and oily grime on its pole piece, and that debris widens the effective air gap and weakens the signal. Pulling the sensor and wiping the tip costs nothing and occasionally is the entire repair.

Add one thought to it, though. Iron filings on the tip came from somewhere — a ring gear, a clutch, a thrust face — so clean it, and then ask what is wearing.

Two things to do after fitting a new one

Buy carefully. Cheap crankshaft sensors have a reputation for exactly the heat-related failure described above, and a sensor that fails the same way in eight months has cost more than the difference.

Check whether the car needs a relearn. Misfire detection works by measuring tiny variations in crank speed, which means the module must first learn the machining tolerance of the reluctor wheel itself — otherwise ordinary manufacturing variation reads as a misfire. On the cars that require it, fitting a new sensor without running the procedure leaves you with P0315 and phantom misfire codes on a healthy engine. It is a scan tool operation, not a part.

A note on push-starting

The original opens its list of the sensor's functions with something that sounds trivial and is not: the first thing the sensor tells the module is simply that the flywheel is moving — whether that motion came from the starter or from four people pushing.

It draws one conclusion from that, which is that an injected car can be push-started, and adds a caution worth repeating: on an engine with a rubber toothed timing belt, dumping the clutch puts a large shock through the belt, and a worn belt or a tired tensioner can jump a tooth. The engine then runs badly with the valve timing out and no fault code to explain it, which is a genuinely nasty place to end up.

There is a second conclusion it does not draw, and it is the one that belongs in this article. If the car will not start because the crankshaft sensor has failed, pushing it will not help. The module is not waiting for the engine to spin faster; it is waiting to be told where the engine is, and nothing about a push supplies that.

Video guides

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