← Back to Sidestep EQ · Funk De Nacht
Sidestep EQ 1.1.0: Sound Quality Specification
92 checks passed, 6 within a documented limit, 0 failed in the signal-path suite, and 96 of 96 checks passed in the real plug-in host test. Every value is measured on the shipping audio code.
Results by area
| Claim | Method | Threshold | Measured | Result |
|---|---|---|---|---|
| Clean bass while ducking Ducking a bass note adds no distortion beyond the volume change itself. | A 55-100 Hz tone is ducked; harmonics and noise are compared with an ideal volume change of the same shape. | warn > -60, fail > -25 | -30.82 dB re carrier worst of 10 checks | WARN |
| No aliasing or intermodulation Boosts and ducking do not create false tones (aliasing) or intermodulation. | Sweeps and two-tone signals pass through; everything that is not the expected filtered signal is measured. | warn > -90, fail > -60 | -64.60 dB re output worst of 5 checks | WARN |
| Smooth parameter moves Moving a knob or switching a value never clicks or buzzes, at any host buffer size. | A pure tone is played while a parameter sweeps or jumps; the high-frequency residue is measured. | warn > -90, fail > -70 | -94.44 dB re carrier worst of 10 checks | PASS |
| Same sound at any buffer size The output does not depend on the host buffer size. | The same signal is rendered at buffer sizes from 32 to 4096 samples and compared. | warn > -140, fail > -120 | -200.00 dB re output worst of 5 checks | PASS |
| Accuracy near the top of the spectrum EQ curves keep their shape up to 20 kHz, even at 44.1 kHz. | Fifteen demanding curves are compared with measurements of the reference EQ, 20 Hz to 20 kHz. | warn > 0.5, fail > 1 | 0.84 dB worst of 7 checks | WARN |
| Linear phase Linear-phase mode reports its true latency and keeps phase intact. | An impulse is sent through; the position of its peak is compared with the reported latency. | warn > 0.5, fail > 2.5 | 0.00 samples worst of 15 checks | PASS |
| Auto Gain Auto Gain keeps the perceived loudness steady when the curve changes. | Pink noise through six typical curves; loudness change is measured with a K-weighted meter. | warn > 0.5, fail > 1 | 0.32 LU worst of 1 checks | PASS |
| Bypass and null accuracy Bypassed or flat settings return the input exactly; Mid/Side and Left/Right bands leave the other channel untouched. | Output is subtracted from the expected signal and the residue is measured. | warn > -120, fail > -100 | -145.81 dB re output worst of 7 checks | PASS |
| Compressor, output gain and Duck mix The compressor follows its settings, and Out gain and Duck mix are exact. | Step signals measure attack and release times; static levels measure the compression law and gain. | warn > 0.01, fail > 0.05 | 0.00 dB worst of 8 checks | PASS |
| Stability Extreme settings and invalid input (NaN, infinity) never produce invalid output. | Random and extreme settings, plus corrupted input, are run; any non-finite output sample is counted. | must be zero | 0.00 samples worst of 8 checks | PASS |
| Vintage and analog-feel modes Vintage mode adds a controlled amount of warmth, is neutral when off and stays level-matched. | Harmonics, drift, loudness and bypass identity are measured with vintage on and off. | must be zero | 0.00 violations worst of 10 checks | PASS |
| Filter behaviour and resonance Low-cut/high-cut resonance and gliding filters behave like their design and stay smooth. | Resonance peaks are compared with the design formula; moving filters are checked for clicks. | warn > 1, fail > 3 | 0.58 dB worst of 12 checks | PASS |
Thresholds are "smaller is better": WARN above the first limit, FAIL above the second; the threshold shown belongs to the worst check in the area.
Plug-in host test
| Scenario | Passed | Result |
|---|---|---|
| Vintage mode | 3 / 3 | PASS |
| Transparent when flat | 10 / 10 | PASS |
| Mono and stereo agree | 5 / 5 | PASS |
| Load time | 1 / 1 | PASS |
| Duck curves match their reference | 16 / 16 | PASS |
| Tempo sync | 3 / 3 | PASS |
| Mix controls | 6 / 6 | PASS |
| Any buffer size | 3 / 3 | PASS |
| 44.1, 48 and 96 kHz | 3 / 3 | PASS |
| Noise input | 3 / 3 | PASS |
| Range and depth | 8 / 8 | PASS |
| Triggers | 4 / 4 | PASS |
| Compressor | 6 / 6 | PASS |
| Per-band Smart Duck | 25 / 25 | PASS |
About the warnings
The duck curve changes level within 1-2 milliseconds. A volume change that fast necessarily spreads some energy around the tone. This is the intended curve shape, not distortion: the harmonics added beyond an ideal volume change measure zero. Softening the edge would change the matched pumping feel.
A 16 kHz bell changes its phase as its gain moves, and the very fast duck edge makes its coefficients glide. With slower edges the same measure improves by 11 dB.
Same cause: the fast edges of the duck curve. Without ducking the same test measures -120 dB.
The worst of 15 deliberately extreme curves (a narrow -12 dB cut at 85 % of the Nyquist frequency) differs from the reference EQ by 0.84 dB at 44.1 kHz and 0.55 dB at 48 kHz. All other curves are within 0.6 dB. The limit for a pass is 0.5 dB.
Known limits
- Measurements are made in a test host with synthetic signals; they describe the signal path, not subjective sound quality.
- Near-Nyquist accuracy was fitted against reference measurements for Bell, shelf, low-pass and notch; slopes steeper than 12 dB/octave, Tilt, Band Pass and low shelves near Nyquist have no reference data.
- The MIDI trigger and the external side-chain bus cannot be reached from the test host; they are covered by separate automated tests.
- The graphical interface is tested separately and is not part of this specification.
How it was measured
The audio code is run offline in a test host with synthetic signals (sines, sweeps, noise, impulses) at 44.1, 48 and 96 kHz and buffer sizes from 32 to 4096 samples. Each result is a number compared with a stated limit. The finished plug-in file is then loaded in a separate third-party host and a further set of checks is run on it. Results are regenerated and published for every release.