Mixing by meters vs mixing by ear: how to use both

A meter has never told anyone whether a mix is good. It tells you what is happening, which is a different and still valuable thing.

The useful split is simple: meters measure facts, ears make decisions. Most mixing problems that come from metering are cases of a meter being asked to make a decision it cannot make.

What meters are genuinely good at

Meters beat ears at anything involving an absolute value, a long time window, or something your monitoring cannot reproduce.

True peak level. You cannot hear an inter-sample overshoot. A true-peak meter can measure it, and it is the difference between a file that is safe and one that distorts after encoding. See true peak vs sample peak.

Integrated loudness. Your ear adapts within seconds; it cannot integrate loudness over three minutes. A LUFS meter can.

Content below your monitoring. If your speakers stop at 50 Hz, there could be enormous energy at 30 Hz and you would never know. A spectrum analyser will show it.

Mono compatibility. A correlation meter reveals phase problems that are invisible in stereo and catastrophic in mono.

Gain reduction. Watching how much a compressor moves and how quickly it recovers teaches you more about its settings than any preset name.

What meters are bad at

Everything that requires judgement.

Balance. No meter knows how loud a vocal should be relative to a guitar in your song.

Tone. A spectrum with a dip at 400 Hz might be exactly right, or wrong. The display cannot tell you which.

Depth and space. Not measurable in any way that helps.

Groove and impact. Two mixes with identical measurements can feel completely different.

Whether something sounds good. Which is the entire job.

“Mixing with your eyes” — what actually goes wrong

The failure mode is specific: you look at a display, decide it looks wrong, and correct it.

The classic examples:

• Flattening a spectrum analyser curve because a bump “looks like too much”, when the bump was the character of the instrument.
• Cutting a frequency because a reference track’s curve is lower there, ignoring that it is a different arrangement in a different key.
• Chasing a LUFS number until the mix is flat, when streaming will normalise it back down anyway.
• Adjusting a compressor to make the gain reduction meter look tidy rather than to make the source sit right.

Every one of those optimises for the picture. The listener never sees the picture.

How to use a meter without being led by it

Listen first, look second. Form an impression, then check whether the meter agrees. If you look first, you have already been told what to hear.

Use meters to answer questions, not to browse. “Is there energy below 40 Hz I cannot hear?” is a question. Staring at a moving spectrum is not.

Turn them off for stretches. Mix a section with the analysers hidden. It is uncomfortable at first and it recalibrates quickly.

Believe the ear when they disagree — unless the meter is measuring something you physically cannot hear, in which case believe the meter. True peak, sub content, phase correlation, integrated loudness: those are the meter’s jurisdiction.

The meters worth having

Meter	Question it answers
True peak	Will this distort after encoding?
Integrated loudness	Am I in the right range for delivery?
Spectrum analyser	Is there content I cannot hear? Where is that resonance?
Correlation / phase	Does this survive mono?
Gain reduction	What is this processor actually doing?

That is enough. More displays mean more things to be distracted by.

Where the ear needs help

Your ears are not neutral instruments, and knowing their failure modes matters more than knowing your meters.

Level bias. Louder always sounds better. Match levels before every comparison — the single most important habit in mixing.

Fatigue. After two hours, judgement of brightness and level drifts predictably. Take breaks. Decisions made at hour four are usually undone at hour one the next day.

Adaptation. Listen to anything long enough and it starts sounding normal. This is why references work: they reset the baseline.

Room and monitoring. Every room lies somewhere. Checking on a second system and in mono catches most of it.

Two supporting habits: mix with everything playing, because balance only exists in context, and reference against a finished record at matched level.

A practical division of labour

Ears decide: balance, tone, depth, arrangement, whether a plugin helped, whether a section works, when the mix is done.

Meters verify: peak safety, loudness delivery, sub content, phase correlation, how hard a processor is working.

Neither decides alone. A mix that measures perfectly and sounds wrong is wrong. A mix that sounds great and true-peaks at +2 dBTP will distort on someone’s phone.

The tool

SPC SpectraScope covers the measuring side — spectrum, loudness with BS.1770 true peak, stereo field and phase analysis — as an analysis-only visualiser that does not touch the audio. Which is the right relationship: the meter observes, you decide.

For the loudness question specifically, how loud should your master be covers targets and why chasing a number is usually the wrong goal, and how to use a spectrum analyser covers reading frequency content without being led by it.