Purpose
Frequency response alone doesn't tell you if a peak is EQ-able or if a dip is a cancellation. Excess group delay is the test: flat = minimum-phase = invertible with EQ. Spiky = non-minimum-phase = don't boost. ETC and Filtered IR reveal reflection timing and decay per band.
Numbered Workflow
- Group Delay: Open Group Delay graph. If you see a large constant offset, you have system delay — use Estimate IR Delay and Offset t=0 to remove it. Otherwise phase slope masks everything.
- Generate Minimum Phase: Controls → Generate Minimum Phase. REW computes minimum-phase version of your magnitude response.
- Excess Group Delay: Enable excess group delay overlay. Flat regions = minimum-phase (EQ works). Peaks/spikes = excess phase (reflections, non-minimum-phase) — don't EQ those dips.
- ETC: Impulse → ETC, smoothing 0. Each spike after direct = reflection. Measure Δt to direct, compute extra path:
distance = Δt × 343 m/s. 3ms spike = ~1m longer path (likely floor/ceiling). - Filtered IR: Open Filtered IR → choose octave or 1/3-octave. View band-limited IR, ETC, and Schroeder integral. Use to see which bands ring and to time-align subs (sub IR peak is broad — use IR start, not peak, for delay).
- Alignment check: Compare excess group delay before/after alignment. Flatter = better integration.
Worked Example — Is This 50Hz Peak EQ-able?
You measure a +12dB peak at 52Hz with long ringing in waterfall. Excess group delay is flat around 52Hz (no spike). → Minimum-phase resonance → cut filter will work. Adjacent -15dB dip at 78Hz has a sharp excess GD spike → non-minimum-phase cancellation → do not boost.
Action: Apply narrow cut at 52Hz, leave 78Hz alone, treat or reposition to address dip.
Outputs & Interpretation
- Group Delay (ms vs Hz): High GD at LF is normal (port, crossover). Sharp spikes = resonances or reflections.
- Excess GD: The decision tool. Flat = EQ-able. Use it to set Match Range in EQ window.
- ETC spikes: Time → distance. Identify first reflections for treatment (panels at reflection points).
- Filtered IR Schroeder: Shows decay slope per band. RT60 metrics (EDT, T20, T30) appear here too.
Deeper Detail — Minimum Phase Math & Filtered IR Modes
Minimum Phase: System where phase is Hilbert transform of log magnitude. Lowest possible delay for that |H(f)|. Can be inverted: H_inv = 1/H. Non-minimum-phase systems have extra all-pass component (excess phase) that EQ cannot invert.
Why sub delay uses IR start: Sub IR is not impulsive — it's a broad hump (low frequencies have long period). Peak is ambiguous. IR start (first rise above noise) is more consistent for delay estimation.
Filtered IR filtering: Forward (causal), time-reversed (zero-phase-ish), zero-phase (non-causal). Filter order affects group delay — higher order = steeper but more ringing. Use order 6 as balanced default.
RT60 metrics: EDT (0 to -10dB), T20 (-5 to -25dB), T30 (-5 to -35dB), Topt (best fit). Curvature indicates non-exponential decay.
Check Your Understanding
Two peaks: A at 45Hz with flat excess GD, B at 120Hz with +15ms excess GD spike. Which do you EQ?