This document describes Jivetalking's gate implementation, inspired by the legendary **Drawmer DS201 Dual Noise Gate**—the industry-standard hardware gate since 1982.

The DS201 pioneered several innovations that became standard in professional gating:

| Parameter | DS201 Range | Notes |

|-----------|-------------|-------|

| **Attack** | 10µs – 1 sec | Microsecond-level attack preserves natural transients |

| **Hold** | 2ms – 2 sec | Keeps gate open after signal drops below threshold |

| **Decay** | 2ms – 4 sec | Smooth fade-out rate after hold expires |

| **Threshold** | +20dB to -54dB | Wide range for various signal levels |

| **Range** | Up to 90dB | Full mute capability |

| **HP Filter** | 25Hz – 4kHz | Side-chain high-pass for frequency-conscious gating |

| **LP Filter** | 250Hz – 35kHz | Side-chain low-pass for frequency-conscious gating |

1. **Frequency-conscious gating** — Variable HP/LP filters on the side-chain allow the gate to respond only to specific frequency ranges, preventing false triggers from bleed or rumble.

2. **Ultra-fast attack** — Opens in microseconds to preserve the natural attack of transients, crucial for drums and percussive speech consonants.

3. **Four-stage envelope** — Attack, Hold, Decay, Range provides precise control over gate behaviour.

4. **Key Listen** — Monitor the filtered side-chain to tune filter settings.

We implement the DS201's philosophy using FFmpeg's filters, with adaptations optimised for spoken word.

Rather than side-chain filtering (which FFmpeg doesn't support), we apply frequency filtering to the audio path before gating:

This achieves the same goal: the gate sees a frequency-filtered signal, preventing false triggers from:

- **Low-frequency rumble** (handled by high-pass)

- **Mains hum** (handled by notch filters at 50/60Hz harmonics)

- **Ultrasonic noise** (handled by low-pass)

Bundles two DS201-inspired filters:

1. **High-pass filter** — Removes subsonic rumble that could hold the gate open

- Adaptive frequency: 60–120Hz based on voice character

- Protects warm voices with gentler slopes and mix blending

2. **Mains hum notch** — Surgical removal of tonal hum

- 50Hz (UK/EU) or 60Hz (US) fundamental

- Up to 4 harmonics with 0.3–2Hz notch width

- Enabled adaptively when silence entropy indicates tonal noise

Removes high-frequency content that could cause false triggers:

- **Adaptive tuning** based on SpectralRolloff:

- Rolloff < 8kHz → disabled (voice already dark)

- Rolloff > 14kHz → enabled at rolloff + 2kHz

- High ZCR + low centroid → possible HF noise, enable at 12kHz

- **Conservative approach** — never cuts below 8kHz to preserve sibilance and air

- **Default: disabled** — only activates when measurements indicate benefit

Here we intentionally depart from the DS201's hard gate behaviour.

The DS201 offers two modes:

- **Hard gate** — Ultra-fast, clean cut, ideal for drums

- **Soft gate** — Gentler expansion for vocals and mixes

For podcast speech, we exclusively use a soft expander approach:

| Aspect | DS201 Hard Gate | Jivetalking DS201Gate |

|--------|-----------------|----------------------|

| **Ratio** | ∞:1 (complete mute) | 1.5:1 – 2.5:1 (gentle reduction) |

| **Knee** | Sharp | Soft (2–5 dB) |

| **Range** | Up to 90dB | -12dB to -36dB |

| **Character** | Absolute silence | Natural fade |

**Rationale:** Hard gating on speech creates unnatural "pumping" artifacts. A soft expander reduces noise between phrases while maintaining the natural room tone that listeners expect.

The DS201's microsecond attack is legendary for preserving transients. We implement adaptive attack based on measured transient characteristics:

| Condition | Attack Time | Use Case |

|-----------|-------------|----------|

| MaxDifference > 40% OR SpectralCrest > 40dB | 0.5ms | Extreme transients (plosives) |

| MaxDifference > 25% OR SpectralCrest > 30dB | 3ms | Sharp consonants |

| MaxDifference > 10% | 7ms | Normal speech |

| Soft delivery | 15ms | Gentle fade-in |

**Measurements used:**

- `MaxDifference` — Sample-to-sample change, catches plosives ("P", "T", "K")

- `SpectralCrest` — Spectral peak-to-RMS ratio, indicates transient energy

- `SpectralFlux` — Frame-to-frame spectral change, biases toward faster attack

The DS201 has a dedicated Hold parameter; FFmpeg's `agate` does not. We compensate by:

- Adding 50ms to the release time as baseline hold compensation

- Adding 75ms extra for tonal noise (hides pumping artifacts)

- Release range: 150–500ms (vs DS201's 2ms–4s decay)

All gate parameters adapt to Pass 1 measurements:

| Parameter | Adaptation Logic |

|-----------|------------------|

| **Threshold** | Based on noise floor + silence peak, with headroom for noise severity |

| **Ratio** | Based on LRA (wide dynamics → gentle ratio to preserve expression) |

| **Attack** | Based on MaxDifference + SpectralCrest + SpectralFlux |

| **Release** | Based on SpectralFlux + ZCR + noise character |

| **Range** | Based on silence entropy (tonal → gentle, broadband → aggressive) |

| **Knee** | Based on SpectralCrest (dynamic → soft knee) |

| **Detection** | RMS for tonal bleed, peak for clean recordings |

| Feature | DS201 | Jivetalking |

|---------|-------|-------------|

| **Frequency-conscious filtering** | ✅ HP/LP on side-chain | ✅ HP/LP on audio path |

| **Ultra-fast attack** | ✅ 10µs minimum | ✅ 500µs minimum |

| **Hold parameter** | ✅ 2ms–2s | ⚠️ Compensated via release |

| **Hard gate mode** | ✅ | ❌ (soft expander only) |

| **Soft gate mode** | ✅ | ✅ (always) |

| **Adaptive parameters** | ❌ (manual) | ✅ (measurement-driven) |

| **Detection modes** | Not specified | ✅ RMS/Peak adaptive |

- [Drawmer DS201 Product Page](https://drawmer.com/products/pro-series/ds201.php)

- [Drawmer DS201 Operator's Manual](https://drawmer.com/uploads/manuals/ds201_operators_manual.pdf)

- FFmpeg [agate filter documentation](https://ffmpeg.org/ffmpeg-filters.html#agate)

- FFmpeg [highpass](https://ffmpeg.org/ffmpeg-filters.html#highpass) / [lowpass](https://ffmpeg.org/ffmpeg-filters.html#lowpass) / [bandreject](https://ffmpeg.org/ffmpeg-filters.html#bandreject) filter documentation