# Resonance Suppressors Can't Tell a Note From a Ring. THE_SMOOTHER 2 Can — Measured.

> The oldest complaint about resonance suppressors is true: they take the good stuff with the bad, because a held note and a ringing resonance look identical to them. THE_SMOOTHER 2's KEY mode knows what key your song is in. Measured on the shipping engine: an in-key note goes from a 16.7 dB cut to 2.6 dB, an out-of-key ring from 16.7 dB to 58 dB.

- Author: [Phil Speiser](https://www.philspeiser.com/about-me)
- Published: 2026-09-02
- Canonical: https://www.philspeiser.com/blog/the-smoother-2-key-mode-resonance-suppression-measured
- Product: [THE_SMOOTHER 2](https://www.philspeiser.com/product/the_smoother-2)

**Key finding:** With KEY mode on, THE_SMOOTHER 2 cuts a held in-key note by 2.6 dB instead of 16.7 dB, and an out-of-key ring by 58 dB instead of 16.7 dB, measured on the shipping engine at HIGH quality.

You put a resonance suppressor on the vocal because the 3 kHz bite was making your ears close up. And it worked — the bite is gone. But now you A/B it for the fifth time and something else is gone too. The vocal is a little thinner. The pad behind it lost some of its body. You can't point at a frequency, you just know the track got smaller.

So you back the depth off until the harshness half-returns, and settle for the compromise.

**"Resonance suppressors take the good stuff with the bad."**

I've heard that complaint for years, and I want to say clearly: it's correct. Not a myth, not user error. It's built into how every resonance suppressor on the market works, mine included — until this update.

## Why the complaint is right

Every automatic resonance suppressor does one job under the hood: it looks for **narrow, steady peaks that stick out above the spectrum around them**. That's the fingerprint of a ring — a room mode, a resonant filter, a harsh formant, a snare wire that won't stop singing. Narrow, stationary, prominent. The detector finds those three properties and pulls them down.

Now describe a sustained note. Narrow. Stationary. Prominent above its surroundings.

**Same three properties. Same fingerprint.**

A held pad chord, the fundamental of a bass line, the third harmonic of a singer holding a note, a guitar's open string ringing under a riff — to a stationarity detector, these are indistinguishable from a resonance. It doesn't know there's a song. It has never been told. So it does exactly what it was built to do and cuts them, and you hear that as the track getting thinner.

I can put a number on it, because I measured it on THE_SMOOTHER 2 itself with the new feature switched off — which is the conventional behaviour every suppressor shares. One band from 200 Hz to 8 kHz at 0.9 depth. A single steady tone at −20 dBFS, one pitch class at a time across an octave:

| Held tone, no key information | Level change |
|---|---|
| C5, in key of A minor | −16.67 dB |
| C#5, out of key | −16.64 dB |
| E5, in key | −16.66 dB |
| F#5, out of key | −16.68 dB |
| A5, in key | −16.66 dB |
| A#5, out of key | −16.69 dB |

Twelve notes, twelve cuts, all within 0.05 dB of each other. The suppressor has no idea which of those notes belong to your song. **It can't. Nothing in the audio tells it.**

That's the whole problem in one table. And it's why I stopped trying to make the detector smarter about *shape* — a note and a ring have the same shape — and started giving it the one piece of information it was missing.

## So I taught it the key of your song

THE_SMOOTHER 2 has always answered one question about every peak: *"is this ringing?"* Since version 1.3.0 it can weigh a second one: **"is this note supposed to be here?"**

Click the new **KEY** cell in the top bar, pick a root and a scale — Major, Minor, Harmonic Minor, Pentatonic Major, Pentatonic Minor or Blues — and switch it on. From that moment every analysis bin carries a weight for how close it sits to an in-scale note, and that weight tilts the suppression:

- **In-key content** — the music — has its suppression depth pulled back by up to the **PROTECT** amount.
- **Out-of-key ringing** — the junk — gets pushed harder, by up to the **BIAS** amount.
- **TOL** sets how many cents off-pitch a peak can be and still count as in key, with a gentle falloff beyond it. That falloff is deliberate: the upper harmonics of a perfectly in-tune note aren't on the equal-tempered grid (the 5th harmonic sits 14 cents flat, the 7th 31 cents flat), and they need protecting too.

Three details matter here, and I'd rather you knew them than found out:

**It's a prior, not a mask.** The stationarity detector is still in charge. If nothing is ringing, KEY does nothing. If an in-key note *is* genuinely ringing out of control, it still gets treated — just more gently. The key doesn't override the detector's judgement, it informs it.

**It only works on the cut side.** Suppress mode and the cutting half of Shape mode. KEY can never boost anything, and it never touches your band settings.

**It fades out where key stops meaning anything.** Below about 120 Hz the analysis bins get coarser than a semitone, and above about 9 kHz partials drift inharmonic and ringing turns noise-like. In both regions the key influence fades to neutral, so it never bites the sub or the air band.

## Measured: the same twelve notes, key switched on

Same signal, same band, same depth. The only change is that the plugin now knows the song is in A minor. HIGH quality, factory PROTECT / BIAS / TOL.

*Figure: One held tone per pitch class — cut depth with KEY off and on. C5 to B5, one tone at a time at −20 dBFS on a noise bed. Key set to A minor. Shaded columns are the five notes outside the scale. Every point measured on the shipping engine; hover for figures. Suppression applied by THE_SMOOTHER 2 to a held tone on each of the twelve pitch classes from C5 to B5, key set to A minor, HIGH quality. With KEY off every note is cut by about 16.7 decibels regardless of pitch. With KEY on at factory defaults the seven in-key notes are cut by 2.6 decibels and the five out-of-key notes by about 58 decibels. With PROTECT at maximum the in-key notes are cut by 0.07 decibels or less. The grey line is THE_SMOOTHER 2 with KEY off — the conventional behaviour of every resonance suppressor. Gold is KEY on at factory settings; dashed cyan is PROTECT at maximum. The table below carries NORMAL quality alongside HIGH.*

Read the gold line against the grey one. **The seven in-key notes go from a 16.7 dB cut to 2.6 dB.** Push PROTECT to maximum and they come through at 0.07 dB or less — effectively untouched. The five out-of-key notes go the other way: **from 16.7 dB to between 56.9 and 59.0 dB**, which in practice means gone.

That's the same detector, the same band, the same 4 seconds of audio. The only thing that changed is that it finally knows which notes are yours.

## Measured: a chord and three rings, all at once

One note at a time is the clean experiment. A mix is not one note at a time. So the second test plays eight steady tones together: an A minor voicing on A4, E5, A5, E6 and C7, plus three out-of-key tones on F#5, C#6 and D#7 sitting in between them — the way a ringing resonance actually sits inside a chord.

*Figure: Held A minor chord + out-of-key tones, played together. Eight simultaneous tones at −24 dBFS each. The five chord tones on the left, the three out-of-key tones in the shaded region. HIGH quality, key A minor. Every point measured. Level change at each of eight simultaneous held tones: an A minor chord on A4, E5, A5, E6 and C7 plus out-of-key tones on F sharp 5, C sharp 6 and D sharp 7. With KEY off all eight are cut between 15.3 and 16.7 decibels. With KEY on in A minor the chord tones are cut between 2.6 and 4.2 decibels while the out-of-key tones are cut between 37.6 and 54.5 decibels. With KEY off the suppressor cuts all eight tones by 15–17 dB — it treats the chord and the rings identically. With KEY on the chord loses 2.6–4.3 dB while the out-of-key tones lose 37.6–54.6 dB.*

With KEY off, look at what happens to the chord: A4 down 16.5 dB, E6 down 16.7 dB, C7 down 16.6 dB. The suppressor has flattened the harmony as thoroughly as the junk. That is the "thinner" you've been hearing.

With KEY on, the chord loses between 2.6 and 4.3 dB and the three out-of-key tones lose between 37.6 and 54.6 dB. The one that fared best for the junk, F#5, sits a whole tone above E5 and a minor third below A5 — surrounded by loud in-key neighbours that raise the local average — and it still went from a 15.8 dB cut to a 37.6 dB cut. **The rings are gone. The chord is still a chord.**

## SMART knows the key too

If you use SMART detection — where THE_SMOOTHER 2 listens and proposes bands for you — the key prior feeds the ranking as well. With KEY on, out-of-key stationary peaks show up as more confident suggestions and in-key peaks as less confident ones. So the brightest markers on the analyzer really are the safest cuts, and a held note in the pad no longer gets flagged as a problem to solve.

While the KEY popup is open, the in-scale note lines are drawn straight onto the spectrum, root octaves brighter. You can line up a band with the music by eye, and you can see at a glance whether the peak you're about to chase sits on a scale note or between two.

## Two honest notes

**Steady tones are the worst case, on purpose.** A steady sine is the purest possible "narrow stationary peak", so the KEY-off cuts above are as bad as it gets. Real notes carry vibrato, decay and onsets, and the detector already spares part of that, so on real material the conventional damage is smaller than 16.7 dB. The direction of the change with KEY on is exactly the same, and you'll hear it most on the material that suffers most: sustained pads, held bass notes, organs, long vocal notes, anything that holds a pitch.

**KEY needs resolution.** It works on the analyser's frequency grid, and on ECO, ULTRA and MAX that grid is 23.4 Hz wide at 48 kHz (ECO because its FFT is small, ULTRA and MAX because they analyse at 96 and 192 kHz). Below roughly 1 kHz that's wider than the tolerance window, so in-key protection weakens there. I measured it rather than guessing: at MAX the same seven in-key notes are still cut by an average of 13.6 dB with KEY on, against 2.6 dB at HIGH. The out-of-key side works at every level. So for now the rule is simple — **use KEY on NORMAL or HIGH**, which is where you're mixing anyway, and I've published the full numbers for all five levels in [the measurement log](https://www.philspeiser.com/measurements/the-smoother-2-key) rather than leaving that one out.

## How I use it

> **Key-aware cleanup in three moves**
>
> **1. Tell it the key.** Click **KEY**, switch it on, set root and scale. If you don't know the key, [WHATTHEKEY 3](https://www.philspeiser.com/product/whatthekey-3) will tell you in seconds.
> 
> **2. Add your bands as usual** — or hit SMART and let it propose them. The markers now carry the key prior.
> 
> **3. Balance PROTECT and BIAS.** Raise PROTECT until the notes stop moving and only the junk does. Raise BIAS when a ring is stubborn. Nudge TOL up for material that drifts a little flat or sharp — an old synth, a live bass.
> 
> Then reach for a deeper DEPTH than you'd ever have dared before. That's the real payoff: you can finally let the suppressor work hard, because it's working on the right things.

And when the song modulates, or a borrowed chord walks through, or the singer bends a note through a passing tone? Out of key isn't the same as wrong — that's why KEY is off by default, why PROTECT and BIAS are separate knobs, and why the detector keeps the final word. You're still the one deciding what the song is. The plugin just stops pretending it doesn't have one.

## Same update, one more thing

Version 1.3.0 also added a **LIVE** quality level: the spectral detector keeps running for analysis, but the audio flows through a time-domain filter bank instead — **0 samples of latency**, for tracking vocals through it or using it on stage. Bounces automatically switch back to the full spectral engine. Both features are opt-in; if you change nothing, your presets and projects sound exactly as before.

## The concern, and the answer

The concern was never wrong. Every resonance suppressor cuts the music with the noise, because by shape alone a held note *is* a resonance. Measured: 16.7 dB off an in-key note, the same as off a ring.

The answer isn't a cleverer shape detector. It's context. Tell [THE_SMOOTHER 2](https://www.philspeiser.com/product/the_smoother-2) the key, and an in-key note comes through at 2.6 dB down — or 0.07 dB with PROTECT maxed — while the out-of-key ring drops 58 dB. Same engine, same band, same audio.

**The junk gets caught. The music is spared. And you get to turn the depth up instead of down.**

> **A resonance suppressor that knows your song.** THE_SMOOTHER 2 with KEY mode — tell it the key, and it takes the music's side. Measured on the shipping engine, log published in full. [THE_SMOOTHER 2](https://www.philspeiser.com/product/the_smoother-2)

## How I measured this

- **Material:** Held sine tones at -20 dBFS on all twelve pitch classes, plus a full A minor chord with three out-of-key tones playing together
- **Versions:** THE_SMOOTHER 2 1.3.0, the shipping engine
- **Measured on:** 2026-09
- **Settings:** Key set to A minor, one band from 200 Hz to 8 kHz at 0.9 depth; KEY off, KEY on at factory PROTECT, BIAS and TOL, and KEY on with PROTECT at maximum; Quality at ECO, NORMAL, HIGH, ULTRA and MAX
- **Held constant:** Same tones, level, band and depth in every run; only the KEY switch and the quality level changed
- **Notes:** Headless harness instantiating the shipping suppressor node in-process, Apple silicon Release build, 48 kHz, direct DFT at each test frequency over the last 2 s of a 4 s render

## FAQ

### Do resonance suppressors remove harmonics and musical content along with resonances?

Yes, by design. A resonance suppressor looks for narrow, steady peaks that stick out of the surrounding spectrum. A ringing room mode or a harsh vocal formant is exactly that, but so is a held note, a sustained pad or the harmonics of a bass line. By shape alone the two are indistinguishable. Measured on THE_SMOOTHER 2 with its key prior switched off, a held tone at -20 dBFS is cut by 16.7 dB on every one of the twelve pitch classes, whether the note belongs to the song or not.

### What does THE_SMOOTHER 2's KEY mode do?

You tell THE_SMOOTHER 2 the root note and scale of your song. Every analysis bin then gets a weight for how close it sits to an in-scale note. In-key content has its suppression depth pulled back by up to the PROTECT amount, out-of-key ringing is pushed harder by up to the BIAS amount, and TOL sets how many cents off-pitch still counts as in key. The key is a prior on the detector, not a hard mask: the stationarity detector still decides whether something is ringing, and KEY only tilts how hard it acts. It works on the cut side only, so it can never boost anything.

### How much difference does KEY mode make, measured?

On the shipping THE_SMOOTHER 2 engine at HIGH quality, key set to A minor, one band from 200 Hz to 8 kHz at 0.9 depth: a held in-key tone is cut by 16.7 dB with KEY off and by 2.6 dB with KEY on at factory settings, or 0.07 dB or less with PROTECT at maximum. A held out-of-key tone is cut by 16.7 dB with KEY off and by 56.9 to 59.0 dB with KEY on. With a full A minor chord and three out-of-key tones playing together, the chord tones are cut between 2.6 and 4.3 dB and the out-of-key tones between 37.6 and 54.6 dB.

### Which quality setting should I use with KEY mode?

NORMAL or HIGH. KEY mode works on the analyser's frequency grid, and on ECO, ULTRA and MAX that grid is 23.4 Hz wide at 48 kHz, which below roughly 1 kHz is wider than the tolerance window, so in-key protection weakens there. Measured at MAX, in-key notes in octave 5 are still cut by an average of 13.6 dB with KEY on. At HIGH the same notes are cut by 2.6 dB. The out-of-key side works at every quality level.

### Were competing resonance suppressors measured for this article?

No. The KEY-off figures are THE_SMOOTHER 2 itself with the key prior disabled, which is the conventional stationarity-only behaviour that every resonance suppressor shares. No rival product was measured, ranked or anonymised.

### Does KEY mode change my existing presets or projects?

No. KEY is off by default, older presets load with it off, and it never touches your band settings. When it is on, key settings are saved with your presets.