You cannot fix a local problem globally

A producer asks me why his lead vocal still sounds harsh after he has “fixed” it three times. I look at the chain. A high shelf pulling the whole top down, the fader riding lower than he wants, a broadband de-esser clamping every “s” in the song. The vocal is now dull, quiet, and lisp-free, and still harsh in the one spot that started all of it: the held note in the chorus where the voice digs in around 3 kHz. He has spent three moves making the whole vocal worse to chase a problem that lives in one narrow place.

The frame

Masking, and harshness with it, is a local event. The fix you reach for is almost always a global one. That mismatch is the whole problem.

Your ear does not hear frequency as one WIDE field. It hears through a bank of overlapping filters, each one narrow, each tuned to a small region. Harry Fletcher proved this back in 1940: a sound is only masked by other energy inside a narrow “critical band” around it. Energy far away in frequency does almost nothing. The interaction is local. A problem at 3 kHz is a problem at 3 kHz, not a problem with “the highs.”

So when you pull a WIDE shelf or drop the whole fader, you are treating a single street with a national policy. You dull the parts that were fine, the air, the breath, the consonants that carried the intimacy, and the actual harsh band, being narrow and dynamic, mostly survives.

The part that makes it worse

There is a wrinkle that makes the global move even more wasteful, and it is one of the older facts in hearing science.

Masking is not symmetric. A loud low sound masks the frequencies above it far more than the ones below. Wegel and Lane measured this in 1924, and people have been refining the numbers ever since: the masking spreads upward, and it spreads further the louder the masker gets.

Read that the right way and it’s a gift. If your vocal sounds harsh or congested up top, the cause is often not up top at all. It’s a buildup lower down, throwing a masking shadow upward over everything above it. Clean a little mud around 250 Hz and the top can open up on its own, without you touching the top at all. The local fix in the right place beats the global fix in the wrong one, and it costs you nothing you wanted to keep.

The same shape, outside the studio

I can’t write about this without seeing it everywhere else, because it’s the same error I make in the rest of my life.

Something goes wrong in one corner, and the instinct is a global response. One client churns, so you rewrite the whole onboarding. One week feels heavy, so you redesign your entire calendar. One note in the mix is harsh, so you turn the whole thing down. The global move feels thorough, responsible, like real action. Usually it just dulls everything that was working and leaves the actual problem, which was always local, mostly intact. The skill, in a mix and out of it, is resisting the satisfying broad gesture long enough to find where the problem actually lives.

The local tools

The in-the-mix version of “go local” is simple to say and a little harder to do: find the exact band, fix it only there, and let it act only when the problem is actually happening. A narrow dynamic move on the offending band leaves the rest of the spectrum untouched, which is the entire point. That is what dynamic resonance suppression does, and on the harsh-vocal case it’s why SMOOTH exists. It watches each band and only pulls the one that spikes, only while it spikes, weighted the way the ear’s own filters are weighted.

The mud-masking-the-top case isn’t even a plugin, it’s a thirty-second check. I keep CHECK, the free analyser, open just to see where the energy actually sits before I reach for anything. Half the time the harsh top is a low-mid problem and I would never have guessed from listening alone.

What this changes

Next time something sounds harsh, before you reach for a global move, ask one question: where does this actually live? Narrow your hearing to the band. Then make the smallest move that fixes that band and nothing else. And check the octaves below the problem before you touch the problem itself, because the upward spread of masking means the cause is often one floor down.

From the studio

Last week’s note on stacking pulled a wave of replies about layering vocals, and a few of you pushed on whether a per-band move is always worth the CPU over a clean static cut. Fair question. That argument is turning into a future issue.

What is the most global fix you have ever thrown at the most local problem? I collect these, and they are almost always funny in hindsight. Reply.

Jonas