⚙️ ENGINEER LEVEL: Statistical Methods and Room Correction
Optimal EQ Filter Placement
Psychoacoustic frequency resolution:
The auditory system resolves frequency in critical bands (Bark scale). EQ Q values should be matched to these bandwidths for most natural-sounding corrections.
| Frequency | Critical BW | Optimal Q |
|---|---|---|
| 60 Hz | 100 Hz | 0.6 |
| 200 Hz | 100 Hz | 2.0 |
| 500 Hz | 110 Hz | 4.5 |
| 1,000 Hz | 160 Hz | 6.3 |
| 4,000 Hz | 700 Hz | 5.7 |
| 10,000 Hz | 2,500 Hz | 4.0 |
Narrow corrections at bass frequencies (high Q at low frequency) produce audible ringing in time domain and rarely correspond to real problems. At midrange, narrower Q is appropriate for resonances.
Least-Squares EQ Optimization
Given a measured response M(ω) and target T(ω), find EQ filter E(ω) that minimizes:
ε = Σ W(ω) × |T(ω) − E(ω) × M(ω)|²
Where W(ω) is a perceptual weighting function (higher weight in midrange where ears are more sensitive).
Constrained solution:
Adding regularization prevents over-correction:
ε = Σ W(ω)|T − E×M|² + λ Σ|E(ω)|²
Higher λ → smoother, more conservative EQ. Lower λ → aggressive, closer to target but with risk of overcorrection.
Iterative refinement:
No closed-form solution works perfectly due to: - Measurement noise - Time-varying acoustics - Finite filter bank constraints
Use iterative algorithms: 1. Compute initial EQ estimate 2. Apply virtually, re-simulate response 3. Compute new error 4. Adjust filter parameters 5. Repeat until error < threshold (typically 1–2 dB)