One listening chain, one independent reference
The listening path and condition measurement are separate. Changing restoration or sound shaping does not change the reference used to compare record sides.
Original remains time-aligned, omits restoration and tone but shares enabled convolution. Recognition, grading, noise acquisition and recording taps do not receive the output tone or room filter. See Sound Shaping.
Fixed NG2 and separate stages
DeClick feeds DeCrackle in one sequential NG2 chain, with independent manual and automatic sensitivities. Each manual sensitivity is an integer from 0–100%. Zero is a transparent, latency-preserving bypass for that stage; higher sensitivity can detect smaller defects but can also alter musical transients. Manual values are retained while Auto owns the effective slider.
The product uses fixed NG2. The enabled click families can be selected while stopped in Settings → Click Families.
A sensible first session
Mid is the current fresh-install default. Begin with both repair Auto controls, S150 and NR Auto, then compare Repaired, Original and the removed signals. Paranoid and Light reduce the built-in click/crackle sensitivity curves; High raises them. Change one control at a time and listen to attacks, percussion and quiet detail.
AutoNG observations
DeClick Auto and DeCrackle Auto use independent curves against the same rolling reference observation, approximately 3.6 seconds long and refreshed about every 1.8 seconds. Until enough fresh input is available, bounded fallback values are used. The stable cumulative media grade uses longer windows and may appear later.
When Auto is enabled, its slider shows the effective sensitivity and cannot be moved manually. Turning Auto off restores the saved manual value. Optional quiet-passage boost can strengthen click/crackle correction without changing the displayed grade or NR limits.
Current correction profiles
AutoNG sensitivity curves decrease from M to P. At a low reference-event rate the policy can look for small defects; at a busy rate it becomes more conservative. These numbers are not the former increasing “maximum correction” table. DeClick and DeCrackle percentages have separate scales and must not be compared as equal processing strength.
DeClick AutoNG sensitivity
| Grade | Paranoid | Light | Mid | High |
|---|---|---|---|---|
| M · Mint | 30% | 35% | 40% | 45% |
| NM · Near Mint | 29% | 34% | 39% | 44% |
| VG+ · Very Good Plus | 28% | 33% | 38% | 43% |
| VG · Very Good | 25% | 30% | 35% | 40% |
| G+ · Good Plus | 20% | 25% | 30% | 35% |
| G · Good | 15% | 20% | 25% | 30% |
| F · Fair | 12% | 17% | 22% | 27% |
| P · Poor | 10% | 15% | 20% | 25% |
DeCrackle AutoNG sensitivity
| Grade | Paranoid | Light | Mid | High |
|---|---|---|---|---|
| M · Mint | 35% | 40% | 45% | 50% |
| NM · Near Mint | 30% | 35% | 40% | 45% |
| VG+ · Very Good Plus | 25% | 30% | 35% | 40% |
| VG · Very Good | 20% | 25% | 30% | 35% |
| G+ · Good Plus | 18% | 23% | 28% | 33% |
| G · Good | 16% | 21% | 26% | 31% |
| F · Fair | 14% | 19% | 24% | 29% |
| P · Poor | 12% | 17% | 22% | 27% |
Custom edits the two AutoNG curves in Settings → Grading. The native grade calibration and shared NR limits remain read-only. Restore Defaults returns to Mid.
NRNG noise reduction
S150 is the conservative profile; S200 is stronger. NR follows both click/crackle stages and low-frequency conditioning. It keeps a prepared, aligned branch while bypassed, so switching the mix does not create a comparison timing jump.
- Manual maximum
- 1–12 dB, preserved while NR Auto is active.
- NR Auto
- Uses the shared grade-dependent maximum below, with the Mint row until a grade is available. These are ceilings, not a promise of uniform attenuation across every frequency.
- Relearn noise
- Starts local acquisition without rebuilding the transport. A newly qualified side also starts acquisition.
Local profile acquisition and fallback
Live observes a four-second acquisition interval, ignores its initial segment and chooses a median-level one-second candidate from qualifying 250 ms windows. The maximum admissible capture level defaults to −55 dBFS. Use a lead-in or a gap with surface noise but no music; raising the threshold can capture quiet programme material.
While a local profile is being acquired, a Generic fallback is available at 44.1, 48, 88.2, 96 and 192 kHz. Relearn can retain an existing local profile until a replacement is ready. If no usable renderer is prepared, the branch remains an aligned bypass; waiting does not always mean NR is silent or inactive. Check the displayed profile state rather than assuming a first-quiet-second capture.
Use input calibration to establish system silence and surface noise before changing the acquisition threshold.
| Grade | NR Auto maximum — all built-in profiles |
|---|---|
| M · Mint | 2 dB |
| NM · Near Mint | 4 dB |
| VG+ · Very Good Plus | 5 dB |
| VG · Very Good | 6 dB |
| G+ · Good Plus | 8 dB |
| G · Good | 10 dB |
| F · Fair | 11 dB |
| P · Poor | 12 dB |
Actual reduction depends on the frequency-dependent gain, music protection and profile confidence. Activity exposes profile state, gain, masking and other diagnostics; they do not replace listening.
Understanding the media grade
The stable grade is the median of complete, non-overlapping reference windows of about ten seconds since the last reset or qualified side change. Thresholds are calibrated for the native session rate, not one universal detection-rate table.
| Grade | 44.1 kHz | 48 kHz | 88.2 kHz | 96 kHz | 176.4 kHz | 192 kHz |
|---|---|---|---|---|---|---|
| M · Mint | 0.55 | 0.75 | 2.75 | 3.75 | 3.925 | 4.025 |
| NM · Near Mint | 1.05 | 1.5 | 9 | 11.9 | 10.175 | 9.725 |
| VG+ · Very Good Plus | 2.05 | 3.025 | 14.35 | 18.275 | 15.025 | 14.275 |
| VG · Very Good | 3.325 | 4.575 | 16.875 | 21.4 | 23.45 | 22.5 |
| G+ · Good Plus | 4.325 | 5.3 | 20.275 | 24.375 | 33.775 | 32.125 |
| G · Good | 8.7 | 9.75 | 28.05 | 32.05 | 44.05 | 42.8 |
| F · Fair | 16.3 | 17.65 | 38.75 | 44.35 | 56.15 | 56 |
| P · Poor | Above the corresponding F boundary | |||||
Other accepted transport rates can process audio but report Uncalibrated grading rather than borrowing the 96 kHz thresholds. Calibration, silence and incomplete windows can temporarily exclude audio from the cumulative measurement.
Reset grading
Reset when beginning a different record or intentionally starting a fresh measurement. Audio continues and the repair history remains separate. Auto uses bounded fallback values while fresh reference observations are collected.
Automatic side detection
Sustained input below the configured silence threshold pauses cumulative grading and recognition, clears Now Playing and retains the previous grade. Two seconds of sustained returning input qualify the next side before grading and recognition restart and fresh noise acquisition begins. The short automatic restoration observation is separate from cumulative grading.