Noise Certification Test

Noise Certification Test (flight-test-operations/planning/noise-certification-test)

Use when the task is planning and analyzing the noise certification flight test for a transport category airplane: laying out the flyover, sideline and approach measurement conditions, reducing the measured tone corrected perceived noise level (PNLT) time history into the effective perceived noise level (EPNL) with the 10 dB down integration rule, and checking the per point and cumulative margins against the stated noise limits. This leaf implements the EPNL math and the test geometry layer in pure Python, stdlib only. The measured PNLT input is the tone corrected perceived noisiness output of the acoustic measurement chain; the noy conversion and the acoustic data acquisition hardware are out of scope. It pairs with flight-test-operations/planning/flight-test-planning for program sequencing, flight-test-operations/planning/flight-test-instrumentation for the acoustic measurement system, and flight-test-operations/planning/position-error-calibration for the airspeed reference flown at each certification condition.

Domain quick reference

The noise certification measurement procedure is summarized here at reference level from the public FAR 36 / ICAO Annex 16 Volume I description (name and paraphrase only, no verbatim regulation text; the certification basis states the applicable limits, which this leaf accepts as inputs).

• Flyover: the airplane accelerates at maximum takeoff thrust along the runway and passes over the reference point on the extended centerline 6500 m from the brake release point, speed reference V2 + 10 kt.
• Sideline: the microphone sits 450 m lateral of the runway centerline at the point of maximum noise during the takeoff run.
• Approach: the airplane descends the 3 degree glide slope to the threshold; the microphone is 1200 m from the threshold under the flight path at 120 m altitude.
• Effective perceived noise level: EPNL = 10 * log10((1 / T0) * sum(10^(PNLT_i / 10)) * dt), T0 = 10 s, summed over the 10 dB down interval where the PNLT stays within 10 EPNdB of its maximum.
• 10 dB down rule: find the PNLT maximum, then walk left and right to the times where the PNLT has dropped 10 dB below that maximum; boundary times are linearly interpolated between the bracketing samples. A series that never drops 10 dB is integrated over its full span and flagged as not truncated.
• Margin to limit: margin_db = limit - EPNL, pass when margin_db >= 0.
• Cumulative rule (typical chapter 4 / stage 4 check at reference level): the sum of the three per point margins must be >= 10 EPNdB and every individual margin must be >= 0. The exact rule set depends on the certification basis.
• Units: geometry in meters and degrees, EPNL in EPNdB, time in seconds, airspeeds in knots.

Workflow

1. Lay out the three measurement conditions with geometry(condition) for flyover, sideline and approach; each returns the reference distances from the module constants.
2. Build the certification test matrix from the weight and speed inputs with test_matrix(takeoff_weight, landing_weight, v2_kt, approach_speed_kt, limits); each row carries the configuration, reference speed, stated limit and target EPNL.
3. For each measured run, integrate the tone corrected PNLT time history with epnl_from_pnlt(pnlt_series, dt) to get the EPNL, the 10 dB down window bounds and the truncation flag.
4. Check every point against its stated limit with margin_to_limit(epnl, limit) to get the margin and the pass or fail verdict.
5. Assess the three point set with cumulative_margin(margins) against the chapter 4 style rule: summed margin at or above the required 10 EPNdB and no negative individual margin.
6. Combine everything with summarize(epnl_by_condition, limits) for the per condition EPNL, margin, verdict and the cumulative verdict.
7. Confirm the deterministic checks with the contract test scripts/test_noise_certification_test.py.

Worked example

A transport airplane noise certification data set, 41 PNLT samples at 0.5 s spacing (20 s per run):

• Constant 90 dB flyover run that never drops 10 dB: the full 20 s series is integrated and EPNL = 90 + 10 * log10(20 / 10) = 93.0103 EPNdB (window 0.0 to 20.0 s, not truncated).
• Single peak flyover run peaking at 92.0 dB at sample 20 with values falling 10 dB below the maximum at the series ends: EPNL = 89.1267 EPNdB over the truncated 10 dB down window from 4.0 s to 16.0 s.
• Pulse approach run at 100 dB from 8 s to 12 s and 80 dB elsewhere: EPNL = 96.3094 EPNdB, below the 100 dB instantaneous peak because the effective duration is below the 10 s reference; the 90 dB crossings are linearly interpolated at 7.75 s and 12.25 s.
• Margin: margin_to_limit(93.01, 95.0) returns margin 1.99 EPNdB, verdict pass; margin_to_limit(96.0, 95.0) fails.
• Cumulative: margins [3.0, 4.0, 4.0] sum to 11.0 EPNdB with all margins at or above zero, verdict pass; [3.0, 3.0, 3.0] fails on the 9.0 EPNdB sum; [-1.0, 6.0, 6.0] fails on the negative individual margin.
• Geometry: geometry("approach") returns distance 1200.0 m, altitude 120.0 m, glide 3.0 degrees.

Pitfalls

• Truncating a run that never drops 10 dB below its peak: the constant 90 dB flyover integrates the full 20 s series to EPNL = 93.0103 EPNdB with a not-truncated flag and the full-span window.
• Reading EPNL as the instantaneous peak: the pulse approach at 100 dB from 8 to 12 s gives EPNL = 96.3094 EPNdB because the effective duration sits below the 10 s reference, and the single-peak 92 dB series lands near 89 EPNdB.
• Forgetting that the 10 dB down window is interpolated: the peaked-run crossings at 7.75 s and 12.25 s bound the truncated integration window and are linearly interpolated, not snapped to sample times.
• Judging compliance from the individual margin alone: margin_to_limit must pass and the cumulative margins must sum to at least the limit with every individual margin at or above zero - a negative single margin fails the cumulative verdict.
• Quoting EPNL differences between runs at different levels: adding c dB to every PNLT sample adds c EPNdB (level-shift identity), so comparisons only mean something at the same measurement conditions.
• Feeding non-physical inputs: an empty or non-finite PNLT series, dt <= 0, an unknown condition, and negative noise limits all raise ValueError.

Verification

• Confirm epnl_from_pnlt([90.0] * 41, 0.5) returns 93.0102999566 within 1e-6, matching 90 + 10 * log10(20 / 10).
• Confirm the single peak 92 dB series stays near 89.0 EPNdB and below the instantaneous peak level.
• Confirm the constant series returns a not truncated flag and the full span bounds; the peaked series returns truncated True with the interpolated window bounds.
• Confirm margin_to_limit(93.01, 95.0) returns 1.99 pass and that a level shift identity holds: adding c dB to every PNLT sample adds c EPNdB to the EPNL.
• Confirm ValueError rejection of an empty or non-finite PNLT series, dt <= 0, an unknown condition, and negative noise limits.
• Run the contract test offline: python3 scripts/test_noise_certification_test.py (33 tests, deterministic).

Related leaves

• flight-test-operations/planning/flight-test-planning: orders the noise certification runs inside the overall build-up sequence.
• flight-test-operations/planning/flight-test-instrumentation: owns the acoustic measurement system that delivers the PNLT series.
• flight-test-operations/planning/position-error-calibration: provides the airspeed reference flown at each certification condition.
• flight-test-operations/planning/test-point-matrix-design: expands the certification conditions into the full flown point matrix.

Behavior contract (gate 3)

Run the deterministic contract test (stdlib unittest, offline):

python3 scripts/test_noise_certification_test.py

The test covers the reference geometry table for the three conditions, the EPNL integration contract (the constant series closed form 93.0103 EPNdB, the ~89.0 EPNdB truncated peak series with interpolated window bounds, the 100 dB pulse staying below 100 EPNdB, the level shift identity), the margin to limit verdicts, the cumulative rule branches (pass, sum shortfall, negative individual margin), the test matrix and summary helpers, and ValueError rejection of empty, non-finite and non-physical inputs.

Compliance

• Standards referenced, not reproduced: the noise certification measurement procedure (FAR 36 / ICAO Annex 16 Volume I) is named and summarized at reference level only, never quoted verbatim; the mapped standards ids far-25 and cs-25 frame the transport type certification context this test supports. The acoustic math above is the standard engineering method, summary-only per standards-map.yaml.
• compliance: STANDARDS-REF, gated: false.