The crank sensor tells the computer where the engine is. The cam sensor tells it which cylinder is which. Correlation is the computer checking that the two stories agree - learn what it is actually comparing, why those codes set, and why a lying signal looks exactly like bad timing.
The crank sensor reports position and speed. The cam sensor reports phase and cylinder identity. Correlation is the computer's third check: do the two signals agree with each other? Every revolution, the module looks at where the cam event lands relative to the crank reference and asks whether that landing spot falls inside a calibrated window. Inside the window, the relationship is good. Outside it, the computer sets a correlation code.
That distinction matters more than anything else in this explainer. Correlation is not a test of the cam signal. It is not a test of the crank signal. It is a test of the relationship between them. Both signals can be individually perfect and still fail correlation - and both signals can look fine on a scan tool while the relationship is wrong.
Codes P0016 through P0019 all mean the same thing in different words: "the cam/crank relationship I measured does not match the relationship I was calibrated to expect." The number tells you which bank and which cam sensor the computer was watching. It does not tell you why the relationship is wrong.
Read the code for exactly what it says. It is the module's claim about a relationship, made from inside the module, using the signals as the module received them. It is not a measurement you made. It is not proof the timing is off, proof a sensor is bad, or proof anything mechanical moved. It is the starting claim - and claims get verified, not believed.
Three questions get tangled together constantly. Untangle them:
The order is load-bearing: presence first, sync second, correlation third. Judging correlation on a signal that drops out is judging a relationship with a missing partner - the answer means nothing. That is why content gets screened before phase is ever judged.
Here is the trap that eats more diagnostic time than any other in this family. The computer judges correlation from the signals as received. If the crank signal drops a pulse, the module's picture of crank position shifts - and the cam event now lands "outside the window" in the module's math. The code says correlation. The reality is a crank-signal dropout.
The same goes the other way: a cam signal with an intermittent dropout or a shifted edge moves the cam event in the module's math, and the code says correlation while the timing hardware sits exactly where it should. A lying input produces a phase error on paper. That is why the signals get proved at the controller input - the exact point where the module sees them - before anyone is allowed to talk about timing.
The related explainer on phase covers why "late" has five different owners (Phase lag vs the physical event: why 'late' has five owners) - correlation codes are one more place where the same discipline applies.
Variable valve timing moves the camshaft relative to the crankshaft on purpose - advancing or retarding valve events while the engine runs. That means the "correct" cam/crank relationship is not one fixed number on a VVT engine. It is a moving target, and the computer's correlation window moves with the commanded phaser position.
Two consequences. First, the correlation check has to be judged at a known VVT state - the window that applies with the phaser parked is not the window that applies with it advanced. Second, a phaser that sticks - fails to park, fails to advance, drifts - puts the real relationship outside the window the computer expects. A stuck phaser reads exactly like jumped timing on every instrument you own. The only thing that splits them is commanding the phaser to park and watching whether the relationship comes home.
Correlation is worked as a relationship question with a strict order: prove the signals first, then judge the phase. The crank and cam signals are captured at the controller input - where the module sees them - and each is checked against the vehicle's own requirements: count, pattern, levels, dropouts. Only with both contents good does the phase comparison happen, and it happens against the service-information correlation window for the known VVT state - never against a generic tolerance.
When the phase is off, it has to be off twice - two independent captures, same direction, still outside the window - before anything mechanical is even discussed. And when the module claims a fault the scope cannot see, the claim gets a currency test: clear it, reproduce the conditions, and see whether the module still says so. A stale claim proves nothing.
The evidence-discipline explainer covers why a later normal reading never erases a reproduced failure (A later normal reading never erases a reproduced failure) - the intermittent side of correlation lives by the same rule.
Correlation is the computer asking, every revolution, whether the cam and crank signals agree on where the engine is. The code is the computer's claim that they do not. Verify the claim the way the computer made it - signals proved at the input, phase judged against the real window, twice - and never let the code pick the part. The relationship is the question; the signals are the evidence; the window is the judge.
Stay in the loop
New QTS builds, tools, and garage stories — straight to your inbox. No spam. Unsubscribe anytime.
Read the Privacy Policy.