Phase correlation is not a pass or fail meter

A correlation meter gives you a number between minus one and plus one. The widespread interpretation is that plus one is good, zero is acceptable and anything negative is a fault to be fixed. That reading will have you flattening a perfectly healthy stereo mix.

What the number means

Correlation measures how similar the left and right channels are. Plus one means they are identical — the signal is mono. Zero means they are unrelated. Minus one means they are identical but opposite in polarity, and will cancel completely if summed.

So the meter is not measuring quality. It is measuring similarity, and the right amount of similarity depends entirely on what you are making.

A mono-heavy pop mix sits high, around plus zero point five to plus zero point eight, and that is correct. A wide ambient piece with genuinely different content on each side can sit near zero for long stretches, and that is also correct. A brief dip into negative during a wide reverb tail or a stereo delay throw is normal and audible to nobody.

The meter tells you how mono your mix is. It does not tell you whether that is the right amount.

What is actually diagnostic

Three readings matter, and none of them is the instantaneous value.

The first is sustained negative correlation. A dip is fine; sitting below zero for a whole section means a substantial part of your mix will vanish when summed. That is worth finding, because it is almost never intentional.

The second is the value in the low end specifically. Broadband correlation averages everything together, and a healthy-looking overall number can hide bass content that is cancelling. Below about two hundred hertz you want correlation close to plus one, and any meter that lets you look at bands separately is worth more than one that does not.

The third is movement. A correlation reading that swings wildly and constantly usually means phase-based processing somewhere — a widener, a chorus, a stereo imager — doing more than you think. A stable reading, wherever it sits, is a mix whose image comes from real placement rather than from manipulation.

The bass case

If you take one thing from correlation metering, take this: check the low end separately and get it close to mono.

Directional hearing at low frequencies barely functions. You gain almost nothing perceptually from stereo bass, and you risk a great deal — cancellation on summed systems, wasted headroom, unpredictable behaviour on club rigs and vinyl. The cost-benefit is one-sided enough that mono bass below a crossover point is a default worth adopting rather than a decision to agonise over.

The common sources of stereo low end are worth knowing because none of them announce themselves: a stereo synth patch with built-in chorus, two bass DIs panned apart, a room mic pair where the kick arrives at different times, and any widener applied across the full range rather than to the top.

Reading it alongside the mono check

The meter and the mono button answer different halves of the same question, and neither is sufficient alone.

The mono button tells you what collapse sounds like, which is what you ultimately care about. But it will not reveal low-frequency cancellation on small monitors, because the speakers are not reproducing enough of that region for the loss to register.

The meter catches exactly that case, and catches it numerically rather than by ear. Use the meter to find where to look and the mono button to decide whether it matters.

The meter finds candidates. Your ears decide. Neither job can be done by the other.

A practical rule

Keep broadband correlation positive on average, with dips allowed. Keep the region below two hundred hertz close to plus one. Investigate anything that sits negative for more than a couple of seconds. Ignore everything else.

That is the whole discipline. It takes no extra time once the meter is on the bus, and it catches the specific class of problem that is invisible in your room and obvious in a club — which is the worst possible place to find out.