Tools for the mix.

1. 0.435 dBHow far a +12 dB bell at 20 kHz with Q 1 departs from the analogue prototype's curve at its worst point at 44.1 kHz in MATCHED (limit 0.5 dB), against 8.159 dB in MINIMUM. At 48, 96 and 192 kHz: 0.445, 0.484 and 0.239 dB in MATCHED, 6.226, 3.295 and 0.951 dB in MINIMUM.BEVEL
2. 0.012 dBHow far a quiet tone, 20 dB under the threshold in the same band as a loud one, moved while a SPECTRAL band turned the loud one down (limit 0.02 dB); the loud tone moved by the curve within 0.5 dB, the limit. A time-domain band would have moved both.BEVEL
3. Bit for bitAt rest, and in bypass, the output is the input delayed by the reported latency, in every voicing and with every look-ahead.CLAMP
4. 240 samplesHow far ahead of a burst reaching the output the reduction has begun with 5 ms look-ahead at 48 kHz: the first 2 ms leave 16.6 dB under the input, against 3.8 dB with no look-ahead.CLAMP
5. 111.4 dBThe least any non-DELTA setting of 57 came out below the output when the new engine was compared with version 1's on the same signals (up to 125.1 dB; the bound is 100 dB). The DELTA settings came out 94.3 to 102.3 dB below the DELTA output, and version 1's DELTA left a rounding trace 138.5 to 139.6 dB under the programme at DEPTH 0, where the new one is silent.PLANE
6. Bit for bitAt DEPTH 0 the output is the input delayed by the reported latency, in each of 48 combinations of settings. Version 1 was 138.0 to 139.5 dB under it. Going from DEPTH 12 to 0, the output was exactly the delayed input after 0.39, 0.41 and 0.44 s in FAST, NORMAL and FINE at RELEASE 120 ms.PLANE
7. 0 of 240Runs in which the copy of LATHE's gate differed from the original in output, trace, meter or history: GATE and EXPAND, every look-ahead, LISTEN on and off, nine sets of settings, four sample rates, noise, bursts and the voice, with the settings moved three times during a run.LATCH
8. Bit for bitWith RANGE 0 the output is the input delayed by the look-ahead in 240 runs: mono and stereo, GATE and EXPAND, five look-aheads, six sample rates, every other control at its default or its hardest. A denormal input number comes out as itself or 0.LATCH
9. 0 of 432Runs in which the copy of LATHE's de-esser module differed from the original: WIDE on and off, FREQUENCY 3 to 12 kHz, Q 0.5 to 8, RANGE 0 to 18 dB, THRESHOLD absolute and relative, ATTACK, RELEASE, LISTEN on and off, four sample rates, noise, sibilant bursts and the voice, with the setting moved three times in a run.SCRAPER
10. Bit for bitWith RANGE 0 the output is the input, with WIDE on or off and a delay of 0, in the engine's 144 runs and in the processor at six sample rates in both layouts. A denormal input number comes out as itself or 0.SCRAPER
11. 0, 25 or 29The delay in samples the processor told the host for OVERSAMPLING OFF, 2X and 4X, at 44.1, 48, 88.2, 96, 176.4 and 192 kHz: the same number at every rate. An impulse at rest came out exactly that late.CHAR
12. Bit for bitWith DRIVE 0, TILT 0 and OUTPUT 0 dB the output is the input delayed by the reported delay in 54 runs and at six sample rates, in both layouts. A denormal input number comes out as itself or 0.CHAR
13. 0 samplesThe delay the processor told the host at every setting tried, at 44.1, 48, 88.2, 96, 176.4 and 192 kHz, and read back from the Audio Unit at 48 kHz. BYPASS does not change it.TAP
14. Bit for bitAt MIX 0 and OUTPUT 0 dB the output is the input in 36 runs and at six sample rates in three layouts, with FEEDBACK 110, FREEZE, DUCK, DRIVE, DIFFUSION, MOD DEPTH, WIDTH 200 and R TO L set. Through the installed Audio Unit 18 of 18 runs were equal. A denormal input number comes out as 0 and a minus zero as plus zero.TAP
15. 0 samplesThe delay the processor told the host at every setting tried, at 44.1, 48, 88.2, 96, 176.4 and 192 kHz. BYPASS does not change it.CASK
16. Bit for bitAt MIX 0 and OUTPUT 0 dB the output is the input in 36 runs (six modes, three layouts, with FREEZE, GATE, DUCK and DECAY 30 s on and off) and at six sample rates. A denormal input number comes out as 0 and a minus zero as plus zero.CASK
17. 241 samplesThe delay told to the host at the defaults at 48 kHz (229, 241, 362, 385, 626 and 673 at 44.1, 48, 88.2, 96, 176.4 and 192 kHz): the told delay, the sum of the modules' own delays and the place of an impulse are one number in 72 structures and rates. The most is 1589 samples at 48 kHz and the least 14, found by trying all 10,080 combinations of the controls that change it.CHEST
18. Bit for bitAt rest, with all eight off, the default order and both levels at 0 dB, the output is the input delayed by the told delay at every block size from 1 to 4096 samples, at six sample rates and in three layouts. Each module alone equals its engine, bit for bit, in 8 modules, two settings and two orders, and all eight on in six orders equal the engines chained by hand.CHEST
19. 11.82 dBHow far the band ESS reads fell, input to output, for a sibilant burst 10 dB over the voice (median of 10 bursts); 3.36 dB of it by ESS itself (limit: at least 3 dB). The vowel beside it moved 0.023 dB with ESS alone (limit 0.3 dB).LATHE
20. 11.94 dBHow far the band under 200 Hz fell, input to output, for a plosive 15 dB over the voice (median of 6); 3.12 dB of it by POP itself (limit: at least 3 dB).LATHE
21. -79 dBAlias of a hard-driven 9.5 kHz tone at 10 kHz with 2x oversampling, against -43 dB with it off.VENEER
22. 25, 29Samples of delay at 2x and 4x, exact at every sample rate; the strip at rest is that delay and nothing else.VENEER
23. Under -50 dBA steady tone changes by less than this with ATTACK and SUSTAIN both at +100. That is the limit the test sets; the change was inside it. Steady sound is left alone.CHISEL
24. 0.33 %Worst pitch error following a voice that glides from 150 to 250 Hz.WEDGE
25. -139 dBReconstruction error when nothing is carved: the track comes back as it went in.WEDGE
26. 0.04 sampleThe same for HOLD, which needs no reference playing: 0.04 at 50 ppm, 0.02 at 20 ppm.SQUARE
27. 0.03 LUThe reference, heard section by section, comes within 0.03 LU of the mix's loudness.GAUGE
28. Exactp-values are the exact binomial: 10 of 10 is 0.0010 and 12 of 16 is 0.0384 (the panel rounds to three decimals).LEVEL

Measured on synthetic signals by each plug-in’s own tests, against the shipped build. They are measurements, not listening tests: they say what the code does to a test signal, and nothing about how it sounds on your session.