측정을 믿어도 되는 이유Why the numbers can be trusted
측정기의 신뢰성은 화면이 예뻐서 생기지 않습니다. CAA는 정답을 이미 아는 신호를 실제 제품과 똑같은 코드에 넣고 답이 맞는지 확인합니다 — 정확히 5.000 ms 지연된 스윕, 잔향시간이 정확히 500 ms인 감쇠, 2.46%와 0.49%의 고조파를 갖도록 만든 왜곡. 아래 표는 그 결과이고, 숫자는 손으로 적은 것이 아니라 검증 프로그램이 출력한 그대로입니다.A measurement tool is not trustworthy because its screens look good. CAA is fedsignals whose answer is already known — a sweep delayed by exactly 5.000 ms, a decay whose reverberation time is exactly 500 ms, a distortion built to have harmonics of 2.46% and 0.49% — through the same code the product ships. The tables below are what came back, printed by the test program rather than typed out.
241개 항목 · 실패 0개241 checks · 0 failing
빌드할 때마다 이 검증을 다시 돌립니다. 하나라도 틀리면 그 빌드는 나가지 않습니다.The suite runs on every build. If one check fails, that build does not ship.
REW와 나란히 측정한 결과Measured beside REW
같은 임펄스 응답 파일을 CAA와 REW V5.40에 각각 넣고 읽은 값입니다. 이론값은 필터 계수와 감쇠식에서 직접 계산한 정답입니다. 크로스오버 응답은 53 dB 아래까지 0.02 dB 이내, 잔향시간은 0.8% 이내로 일치합니다.The same impulse-response files, measured in CAA and in REW V5.40. The theory column is the exact answer from the filter coefficients and the decay law. The crossover response agrees to within 0.02 dB even 53 dB down, and the reverberation times to within 0.8%.
| 항목Measurement | 이론값Theory | CAA | REW | 차이CAA − REW |
|---|
| LR4 2.5 kHz, 1 kHz | -0.213 | -0.213 | -0.213 | -0.000 |
| LR4 2.5 kHz, at fc | -6.021 | -6.024 | -6.019 | -0.005 |
| LR4 2.5 kHz, 5 kHz | -25.516 | -25.512 | -25.504 | -0.008 |
| LR4 2.5 kHz, 10 kHz | -53.397 | -53.400 | -53.377 | -0.022 |
| T20, 500 ms decay | 500.000 | 502.728 | 498.850 | +3.878 |
| T30, 500 ms decay | 500.000 | 499.931 | 499.172 | +0.759 |
| C50 | 4.744 | 4.810 | 4.742 | +0.068 |
| C80 | 9.096 | 9.176 | 8.945 | +0.231 |
단위는 항목별로 dB 또는 ms입니다. REW는 REST API로 같은 파일을 불러 측정했고, 재현에 쓴 스크립트는 rew_compare.py에 들어 있습니다.Units are dB or ms per row. REW measured the same files through its REST API; the script that does it is rew_compare.py.
검증 항목 전체Every check
레벨 스케일Level scale (WebAnalyser) · 2
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| sine, summed bin power vs -20 dBFS RMS | -20 | -20.0000 | 0.1 | PASS |
| white noise, summed bin power vs RMS | -16.8236 | -16.8034 | 0.1 | PASS |
밴드 표와 A 가중Band tables (ISO 266 / IEC 61672) · 21
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| 31-band table size | 31 | 31.0000 | 0 | PASS |
| band 0 = 20 Hz | 20 | 20.0000 | 0 | PASS |
| band 17 = 1 kHz | 1000 | 1000.0000 | 0 | PASS |
| band 30 = 20 kHz | 20000 | 20000.0000 | 0 | PASS |
| A-weighting @ 31 Hz (class 1) | -39.49 | -39.5000 | 1.5 | PASS |
| A-weighting @ 31 Hz (tight) | -39.49 | -39.5000 | 0.12 | PASS |
| A-weighting @ 63 Hz (class 1) | -26.2551 | -26.2000 | 1 | PASS |
| A-weighting @ 63 Hz (tight) | -26.2551 | -26.2000 | 0.12 | PASS |
| A-weighting @ 125 Hz (class 1) | -16.219 | -16.2000 | 1 | PASS |
| A-weighting @ 125 Hz (tight) | -16.219 | -16.2000 | 0.12 | PASS |
| A-weighting @ 250 Hz (class 1) | -8.7002 | -8.7000 | 1 | PASS |
| A-weighting @ 250 Hz (tight) | -8.7002 | -8.7000 | 0.12 | PASS |
| A-weighting @ 500 Hz (class 1) | -3.2698 | -3.2000 | 1 | PASS |
| A-weighting @ 500 Hz (tight) | -3.2698 | -3.2000 | 0.12 | PASS |
| A-weighting @ 1000 Hz (class 1) | -0.0002 | 0.0000 | 0.7 | PASS |
| A-weighting @ 1000 Hz (tight) | -0.0002 | 0.0000 | 0.12 | PASS |
| A-weighting @ 2000 Hz (class 1) | 1.2704 | 1.2000 | 1 | PASS |
| A-weighting @ 2000 Hz (tight) | 1.2704 | 1.2000 | 0.12 | PASS |
| A-weighting @ 4000 Hz (class 1) | 1.2591 | 1.0000 | 1 | PASS |
| A-weighting @ 8000 Hz (class 1) | -0.447 | -1.1000 | 1.5 | PASS |
| A-weighting @ 12500 Hz (class 1) | -4.4105 | -4.3000 | 2 | PASS |
스윕 디컨볼루션 · 도달 시간Sweep deconvolution -- arrival time · 5
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| 0 samples -> 0.000 ms | 0 | 0.0000 | 0.005 | PASS |
| 240 samples -> 5.000 ms | 5 | 5.0000 | 0.005 | PASS |
| 240.5 samples (sub-sample) | 5.0104 | 5.0104 | 0.02 | PASS |
| 961.3 samples (sub-sample) | 20.0247 | 20.0271 | 0.02 | PASS |
| inverted input -> negative impulse peak | 1 | 1.0000 | 0 | PASS |
크로스오버 · 설계와 실측Crossover -- designed and measured back · 16
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| LR4 low-pass at fc | -6.0206 | -6.0200 | 0.1 | PASS |
| LR4 high-pass at fc | -6.0206 | -6.0200 | 0.1 | PASS |
| Butterworth 24 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 6 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 12 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 18 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 24 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 30 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 36 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 42 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| Butterworth 48 dB/oct at fc | -3.0103 | -3.0100 | 0.1 | PASS |
| measured vs designed @ 200 Hz | -0.0004 | -0.0003 | 0.3 | PASS |
| measured vs designed @ 1000 Hz | -0.2135 | -0.2131 | 0.3 | PASS |
| measured vs designed @ 2500 Hz | -6.0241 | -6.0206 | 0.3 | PASS |
| measured vs designed @ 5000 Hz | -25.5121 | -25.5155 | 0.3 | PASS |
| measured vs designed @ 10000 Hz | -53.3995 | -53.3971 | 0.3 | PASS |
고조파 왜율Harmonic distortion (Farina packets) · 4
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| distortion analysis valid | 1 | 1.0000 | 0 | PASS |
| H2 @ 1 kHz | 2.463 | 2.4631 | 0.25 | PASS |
| H3 @ 1 kHz | 0.4926 | 0.4926 | 0.15 | PASS |
| THD @ 1 kHz | 2.5118 | 2.5118 | 0.3 | PASS |
실내 음향 (ISO 3382-1)Room acoustics (ISO 3382-1) · 7
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| T20 of an exact 500 ms decay | 502.7281 | 500.0000 | 15 | PASS |
| T30 of an exact 500 ms decay | 499.9307 | 500.0000 | 15 | PASS |
| T30 regression fit r^2 | 1 | >=0.9900 | 0 | PASS |
| C50 vs closed form | 4.81 | 4.7437 | 0.5 | PASS |
| C80 vs closed form | 9.1764 | 9.0956 | 0.5 | PASS |
| octave-band T30 @ 125 Hz (700 ms) | 700.0123 | 700.0000 | 60 | PASS |
| octave-band T30 @ 4 kHz (250 ms) | 250.0031 | 250.0000 | 40 | PASS |
NC 등급NC rating (noise criterion) · 3
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| spectrum on the NC-30 line | 30 | 30.0000 | 0 | PASS |
| 6 dB quieter rates better than NC-30 | 5 | >=1.0000 | 0 | PASS |
| 6 dB louder rates worse than NC-30 | 10 | >=1.0000 | 0 | PASS |
2채널 FFT 전달함수Dual-FFT transfer function (IRMeasurement) · 10
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| result produced (unfiltered) | 1 | 1.0000 | 0 | PASS |
| delay of an exact 8.000 ms path | 8 | 8.0000 | 0.05 | PASS |
| result produced (filtered) | 1 | 1.0000 | 0 | PASS |
| H1 magnitude vs design @ 500 Hz | 0.0095 | -0.0134 | 0.6 | PASS |
| H1 magnitude vs design @ 1000 Hz | -0.183 | -0.2131 | 0.6 | PASS |
| H1 magnitude vs design @ 2500 Hz | -6.0346 | -6.0206 | 0.6 | PASS |
| H1 magnitude vs design @ 5000 Hz | -25.4711 | -25.5155 | 0.6 | PASS |
| delay added by the LR4 itself | 0.1875 | 0.1800 | 0.03 | PASS |
| coherence in the pass band (500 Hz) | 0.9999 | >=0.9500 | 0 | PASS |
| coherence in the stop band (12 kHz) is low | 1 | 1.0000 | 0 | PASS |
최소위상 · 초과위상Minimum / excess phase · 5
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| decomposition valid | 1 | 1.0000 | 0 | PASS |
| LR4 alone: excess delay is zero | 0 | 0.0000 | 0.02 | PASS |
| pure 2.000 ms delay | 2 | 2.0000 | 0.02 | PASS |
| LR4 + 2.000 ms: delay separated from filter | 2 | 2.0000 | 0.03 | PASS |
| minimum phase reproduces the LR4's phase | 0.0898 | 0.0000 | 2 | PASS |
채널별 목표와 점수Per-channel target and score · 26
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| target curve built | 240 | 240.0000 | 0 | PASS |
| perfect response scores 100 | 99.9145 | 100.0000 | 0.5 | PASS |
| tweeter is judged from ~3.1 kHz up | 3149.8025 | 3149.7500 | 5 | PASS |
| perfect response needs no fixes | 0 | 0.0000 | 0 | PASS |
| a uniformly quiet channel loses half its score | 49.9853 | 50.0000 | 1 | PASS |
| score follows RMS deviation | 82.51 | 82.5100 | 0.1 | PASS |
| a real error costs points | 17.49 | >=5.0000 | 0 | PASS |
| worst band identified | 3150 | 3150.0000 | 0 | PASS |
| correction has the right sign and size | -3.5 | -4.0000 | 1.5 | PASS |
| correction never overshoots the error | 0.5 | >=0.0000 | 0 | PASS |
| a spike narrower than the band is under-corrected | 0.5 | >=0.0000 | 0 | PASS |
| sub is judged below its corner | 63.496 | 63.4971 | 1 | PASS |
| out-of-band content does not change the score | 98.3639 | 98.3639 | 0.01 | PASS |
| band read back from a 2.5 kHz high-pass | 2181.5535 | 2500.0000 | 500 | PASS |
| ...and it reports no low-pass | 0 | 0.0000 | 0 | PASS |
| band-pass low corner | 400.7899 | 400.0000 | 80 | PASS |
| band-pass high corner | 2278.4199 | 2500.0000 | 500 | PASS |
| sub low-pass corner | 83.8825 | 80.0000 | 16 | PASS |
| ...and no high-pass on a sub | 0 | 0.0000 | 0 | PASS |
| a full-range driver reports no corners | 0 | 0.0000 | 0 | PASS |
| summed target built | 240 | 240.0000 | 0 | PASS |
| the gap between two drivers is in the target | 24.1952 | >=10.0000 | 0 | PASS |
| a single spanning band would have missed it | 29.1664 | >=10.0000 | 0 | PASS |
| one driver sums to itself | 0 | 0.0000 | 0.01 | PASS |
| a tweeter with no crossover set still has a band | 2500 | 2500.0000 | 0 | PASS |
| a sub with no crossover set still has a band | 80 | 80.0000 | 0 | PASS |
AUTO 모드 순서AUTO mode sequence · 14
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| the guided sequence exists | 19 | 19.0000 | 0 | PASS |
| the crossover is set before any sweep | 7 | >=1.0000 | 0 | PASS |
| time alignment comes before any EQ | 4 | >=1.0000 | 0 | PASS |
| per-channel tone before the global EQ | 5 | >=1.0000 | 0 | PASS |
| per-channel tone after the crossover is set | 6 | >=1.0000 | 0 | PASS |
| per-channel tone before time alignment | 1 | >=1.0000 | 0 | PASS |
| the channel-target step needs no loop-back | 1 | 1.0000 | 0 | PASS |
| the step opens the Channel Target screen | 1 | 1.0000 | 0 | PASS |
| a car that measures right passes | 1 | 1.0000 | 0 | PASS |
| every channel is reported | 2 | 2.0000 | 0 | PASS |
| a badly-off channel blocks the step | 1 | 1.0000 | 0 | PASS |
| the failing channel is named first | 1 | 1.0000 | 0 | PASS |
| with a band and an amount to change | 1 | 1.0000 | 0 | PASS |
| no measurement is a warning, not a pass | 1 | 1.0000 | 0 | PASS |
위상 · 지연 램프 · 시간 기준Phase, delay ramp and timing reference · 7
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| bin spacing of a 16384-point FFT | 2.9297 | 2.9297 | 0 | PASS |
| the last bin is Nyquist | 24000 | 24000.0000 | 0 | PASS |
| a tiny transform too | 24000 | 24000.0000 | 0 | PASS |
| a 5 ms delay wraps every 200 Hz | 3 | 2.0000 | 1 | PASS |
| wraps crowd together toward the top | 22 | >=15.0000 | 0 | PASS |
| removing the delay leaves flat phase | 0 | 0.0000 | 1 | PASS |
| half a sample of error is visible | 59.9963 | >=25.0000 | 0 | PASS |
C 가중 (IEC 61672-1)C-weighting (IEC 61672-1) · 10
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| C-weighting @ 31 Hz | -2.9875 | -3.0000 | 1.5 | PASS |
| C-weighting @ 63 Hz | -0.8508 | -0.8000 | 1 | PASS |
| C-weighting @ 125 Hz | -0.2011 | -0.2000 | 1 | PASS |
| C-weighting @ 250 Hz | -0.0241 | 0.0000 | 1 | PASS |
| C-weighting @ 500 Hz | 0.0154 | 0.0000 | 1 | PASS |
| C-weighting @ 1000 Hz | 0 | 0.0000 | 0.7 | PASS |
| C-weighting @ 2000 Hz | -0.1038 | -0.2000 | 1 | PASS |
| C-weighting @ 4000 Hz | -0.5346 | -0.8000 | 1 | PASS |
| C-weighting @ 8000 Hz | -2.3518 | -3.0000 | 1.5 | PASS |
| Z-weighting passes the signal untouched | 0 | 0.0000 | 0 | PASS |
마이크 보정 파일Microphone calibration file · 8
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| calibration file loads | 1 | 1.0000 | 0 | PASS |
| no correction where the mic is flat | 0 | 0.0000 | 0.01 | PASS |
| a +6 dB mic error corrects by -6 dB | -6 | -6.0000 | 0.01 | PASS |
| interpolated between points | -2.9998 | -3.0000 | 0.3 | PASS |
| flat outside the file's range | -6 | -6.0000 | 0.01 | PASS |
| spectrum corrected at 10 kHz | -5.9997 | -6.0000 | 0.05 | PASS |
| spectrum untouched at 1 kHz | 0 | 0.0000 | 0.05 | PASS |
| cleared calibration corrects nothing | 0 | 0.0000 | 0 | PASS |
샘플레이트별 필터 안정성Biquad stability across sample rates · 6
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| worst pole margin below the unit circle | 0.0009 | >=0.0000 | 0 | PASS |
| 20 kHz section on a 32 kHz device stays bounded | 0.9565 | 1.0000 | 2 | PASS |
| ...and its impulse response decays away | 0 | 0.0000 | 0 | PASS |
| 48 kHz / 1 kHz untouched | 1000 | 1000.0000 | 0.01 | PASS |
| 48 kHz / 20 kHz untouched | 20000.0009 | 20000.0000 | 0.01 | PASS |
| 32 kHz / 20 kHz clamped below Nyquist | 15680.0011 | 15680.0000 | 1 | PASS |
임피던스 · 유닛 모델의 공진Impedance -- resonance from a driver model · 7
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| a resonance is found | 1 | 1.0000 | 0 | PASS |
| Fs recovered from the curve | 44.9552 | 45.0000 | 1.5 | PASS |
| the plain maximum really is the inductive rise | 20000 | >=10000.0000 | 0 | PASS |
| Z_min is close to Re | 6.0708 | 6.0000 | 0.6 | PASS |
| a tweeter's Fs too | 894.9113 | 900.0000 | 25 | PASS |
| Fs survives 4% ripple on the curve | 44.8776 | 45.0000 | 2.5 | PASS |
| a resistor reports no resonance | 1 | 1.0000 | 0 | PASS |
스윕 품질 · 클리핑과 SNRSweep quality -- clipping and SNR · 12
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| 0.95 peak flagged as input overload | 1 | 1.0000 | 0 | PASS |
| 0.50 peak is not | 0 | 0.0000 | 0 | PASS |
| pre-roll does not shift arrival (240 smp -> 5.000 ms) | 5 | 5.0000 | 0.005 | PASS |
| SNR of a take with noise 40 dB under the sweep | 82.3215 | 82.1986 | 1.5 | PASS |
| reverb tail is not counted as noise | 120 | >=50.0000 | 0 | PASS |
| hum: 1 kHz band clearly above the 63 Hz band | 63.0209 | >=20.0000 | 0 | PASS |
| hum: 1 kHz band itself is clean | 74.7622 | >=40.0000 | 0 | PASS |
| averaging 2 takes gains about 3 dB SNR | 3.1302 | 3.0000 | 1 | PASS |
| clipped take: lower the output | -1 | -1.0000 | 0 | PASS |
| quiet, noisy take: raise it | 1 | 1.0000 | 0 | PASS |
| raise by enough to reach -12 dBFS | 20 | 20.0000 | 0.01 | PASS |
| a good take: leave it | 0 | 0.0000 | 0 | PASS |
AI 어시스트 진단 규칙Tuning assistant (offline rules) · 25
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| no data -> no items | 0 | 0.0000 | 0 | PASS |
| inverted first peak -> polarity rule | 1 | 1.0000 | 0 | PASS |
| clipped take -> overload rule | 1 | 1.0000 | 0 | PASS |
| narrow -8 dB notch -> 'don't EQ' rule | 1 | 1.0000 | 0 | PASS |
| ...and no boost suggestion for it | 0 | 0.0000 | 0 | PASS |
| symptom matched (쏠려요) | 1 | 1.0000 | 0 | PASS |
| spelling variant -> same symptom | 1 | 1.0000 | 0 | PASS |
| unrelated question -> honest no-answer | 1 | 1.0000 | 0 | PASS |
| ...offering 3 nearby symptoms | 3 | 3.0000 | 0 | PASS |
| tweeter HPF 800 Hz / 12 dB -> warned | 1 | 1.0000 | 0 | PASS |
| tweeter HPF 2.5 kHz / 24 dB -> quiet | 0 | 0.0000 | 0 | PASS |
| sub LPF 160 Hz -> warned | 1 | 1.0000 | 0 | PASS |
| mid ends 2 kHz, tweeter starts 4 kHz -> gap | 1 | 1.0000 | 0 | PASS |
| mid to 3.5 kHz, tweeter from 1.8 kHz -> overlap | 1 | 1.0000 | 0 | PASS |
| matched 2.5 kHz handover -> neither | 0 | 0.0000 | 0 | PASS |
| +8 dB EQ boost -> warned | 1 | 1.0000 | 0 | PASS |
| band at -12 dB rail -> flagged | 1 | 1.0000 | 0 | PASS |
| correction added a 7 dB bump -> 'worse after' | 1 | 1.0000 | 0 | PASS |
| bump removed by correction -> 'better after' | 1 | 1.0000 | 0 | PASS |
| R 6 dB louder at 1 kHz only -> L/R colour rule | 1 | 1.0000 | 0 | PASS |
| identical L/R -> quiet | 0 | 0.0000 | 0 | PASS |
| '트위터 터질까 봐요' -> tweeter_safe | 1 | 1.0000 | 0 | PASS |
| '보컬이 멀게 들려요' -> vocal_far | 1 | 1.0000 | 0 | PASS |
| '저음이 트렁크에서 들려요' -> sub_where | 1 | 1.0000 | 0 | PASS |
| '보정했더니 더 이상해요' -> after_worse_q | 1 | 1.0000 | 0 | PASS |
EQ 추천 = 실제 적용 필터EQ fit -- recommendation equals the applied filter · 16
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| +3 dB band reads +3 dB at its centre | 3 | 3.0000 | 0.01 | PASS |
| 63 Hz peak +5 -> band gain | -4 | -4.3497 | 0.5 | PASS |
| 1 kHz peak +4 -> band gain | -2.9 | -3.3497 | 0.5 | PASS |
| 8 kHz dip -2 -> band gain | 3 | 2.6503 | 0.5 | PASS |
| residual after fit | 0.3858 | 0.0000 | 0.5 | PASS |
| re-measure round changes nothing | 0 | 0.0000 | 0.3 | PASS |
| boost capped at +3 dB (3 - max) | 0 | >=-0.0010 | 0 | PASS |
| cap reported | 1 | 1.0000 | 0 | PASS |
| narrow dip reported | 1 | 1.0000 | 0 | PASS |
| dip centre | 1974.0299 | 2000.0000 | 100 | PASS |
| no boost into the dip (-gain @2k) | 0 | >=0.0000 | 0 | PASS |
| tweeter passband low edge | 1810.1934 | 1900.0000 | 400 | PASS |
| tweeter: nothing below 1.6 kHz | 0 | 0.0000 | 0 | PASS |
| tweeter: no boost at/below its corner | 0 | 0.0000 | 0 | PASS |
| tweeter: 6.3 kHz bump taken out | -4.3991 | -4.0000 | 1 | PASS |
| band count respected | 1 | 1.0000 | 0 | PASS |
세션 파일 · 형식 2 왕복, 형식 1 변환Session file -- format 2 round trip, format 1 migration · 15
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| format 2 loads | 1 | 1.0000 | 0 | PASS |
| format recorded | 2 | 2.0000 | 0 | PASS |
| magnitude bit-exact (bin 1023) | 0 | 0.0000 | 0 | PASS |
| before-take kept separately | 3 | 3.0000 | 0 | PASS |
| Live EQ band survives | -3.4 | -3.4000 | 0 | PASS |
| EQ on flag survives | 1 | 1.0000 | 0 | PASS |
| recommend round survives | 2 | 2.0000 | 0 | PASS |
| Korean title survives | 1 | 1.0000 | 0 | PASS |
| crossover corner survives | 63.5 | 63.5000 | 0 | PASS |
| crossover slope + type survive | 1 | 1.0000 | 0 | PASS |
| crossover PEQ survives | -2.5 | -2.5000 | 0 | PASS |
| format 1 still loads | 1 | 1.0000 | 0 | PASS |
| format 1: channel delay read | 3.5 | 3.5000 | 0 | PASS |
| format 1: no before-takes | 0 | 0.0000 | 0 | PASS |
| format 1: no corrections | 0 | 0.0000 | 0 | PASS |
AUTO · 시작 점검, 재측정, 진행 저장AUTO -- start checklist, re-measure, saved progress · 8
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| first step is the start checklist | 1 | 1.0000 | 0 | PASS |
| an unprepared car: warned, not blocked | 1 | 1.0000 | 0 | PASS |
| checklist lists six items | 6 | 6.0000 | 0 | PASS |
| a prepared car passes | 1 | 1.0000 | 0 | PASS |
| re-measure recognised (residual wording) | 1 | 1.0000 | 0 | PASS |
| converged re-measure passes | 1 | 1.0000 | 0 | PASS |
| AUTO step survives the session file | 7 | 7.0000 | 0 | PASS |
| step verdicts kept by step id | 1 | 1.0000 | 0 | PASS |
라이브 EQ · 들리는 소리 = 화면 곡선Live EQ -- what plays equals what is shown · 6
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| 48 kHz: heard vs shown, worst point | 0.0041 | 0.0000 | 0.05 | PASS |
| 44.1 kHz: heard vs shown, worst point | 0.0199 | 0.0000 | 0.05 | PASS |
| curve set before the device opened | 0.0199 | 0.0000 | 0.05 | PASS |
| device changed 48 -> 44.1 kHz | 0.0199 | 0.0000 | 0.05 | PASS |
| flat curve passes samples unchanged | 1 | 1.0000 | 0 | PASS |
| all bands at +/-12 dB stay stable | 1 | 1.0000 | 0 | PASS |
Target score -- the system's shape, not the signal'sTarget score -- the system's shape, not the signal's · 4
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| perfect system, pink noise per bin | 99.6562 | 100.0000 | 0.5 | PASS |
| perfect system, mic minus source | 99.7679 | 100.0000 | 0.5 | PASS |
| level does not change the score | -0.2321 | 0.0000 | 0.5 | PASS |
| 1 kHz dip costs points, not the whole score | 92.3397 | 92.0000 | 6 | PASS |
Room acoustics -- impulse-to-noise (when T20/T30 can be trusted)Room acoustics -- impulse-to-noise (when T20/T30 can be trusted) · 4
| 항목check | 측정measured | 이론expected | 허용tol | |
|---|
| noise -30 dB -> INR read back | 29.7406 | 29.7047 | 1 | PASS |
| noise -50 dB -> INR read back | 49.7002 | 49.7047 | 1 | PASS |
| INR 30 dB: T20 flagged | 1 | 1.0000 | 0 | PASS |
| INR 50 dB: T20 and T30 trusted | 1 | 1.0000 | 0 | PASS |
직접 확인해 보기Check it yourself
위의 두 표는 계산을 검증한 것입니다. 마이크와 차량까지 포함한 진짜 측정은 각자의 장비로만 확인할 수 있으므로, 그 절차와 도구를 함께 넣어 두었습니다. 같은 마이크를 같은 자리에 고정한 채 한 채널을 CAA로 한 번, REW로 한 번 재고field_compare.py를 돌리면 두 곡선과 차이 곡선이 나옵니다. 절차와 결과 읽는 법은 FIELD-COMPARE.md에 있습니다.The two tables above check the arithmetic. Whether a real measurement through your microphone and your car agrees can only be established on your own rig, so the procedure and the tool ship with the app. Fix one microphone in one place, measure one channel once in CAA and once in REW, and run field_compare.py: it draws both curves and the difference between them. The method, and how to read the result, is inFIELD-COMPARE.md.
아래 그림은 그 대조 도구가 스스로 만들어 내는 오차입니다 —같은 임펄스 응답 파일 하나를 CAA 세션과 REW에 각각 넣고 비교한 것으로, 평균 0.007 dB, 최대 0.054 dB(16 kHz)입니다. 실측에서 이보다 큰 차이가 나온다면 그것은 도구가 아니라 측정 조건입니다.The picture below is the error the comparison tool itself contributes: one and the same impulse response put through a CAA session and through REW, differing by 0.007 dBon average and 0.054 dB at worst (at 16 kHz). Anything larger in a real comparison is the measurement conditions, not the tools.
이 검증이 말하지 않는 것What this does not prove
여기서 확인한 것은 계산이 맞다는 것입니다. 측정 전체의 정확도는 마이크와 인터페이스에도 달려 있습니다. 절대 음압(dB SPL)은 교정기(calibrator)로 마이크를 맞춘 만큼만 정확하고, 교정기가 없으면 상대값 — 채널 간 레벨 차, 도달 시간 차, 극성, 응답 모양 — 만 유효합니다. 다행히 DSP에 넣는 값은 모두 그 상대값입니다. CAA는 공인 교정 성적서를 발급하는 계측기가 아니며, 그런 용도로 쓰면 안 됩니다.What is established here is that the arithmetic is right. The accuracy of a real measurement also depends on the microphone and the interface. Absolute sound pressure is only as good as the calibrator used to set the microphone up; without one, only relative figures hold — level differences between channels, arrival-time differences, polarity, the shape of a response. Those happen to be every number you type into a DSP. CAA is not a certified instrument and must not be used as one.
