> **This is the raw measurement log for THE_SMOOTHER 2's KEY mode, published in full.**
> It is the document behind the article
> [Resonance Suppressors Can't Tell a Note From a Ring. THE_SMOOTHER 2 Can — Measured.](/blog/the-smoother-2-key-mode-resonance-suppression-measured).
> Nothing has been removed. The result that does not flatter the feature is in section 7.

# THE_SMOOTHER 2 - KEY mode measurement log

Device under test: PsResonanceSuppressor, the shipping THE_SMOOTHER 2 1.3.0 suppressor node (the same code compiled into the VST3/AU/AAX).
Instrument: a headless console harness (smoother_key_test) that instantiates the node in-process, drives it with synthetic audio and measures the output with a direct DFT at each test frequency.
Platform: Apple silicon, Release build, 48 kHz, 512-sample blocks. Date: 2026-09-02.

All figures below are MEASURED on the node, not computed from the formula. 'Level change' is output level minus input level at the exact test frequency, measured over the last 2 s of a 4 s render so latency and attack are excluded. More negative = cut harder. 0.00 dB = untouched.

## 1. Configuration

| Setting | Value |
|---|---|
| Band | one band, 200 Hz - 8 kHz, Suppress mode, DEPTH 0.90, all other band controls at factory default |
| Global | factory defaults (ADAPTIVE on, TRANSIENTS 0.5, MIX 100 %) |
| KEY | A minor (root A, Natural Minor). PROTECT 0.70, BIAS 0.35, TOL 25 cents = factory defaults. 'PROTECT max' column = PROTECT 1.00, BIAS and TOL unchanged |
| Quality levels | ECO (FFT 2048), NORMAL (FFT 4096), HIGH (FFT 8192), ULTRA (FFT 4096 at 2x oversampling), MAX (FFT 8192 at 4x oversampling) |
| Analysis bin width at 48 kHz | ECO 23.4 Hz, NORMAL 11.7 Hz, HIGH 5.9 Hz, ULTRA 23.4 Hz, MAX 23.4 Hz (the oversampled levels analyse at 96/192 kHz, so their bins are wider in Hz) |

## 2. Signals

- **M1, one note at a time.** A steady sine at -20 dBFS on a white-noise bed (Gaussian, sigma 0.01), one render per pitch class from C5 (523.3 Hz) to B5 (987.8 Hz). A steady sine is the purest possible 'stationary narrow peak' - the worst case for a resonance suppressor, because it is indistinguishable from a ring by shape alone.
- **M2, all at once.** Eight steady sines at -24.4 dBFS each on the same noise bed, rendered together: an A minor voicing on A4, E5, A5, E6, C7 and three out-of-key tones on F#5, C#6, D#7.
- Every configuration sees the identical input. The only variable between columns is the KEY setting.

## 3. M1 - one held tone per pitch class, NORMAL quality (FFT 4096, 1x, 11.7 Hz bins)

| Note | Hz | In A minor | KEY off | KEY on (defaults) | KEY on (PROTECT max) |
|---|---|---|---|---|---|
| C5 | 523.2 | yes | −16.52 dB | −5.35 dB | −4.77 dB |
| C#5 | 554.4 | no | −16.47 dB | −47.02 dB | −46.80 dB |
| D5 | 587.3 | yes | −16.54 dB | −3.44 dB | −1.23 dB |
| D#5 | 622.2 | no | −16.49 dB | −49.79 dB | −49.79 dB |
| E5 | 659.3 | yes | −16.55 dB | −4.35 dB | −3.77 dB |
| F5 | 698.5 | yes | −16.56 dB | −4.38 dB | −3.84 dB |
| F#5 | 740.0 | no | −16.50 dB | −51.07 dB | −51.07 dB |
| G5 | 784.0 | yes | −16.52 dB | −2.60 dB | −0.05 dB |
| G#5 | 830.6 | no | −16.47 dB | −49.14 dB | −49.14 dB |
| A5 | 880.0 | yes | −16.52 dB | −2.56 dB | 0.00 dB |
| A#5 | 932.3 | no | −16.51 dB | −53.34 dB | −53.34 dB |
| B5 | 987.8 | yes | −16.50 dB | −2.57 dB | −0.15 dB |

Mean over the 7 in-key notes: KEY off −16.53 dB -> KEY on −3.61 dB -> PROTECT max −1.97 dB.  
Mean over the 5 out-of-key notes: KEY off −16.49 dB -> KEY on −50.07 dB -> PROTECT max −50.03 dB.

## 4. M1 - one held tone per pitch class, HIGH quality (FFT 8192, 1x, 5.9 Hz bins)

| Note | Hz | In A minor | KEY off | KEY on (defaults) | KEY on (PROTECT max) |
|---|---|---|---|---|---|
| C5 | 523.2 | yes | −16.67 dB | −2.57 dB | −0.07 dB |
| C#5 | 554.4 | no | −16.64 dB | −56.85 dB | −56.85 dB |
| D5 | 587.3 | yes | −16.68 dB | −2.57 dB | −0.03 dB |
| D#5 | 622.2 | no | −16.67 dB | −57.82 dB | −57.82 dB |
| E5 | 659.3 | yes | −16.66 dB | −2.57 dB | 0.00 dB |
| F5 | 698.5 | yes | −16.68 dB | −2.57 dB | 0.00 dB |
| F#5 | 740.0 | no | −16.68 dB | −57.68 dB | −57.68 dB |
| G5 | 784.0 | yes | −16.68 dB | −2.57 dB | 0.00 dB |
| G#5 | 830.6 | no | −16.67 dB | −57.45 dB | −57.45 dB |
| A5 | 880.0 | yes | −16.66 dB | −2.56 dB | 0.00 dB |
| A#5 | 932.3 | no | −16.69 dB | −58.96 dB | −58.96 dB |
| B5 | 987.8 | yes | −16.67 dB | −2.57 dB | 0.00 dB |

Mean over the 7 in-key notes: KEY off −16.67 dB -> KEY on −2.57 dB -> PROTECT max −0.01 dB.  
Mean over the 5 out-of-key notes: KEY off −16.67 dB -> KEY on −57.75 dB -> PROTECT max −57.75 dB.

## 5. M2 - held A minor chord plus out-of-key tones, simultaneous, NORMAL quality

| Tone | Hz | In A minor | KEY off | KEY on (defaults) | KEY on (PROTECT max) |
|---|---|---|---|---|---|
| A4 | 440.0 | yes | −14.31 dB | −2.40 dB | −0.01 dB |
| E5 | 659.3 | yes | −12.21 dB | −4.78 dB | −4.52 dB |
| A5 | 880.0 | yes | −12.88 dB | −2.35 dB | −0.13 dB |
| E6 | 1318.5 | yes | −14.88 dB | −2.45 dB | 0.00 dB |
| C7 | 2093.0 | yes | −16.10 dB | −2.54 dB | 0.00 dB |
| F#5 | 740.0 | no | −11.97 dB | −22.14 dB | −21.94 dB |
| C#6 | 1108.7 | no | −14.67 dB | −30.10 dB | −30.12 dB |
| D#7 | 2489.0 | no | −16.53 dB | −53.47 dB | −53.48 dB |

## 6. M2 - held A minor chord plus out-of-key tones, simultaneous, HIGH quality

| Tone | Hz | In A minor | KEY off | KEY on (defaults) | KEY on (PROTECT max) |
|---|---|---|---|---|---|
| A4 | 440.0 | yes | −16.54 dB | −2.57 dB | −0.01 dB |
| E5 | 659.3 | yes | −15.30 dB | −4.25 dB | −4.10 dB |
| A5 | 880.0 | yes | −16.49 dB | −2.56 dB | 0.00 dB |
| E6 | 1318.5 | yes | −16.68 dB | −2.57 dB | 0.00 dB |
| C7 | 2093.0 | yes | −16.61 dB | −2.57 dB | 0.00 dB |
| F#5 | 740.0 | no | −15.78 dB | −37.56 dB | −37.51 dB |
| C#6 | 1108.7 | no | −16.51 dB | −48.59 dB | −47.94 dB |
| D#7 | 2489.0 | no | −16.64 dB | −54.55 dB | −54.57 dB |

## 7. Finding: KEY protection depends on analysis resolution (ECO, ULTRA, MAX)

ECO, ULTRA and MAX all analyse with 23.4 Hz bins at a 48 kHz session (ECO because its FFT is small, ULTRA and MAX because they analyse at 96 and 192 kHz). At 523 Hz a 23.4 Hz bin is 77 cents wide - wider than the TOL window (25 cents, with a falloff to 50 cents) - so bins next to an in-key note fall outside the protected zone and the note is still cut. The three levels produce identical figures because their bin grid is identical.

| Note | In A minor | KEY off | KEY on (defaults) | KEY on (PROTECT max) |
|---|---|---|---|---|
| C5 | yes | −16.48 dB | −19.70 dB | −18.97 dB |
| C#5 | no | −16.52 dB | −38.80 dB | −38.28 dB |
| D5 | yes | −16.51 dB | −17.04 dB | −16.72 dB |
| D#5 | no | −16.51 dB | −39.13 dB | −38.66 dB |
| E5 | yes | −16.52 dB | −16.40 dB | −16.08 dB |
| F5 | yes | −16.49 dB | −14.94 dB | −14.65 dB |
| F#5 | no | −16.47 dB | −36.66 dB | −36.25 dB |
| G5 | yes | −16.50 dB | −11.39 dB | −10.90 dB |
| G#5 | no | −16.49 dB | −36.17 dB | −35.79 dB |
| A5 | yes | −16.52 dB | −10.10 dB | −9.67 dB |
| A#5 | no | −16.48 dB | −50.13 dB | −50.09 dB |
| B5 | yes | −16.47 dB | −5.84 dB | −5.03 dB |

Mean in-key: KEY off −16.50 dB -> KEY on −13.63 dB. Mean out-of-key: KEY off −16.49 dB -> KEY on −40.18 dB. (Identical at ECO, ULTRA and MAX.)

Consequence: the out-of-key side still works at every level, but in-key protection below roughly 1 kHz is only reliable on NORMAL and HIGH. HIGH is the cleanest of the five for KEY. This is a resolution limit of the current release, not a tuning choice, and it is published here rather than omitted.

The M2 chord render at these three levels is also less selective with KEY off (A4, E5 and F#5 are cut by less than 0.4 dB because, at 23.4 Hz resolution, tones 81 to 219 Hz apart merge into one broad hump that the detector does not treat as a narrow peak). The full M2 rows for MAX:

| Tone | In A minor | KEY off | KEY on (defaults) | KEY on (PROTECT max) |
|---|---|---|---|---|
| A4 | yes | −0.36 dB | −0.11 dB | −0.01 dB |
| E5 | yes | −0.04 dB | −0.04 dB | −0.05 dB |
| A5 | yes | −4.11 dB | −1.05 dB | −0.01 dB |
| E6 | yes | −9.52 dB | −1.97 dB | −0.40 dB |
| C7 | yes | −15.79 dB | −2.52 dB | −0.01 dB |
| F#5 | no | −0.15 dB | −0.06 dB | −0.03 dB |
| C#6 | no | −9.15 dB | −17.08 dB | −17.08 dB |
| D#7 | no | −16.42 dB | −48.85 dB | −48.85 dB |

## 8. Notes on method and limits

- No competing product was measured. The 'KEY off' column is THE_SMOOTHER 2 itself with the key prior disabled, which is the conventional stationarity-only behaviour every resonance suppressor shares. It stands in for the category, not for any named rival.
- Steady sines are the worst case. Real notes carry vibrato, decay and onsets, all of which the stationarity detector already partly spares, so the KEY-off cuts on real material are smaller than the tabletop figures here. The direction of the KEY-on change is the same.
- 'Out of key' is a musical judgement, not a physical one. A chromatic passing note, a borrowed chord or a modulation is out of key and is still music. That is why KEY is off by default, why PROTECT and BIAS are separate controls, and why the key acts as a prior on the detector rather than a hard mask.
- With BIAS above zero the out-of-key depth can be pushed past the band's DEPTH setting; the node caps the multiplier at 1.25x so the gain interpolation stays well-behaved. That is why out-of-key cuts land far deeper than the KEY-off figure rather than a fixed number of dB below it.
- The harness is a test build that links the shipping suppressor source directly; it is not the plugin UI. Band, key and quality parameters are set through the same parameter interface the plugin uses.