You don't need this chapter to get a great result β the defaults are chosen to work well in most rooms. Come here only if you want to understand or nudge how the correction behaves. These options live in Settings β Calibration: an Applied Correction summary sits at the top, and below it the options are grouped into Measurement, Alignment, Room Correction and Display.
Measurement
- Quick Positions / Full Positions β how many microphone positions to measure per speaker. More positions average out the room better but take longer. The How many positions should I use? panel explains the trade-off; 5 is the recommended balance for most rooms.
- Clock Drift Correction β Enable only when your mic input and speaker output are on different audio interfaces. Default Disable. (Same setting as on the measurement screen β see Chapter 8 β Measuring your system.)
Alignment
How the app matches level and timing between your speakers.
- Gain Estimation β how playback level is matched: Target Level Match (default-style, matches to the target), Minimum Summed Error (Least Squares), or Matched Amplitude (RMS).
- Delay Estimation β how the alignment delay is found: Minimum Phase Error (Least Squares) or Cross Correlation Peak (GCC-PHAT).
- Time-Align Tops β whether your left and right speakers are time-aligned to each other, or only the subwoofer is aligned to the tops. Default Subwoofer only. Choose All speakers if your left and right are at noticeably different distances.
- Phase Correction β after alignment, an extra pass fine-tunes the phase still left between each speaker and the reference, which can tighten the stereo image. Default Off: it makes calibration take noticeably longer, and it only helps where the remaining phase error is large enough to hear.
Leave Gain and Delay Estimation on their defaults unless you're comparing methods deliberately β they change how alignment is computed, not whether it happens.
Room Correction
How hard and how accurately the optimiser works.
- Max Generations β Overkill (80) / Best (50) / Balanced (30) / Fastest (15). More generations = more accurate, slower.
- Fitness Threshold β Best (10%) / Balanced (20%) / Fastest (30%). Lower = more accurate, slower.
- Confidence Weighting β which frequencies set the bar for a "trustworthy" measurement: 95% Strict (clearest bins only), 70% Balanced (the default), 50% Relaxed, 10% Permissive. Tighten to Strict if correction seems to chase measurement noise; only relax below Balanced if correction is leaving obvious problems untreated.
- Correction Filters β the maximum number of EQ filters the correction may use per channel (a number from 1 to 48; default 24). The device itself always caps it too β a channel never gets more filters than its hardware has slots for, so the default simply means "use every slot the device offers". Fewer filters give a simpler, easier-to-review correction and leave slots free for manual EQ, at the cost of finer detail.
- Boost Trust β how noisy a frequency's measurement can be before a boost stops counting fully: Off (clearest frequencies only), Slight (80%, default), Moderate (65%), Maximum (50%). Higher settings let the app boost into less-certain regions β use with care.
- Cut Trust β the same idea for cuts: Slight (80%), Moderate (65%), Standard (50%, default), Widest (30%). Cutting is safer than boosting, so this can be more relaxed.
- Boost Restraint β how strongly the optimiser is discouraged from boosting into deep, unfixable dips: Off, Gentle (recommended), Strong.
- Target Curve β the tonal balance to aim for (see Chapter 14 β Custom target curves).
The two knobs worth knowing
If correction sounds over-processed or "boosty", set Boost Restraint β Strong and/or Boost Trust β Off, then Calculate Again. If it's leaving obvious problems uncorrected, try Confidence Weighting β 70% Balanced β and re-measure first if the diagnostics blamed a weak measurement.
After changing any of these, go to the results and click Calculate Again to re-run with the new settings. Your saved tuning is preserved across app updates.
Display
These affect only the chart view, never the captured data or the correction:
- Smoothing β fractional-octave smoothing of the displayed curve (default 1/6 Oct).
- Coherence β hide low-confidence frequencies (default Low = show all).
- Graph Scale β vertical zoom: Full Scale, Β±24dB Norm, or Customβ¦ β set your own Β±dB window (Scale) and frequency range (From / To); an Edit button reopens the custom view while it's active.
The full range and default tables are in Appendix A β Reference tables.
