CANM's aircraft engine implements ECAC.CEAC Doc 29 4th ed. Vol 2 segmentation method (ICAO Doc 9911 equivalent; SAE-AIR-1845 / AIR-5662). This page runs the ECAC Doc 29 Volume 3 Part 1 reference cases - the published conformance benchmark for this method - and compares the result against the published answers. It computes on every page load, so it cannot drift out of date.
Aircraft method revision 2026-08-25-a
The rest of CANM predicts a steady level from a fixed source: a sound power level in, ISO 9613-2 attenuation off it, a level out. None of that framing survives contact with an aircraft.
ECAC Doc 29, ICAO Doc 9911 and United States federal practice are not separate methods. Each computes the segmentation chain defined by SAE-AIR-1845, with lateral attenuation from SAE-AIR-5662, the current revision superseding AIR-1751. CANM implements those same two standards, so the question of agreement reduces to one that can be settled arithmetically: are the coefficients the published ones?
The checks below settle it. Each takes a point where the standard fixes the answer in advance - a join the coefficients have to reach on their own, or a designed value - and evaluates it now. These are the points an approximation cannot fake: the air-to-ground term is not clamped to zero at 50°, the published coefficients arrive there themselves, and a fuselage-mounted engine loses exactly half its power in the wing plane. Round any coefficient and the joins open by a visible amount.
| Standard | Quantity | Published | CANM | Difference | Why this one | |
|---|---|---|---|---|---|---|
| SAE-AIR-5662 | Air-to-ground attenuation at 0 degrees elevation | 10.8570 | 10.8570 | 1.78e-15 | holds | The maximum the model allows. An aircraft on the horizon, far to the side, loses this much and no more. |
| SAE-AIR-5662 | Air-to-ground term where the standard sets it to vanish (50 deg) | 0.0000 | 0.0000 | 2.00e-05 | holds | The formula is not clamped to zero here - the published coefficients arrive at zero on their own. Round any one of them and this join opens. |
| SAE-AIR-5662 | Ground-distance factor at 914 m (3000 ft) | 1.0000 | 1.0000 | 1.67e-06 | holds | Beyond 914 m the factor is unity by definition. The coefficients reach it without being told to, which is the same test applied to the other join. |
| SAE-AIR-5662 | Attenuation above 50 degrees elevation | 0.0000 | 0.0000 | 0.00e+00 | holds | Directly overhead there is no lateral attenuation at all. This is the branch the continuity check above justifies. |
| SAE-AIR-1845 | Impedance adjustment at the NPD reference atmosphere (25 C) | 0.0000 | 0.0000 | 4.22e-06 | holds | NPD data are normalised to 25 C at sea level. At exactly those conditions the adjustment has to vanish, and it only does so if both the 409.81 Ns/m3 reference impedance and the 416.86 constant are the published ones. |
| SAE-AIR-1845 | Impedance adjustment at ISA sea level (15 C) | 0.0741 | 0.0741 | 2.33e-05 | holds | A real, non-zero adjustment, shown alongside the one above so the zero is visibly a property of the reference condition rather than the term being inert. |
| SAE-AIR-1845 | Duration correction at the reference speed | 0.0000 | 0.0000 | 0.00e+00 | holds | An aircraft flying at the speed the NPD curve was measured at is over the receiver for exactly the tabulated time. |
| SAE-AIR-1845 | Duration correction at half the reference speed | 3.0103 | 3.0103 | 4.34e-08 | holds | Half the speed is twice the exposure time, which is 3 dB on an energy quantity. The sign matters: slower is louder, not quieter. |
| Doc 29 / AIR-1845 | Installation correction directly beneath, wing-mounted | 0.0000 | 0.0000 | 0.00e+00 | holds | Under the flight path there is no installation effect for either mount type. The term exists to describe what happens to the SIDE. |
| Doc 29 / AIR-1845 | Installation correction directly beneath, fuselage-mounted | 0.0000 | 0.0000 | 0.00e+00 | holds | The same join for the other airframe, which the coefficients have to reach independently. |
| Doc 29 / AIR-1845 | Installation correction in the wing plane, fuselage-mounted | -3.0000 | -3.0000 | 3.23e-05 | holds | Out to the side a fuselage-mounted engine loses exactly half its power. Landing on 3.0000 dB is not a coincidence - it is a designed value of the standard, and it is what an approximated coefficient would miss. |
| Doc 29 / AIR-1845 | Installation correction for a propeller aircraft | 0.0000 | 0.0000 | 0.00e+00 | holds | Doc 29 4-16 applies no directivity correction to propellers at any angle, so this term must be identically zero rather than small. |
The checks above test the terms one at a time. These run a real movement past real receivers and test what has to be true of the whole chain - the properties a wrong assembly breaks even when every individual coefficient is right.
The first is the one that matters most. Doc 29's segmentation approximates a continuous flight path by summing finite pieces, and it is only valid if the answer does not depend on how finely the path is cut. Refining from 1000 m pieces to 25 m is a factor of forty, and the level has to sit still through it. If that sum did not converge, the number it produced would be an artefact of the segmentation rather than a property of the flight.
| What must be true | Fixed by | Required | Computed | Difference | Why it matters | |
|---|---|---|---|---|---|---|
| Answer is independent of how finely the path is cut | Doc 29 4.4 (segmentation) | 90.0314 | 90.0420 | 1.06e-02 | holds | A flight path cut into 1000 m pieces and the same path cut into 25 m pieces must give the same level. If the sum did not converge, the answer would be an artefact of the segmentation rather than a property of the flight. |
| Doubling the movements adds 3 dB | Doc 29 4.6 / ISO 1996-1 (energy addition) | 3.0103 | 3.0103 | 4.34e-08 | holds | Levels add on energy, not arithmetically. Two identical movements are 3 dB above one, and this is the property a wrong summation breaks first. |
| Night movements carry a 10 dB penalty | Directive 2002/49/EC Annex I (Lden) | 10.0000 | 10.0000 | 0.00e+00 | holds | One movement at night counts as ten would by day. The evening penalty is 5 dB by the same definition. |
| Evening movements carry a 5 dB penalty | Directive 2002/49/EC Annex I (Lden) | 5.0000 | 5.0000 | 0.00e+00 | holds | The other half of the same weighting. |
| A barrier between runway and receiver changes nothing | Doc 29 4.5.3 fn.17 and 2.5(6) | 0.0000 | 0.0000 | 0.00e+00 | holds | The segmentation geometry neglects obstacles on the path, and the noise-power-distance data are themselves normalised to open ground. Ground noise - taxiing, run-ups, auxiliary power - is excluded from this method and is screened by the industrial engine instead. |
One method, several statements. ECAC Doc 29, ICAO Doc 9911 and United States federal practice are not separate methods. Each computes the segmentation chain defined by SAE-AIR-1845, with lateral attenuation from SAE-AIR-5662. CANM implements those same two standards, so the algorithms agree term for term, and the checks above are what shows the coefficients do too.
The lateral term is the current revision, not the legacy one. SAE-AIR-5662 superseded AIR-1751, and the difference is structural rather than cosmetic: AIR-1751 folds the engine installation effect into the lateral term, where AIR-5662 keeps them separate, and the two give different answers to the side of the track. CANM implements AIR-5662, separated, which is why installation is its own term in the chain.
ICAO Annex 16 is certification, not a contour method. Annex 16 Volume I fixes the effective perceived noise levels a type is certificated to at the lateral, flyover and approach reference points. It does not compute contours and is not a substitute for the method above. Its role here is to anchor the reference aircraft to certificated levels, so a substitute type is offset from a published figure rather than from an assumption.
The data decide the answer. The method being right is necessary and not sufficient. A contour is only as good as the noise-power-distance curves fed to it, and the authoritative sets are published separately and not redistributed here. Every source records which set its curve came from, and a substitute curve is labelled as one wherever it appears.
The checks above show the coefficients are right. These show the whole chain assembled: the published conformance benchmark for the method, run end to end and compared against the answers in Doc 29 Volume 3.
| Receptor | x, y (m) | Doc 29 SEL | CANM SEL | Difference | What this receptor tests | |
|---|---|---|---|---|---|---|
| R01 | 6500, 0 | 90.03 | 90.04 | 0.01 | match | Under the climb-out, 6.5 km from start of roll. Airborne segments only. |
| R02 | 0, 200 | 102.82 | 100.91 | -1.91 | not quotable | Beside the start of roll. Takeoff ground roll dominates.
0.64 of this level comes from behind the start of roll. The rearward directivity of Doc 29 equation 4-24a is implemented, and it removed most of this error: receptor R03 was 28 dB high without it. What remains is a known convergence defect in the behind-roll noise fraction, characterised but not yet closed, so CANM marks a receiver fed from that region as not quotable rather than reporting a confident number over it. |
| R03 | -500, 0 | 76.13 | 74.83 | -1.30 | not quotable | Directly behind the start of roll, on the extended centreline. Needs the rearward jet lobe.
0.84 of this level comes from behind the start of roll. The rearward directivity of Doc 29 equation 4-24a is implemented, and it removed most of this error: receptor R03 was 28 dB high without it. What remains is a known convergence defect in the behind-roll noise fraction, characterised but not yet closed, so CANM marks a receiver fed from that region as not quotable rather than reporting a confident number over it. |
| R04 | -500, 500 | 83.20 | 81.47 | -1.73 | not quotable | Behind and to the side. Rearward lobe again, but off axis.
0.98 of this level comes from behind the start of roll. The rearward directivity of Doc 29 equation 4-24a is implemented, and it removed most of this error: receptor R03 was 28 dB high without it. What remains is a known convergence defect in the behind-roll noise fraction, characterised but not yet closed, so CANM marks a receiver fed from that region as not quotable rather than reporting a confident number over it. |
| R05 | 3000, 500 | 92.68 | 92.65 | -0.03 | match | Sideline at the runway end. Ground roll and airborne mixed. |
| Receptor | x, y (m) | Doc 29 SEL | CANM SEL | Difference | What this receptor tests | |
|---|---|---|---|---|---|---|
| R02 | 0, 200 | 91.09 | 89.96 | -1.13 | differs | Beside the start of roll. Landing roll. |
| R03 | -500, 0 | 104.59 | 106.46 | 1.87 | differs | Behind the start of roll, at the end of the landing roll. |
| R04 | -500, 500 | 82.11 | 81.43 | -0.68 | differs | Behind and to the side. |
| R05 | 3000, 500 | 64.22 | 65.94 | 1.72 | differs | Sideline. Late approach plus landing roll. |
| R13 | -24400, -500 | 78.22 | 78.90 | 0.68 | differs | Under the approach, 24 km out. Airborne only. |
| R18 | -2000, 0 | 98.45 | 99.23 | 0.78 | differs | 2 km from the threshold. Airborne, low and slow. |
The airborne receptors match. R01, under the climb-out, comes out within 0.01 dB of the published answer, and it stays there whether the flight path is cut into 40 segments or 1,400. That single number exercises almost the whole method at once: NPD interpolation (linear in power, logarithmic in distance), the impedance adjustment, the duration correction, the engine installation term, lateral attenuation, the finite segment energy fraction, and the decibel summation over segments. If any one of those were wrong, R01 would not land on 90.0.
The receptors behind the start of roll do not match, and we know exactly why. Doc 29 treats an observer behind the takeoff roll as a package: a reduced energy fraction that deliberately over-counts, brought back down by a rearward directivity function (equation 4-24a) that models the lobed jet exhaust pattern. CANM applies the first half and not the second, because the 4-24a coefficients are typeset in the published PDF with symbol-font glyphs that do not extract reliably - and a directivity function reconstructed from a bad character read would be wrong by an unknown amount in exactly the place where the most exposed dwellings are.
The size and pattern of the gap confirm the diagnosis: it is largest at R03, directly on the extended centreline where the lobe is strongest, and much smaller at R04, off to the side. Supply the function and both close; the geometry that feeds it (equation 4-22) is already implemented and tested.
The arrival cases carry a constant offset of about 0.7 dB. It is identical at R13 (24 km out) and R18 (2 km out), and it does not move at all when the segmentation is refined from 1,000 m sub-segments down to 25 m - so it is a systematic term, not a numerical artefact. At those receptors the aircraft passes overhead, where the lateral attenuation and the installation correction are both exactly zero, which narrows it to the NPD, duration and energy-fraction chain. Pinning it down needs the reference segmental results - worksheet B-2 of the Doc 29 workbook, which lists every intermediate quantity per segment and exists for precisely this purpose. That workbook is distributed separately from the PDF and is not in hand.
| Doc 29 Vol 2 | Element | Status |
|---|---|---|
| 4-3, 4-4, 4-5 | NPD interpolation and extrapolation | Implemented, unit-tested |
| 4-6, 4-7 | Acoustic impedance adjustment | Implemented; reproduces the documented 0.074 dB |
| 4-8, 4-10, 4-11 | Segment levels, LAmax and LE | Implemented |
| 4-12, 4-13 | Segment power and speed | Implemented |
| 4-14 | Duration correction | Implemented |
| 4-15, 4-16 | Engine installation, wing / fuselage / prop | Implemented |
| 4-18, 4-19 | Lateral attenuation (SAE-AIR-5662) | Implemented |
| 4-20, 4-21 | Finite segment energy fraction | Implemented; energy-conserving to 1e-9 |
| 4-22 | Start-of-roll azimuth geometry | Implemented |
| 4-24a, 4-24b | Start-of-roll directivity coefficients | Not shipped - injectable |
| Appendix B | Flight performance from procedural steps | Not implemented - fixed-point profiles only |
| 3.6.6 | Curved ground tracks and bank angle | Bank accepted; turn construction not implemented |
Reference cases from ECAC.CEAC Doc 29, 4th Edition, Volume 3 Part 1 (December 2016), Tables A-6, A-7, A-9 to A-12 and B-1. The three reference aircraft are hypothetical, defined by the document so that conformance testing does not depend on the ANP database.
