PyOctaveBand/tests/data/cnossos/ at main · jmrp.io/PyOctaveBand · Tangled

jmrp.io / PyOctaveBand

0

[READ-ONLY] Mirror of https://github.com/jmrplens/PyOctaveBand. [Python3] Octave-Band and Fractional Octave-Band filter. For signal in time domain. jmrplens.github.io/PyOctaveBand/

acoustics audio filter frequency frequency-analysis frequency-domain octave python3 signal time-domain

0

PyOctaveBand / tests / data / cnossos /

at main 8 files

José M. Requena Plens Read the glossary on a phone, and file the code the way the pages are filed 2mo ago

Read the glossary on a phone, and file the code the way the pages are filed Read the glossary on a phone, and file the code the way the pages are filed The glossary was 120 terms spread across markdown tables that read badly on a phone. It now comes from one source rendered as cards, each term carrying its symbol, unit, standard, clause and the guide that teaches it, with the guide link text read from the content collection at build time so a slug that stops existing fails the build instead of rotting. Three example fiches the guides described but never showed are generated now, found by a new per-language check that every committed fiche is actually shown on a page. A topic's landing page lists what it holds on a phone, and the byline moved off the top of every page into the footer. Then the code got the same treatment. scripts/generate_graphs.py had reached 21020 lines; three rounds of splitting brought the largest file in the repository down to 2862, which is the flat API surface and cannot be shorter. Files over a thousand lines went from 41 to 40, and that is the point: splitting a 21000-line file yields several of a thousand, and what changed is that none of the remaining ones has an honest seam left. CONTRIBUTING now states the rule those splits followed: there is no line limit, the test is cohesion, and a split is a move proven by byte-identical generated output. Splitting a module moves names between modules of the same package, and a deprecated 3.x import path names only one of them, so 47 names would have stopped resolving through a path that still promises a warning until 5.0. _compat.py now knows which modules a split fed, and a test holds every name to that promise. Auditing the layout afterwards turned up four things the splits had damaged and nothing had noticed: the pre-commit hook had quietly stopped regenerating the conformance report, CONTRIBUTING pointed contributors at a file that holds no figure builders any more, one figure could not be generated at all because moving its builder left a parent count one level short, and one renderer was filed under a result name instead of its domain. Six test files move to directories that already carried their subject, two guides move into the directory their sidebar group is named after, and eight section indexes stop listing pages from other sections under a heading that reads as an inventory. The ANP fleet database, the bridge from the EASA tables to the ECAC Doc 29 chain, had five public names and no guide: it appeared only in the generated reference, where you have to already know the name to find it. It has a page in both languages now, with two figures drawn off a real aircraft record, and it states the limit nothing user-facing did: only the 13 aircraft with fixed-point profiles have a ready-to-use trajectory. Verified throughout by regenerating what the code produces: 1700 committed figures match within tolerance, 70 fiches byte for byte, docs/CONFORMANCE.md identically with its 533 checks passing, 8011 tests, and the site building and validating in both languages.

2 months ago

rail_emission_cases.csv

Add the CNOSSOS-EU railway source emission model (Annex II 2.3 and Appendix G) (#456) * Add the CNOSSOS-EU railway source emission model of Annex II 2.3 * Cover every committed coefficient row in a regime that exposes it * Regenerate the conformance report, api docs and llms manifests * Give each workbook case its vehicle, so a repeated case number cannot shadow another * Record what the published test set can and cannot resolve * Refresh the generated manifests after rebase * State exactly how much of the published test set is reproduced * Import the shared reference data the way the rest of the suite does * Regenerate the railway emission figure after its axis change * Guard the logarithms on sign rather than on float equality, and split the assembly of a vehicle's spectra out of the source-power function The three comparisons Sonar flagged as float equality sit behind a validation that already rejects a negative value, so <= 0 selects exactly the same inputs as == 0 and returns the same -inf. The assembly function is split into the speed floor of 2.3.2, the flow term of (2.3.2)/(2.3.4) and the per-vehicle spectra, in the order the original built them. A 16 760-case sweep of railway_source_power over vehicles, tracks, speeds, angles, running conditions, interpolation rules and directivity editions returns byte-identical arrays before and after. * Regenerate the conformance report and manifests after the rebase

2 months ago

rail_frequency_tables_2015.csv

Add the CNOSSOS-EU railway source emission model (Annex II 2.3 and Appendix G) (#456) * Add the CNOSSOS-EU railway source emission model of Annex II 2.3 * Cover every committed coefficient row in a regime that exposes it * Regenerate the conformance report, api docs and llms manifests * Give each workbook case its vehicle, so a repeated case number cannot shadow another * Record what the published test set can and cannot resolve * Refresh the generated manifests after rebase * State exactly how much of the published test set is reproduced * Import the shared reference data the way the rest of the suite does * Regenerate the railway emission figure after its axis change * Guard the logarithms on sign rather than on float equality, and split the assembly of a vehicle's spectra out of the source-power function The three comparisons Sonar flagged as float equality sit behind a validation that already rejects a negative value, so <= 0 selects exactly the same inputs as == 0 and returns the same -inf. The assembly function is split into the speed floor of 2.3.2, the flow term of (2.3.2)/(2.3.4) and the per-vehicle spectra, in the order the original built them. A 16 760-case sweep of railway_source_power over vehicles, tracks, speeds, angles, running conditions, interpolation rules and directivity editions returns byte-identical arrays before and after. * Regenerate the conformance report and manifests after the rebase

2 months ago

rail_vehicles_2015.csv

Add the CNOSSOS-EU railway source emission model (Annex II 2.3 and Appendix G) (#456) * Add the CNOSSOS-EU railway source emission model of Annex II 2.3 * Cover every committed coefficient row in a regime that exposes it * Regenerate the conformance report, api docs and llms manifests * Give each workbook case its vehicle, so a repeated case number cannot shadow another * Record what the published test set can and cannot resolve * Refresh the generated manifests after rebase * State exactly how much of the published test set is reproduced * Import the shared reference data the way the rest of the suite does * Regenerate the railway emission figure after its axis change * Guard the logarithms on sign rather than on float equality, and split the assembly of a vehicle's spectra out of the source-power function The three comparisons Sonar flagged as float equality sit behind a validation that already rejects a negative value, so <= 0 selects exactly the same inputs as == 0 and returns the same -inf. The assembly function is split into the speed floor of 2.3.2, the flow term of (2.3.2)/(2.3.4) and the per-vehicle spectra, in the order the original built them. A 16 760-case sweep of railway_source_power over vehicles, tracks, speeds, angles, running conditions, interpolation rules and directivity editions returns byte-identical arrays before and after. * Regenerate the conformance report and manifests after the rebase

2 months ago

rail_wavelength_tables_2015.csv

Add the CNOSSOS-EU railway source emission model (Annex II 2.3 and Appendix G) (#456) * Add the CNOSSOS-EU railway source emission model of Annex II 2.3 * Cover every committed coefficient row in a regime that exposes it * Regenerate the conformance report, api docs and llms manifests * Give each workbook case its vehicle, so a repeated case number cannot shadow another * Record what the published test set can and cannot resolve * Refresh the generated manifests after rebase * State exactly how much of the published test set is reproduced * Import the shared reference data the way the rest of the suite does * Regenerate the railway emission figure after its axis change * Guard the logarithms on sign rather than on float equality, and split the assembly of a vehicle's spectra out of the source-power function The three comparisons Sonar flagged as float equality sit behind a validation that already rejects a negative value, so <= 0 selects exactly the same inputs as == 0 and returns the same -inf. The assembly function is split into the speed floor of 2.3.2, the flow term of (2.3.2)/(2.3.4) and the per-vehicle spectra, in the order the original built them. A 16 760-case sweep of railway_source_power over vehicles, tracks, speeds, angles, running conditions, interpolation rules and directivity editions returns byte-identical arrays before and after. * Regenerate the conformance report and manifests after the rebase

2 months ago

road_coefficients_2015.csv

Add the CNOSSOS-EU road traffic source emission model (#446) * Add the CNOSSOS-EU road traffic source emission model * Test the road emission model against the published EC test set, and document it * Assert every Table F-4 speed range and the temperature coefficients * Add the CNOSSOS-EU road emission figures * Move the source-line legend clear of the data and correct the figure captions * Regenerate the conformance report, API reference and llms indexes * Range-check the studded-tyre share and season * Drop the em-dashes from the changelog entry * Say precisely how many surfaces Table F-4 carries * Share the committed CNOSSOS road oracle loaders between the suite and the report * Reject an incomplete substituted database and fix the guide snippet * Regenerate the llms indexes for the guide edits * Note that the zero rolling row of category 4 is Table F-1's own * Regenerate the API reference page * Import the shared reference data the way every other suite does * Address the review: restore the section markers, guard partial tables, sharpen the prose * Say the workbook oracle uses the superseded 2015 database, not the shipped one * Take the remaining static-analysis findings on the new code * Pin the gradient dead band on both sides of every breakpoint The correction is continuous, so it is exactly zero at every breakpoint and a test that samples the edges cannot tell a widened branch from the printed one: at s = 2 % the category 1 uphill branch returns (2 - 2)/1,5 = 0 whether it opens at 0 % or at 2 %. The workbook cases use -15, -5, 0, +5 and +15 %, none of which lands in a gap, so five breakpoint mutations survived the whole suite, each of them turning the correction negative on the gentle slopes that make up most of a real network, by up to 0,70 dB. Two tests close the class rather than the five instances: one samples strictly inside each dead band and just past each breakpoint, and one sweeps the whole gradient range asserting the correction never subtracts, which is what every branch of (2.2.13) to (2.2.15) guarantees by construction. The energy sum of (2.2.3) also lived in two places, and only one of them was pinned by an exact-equality test. It moves into a helper both call sites use; the outputs are unchanged bit for bit over a 7488-case sweep of both public entry points. * Hoist the array out of the raises block in the segment-length test * Share the 2015 oracle database and state what speed_range does not do The wrapping of the superseded Appendix F rows into the library's coefficient types existed twice, once in the test suite and once in the conformance report. Two copies is a drift hazard rather than an untidiness: if one gained a field the other did not, the report would validate against a different database from the tests while both still passed. It now lives in tests/cnossos_road_oracle.py, which is the only module there that imports the library, so reference_data.py stays free of the code it is used to check. The report regenerates byte for byte identical. speed_range is carried for the reader and never enforced, because Table F-4 prints those columns without attaching a consequence; the docstring now says so instead of leaving it to be inferred. * Rebuild the exports and the plot module structurally after the rebase Resolving the rebase by taking both sides of each conflict block broke three Python files: two conflicts landed inside a parenthesised import list and one inside a dict literal, so the union produced unbalanced brackets and an unterminated string. It also left the CNOSSOS road labels declared in __all__ but never imported, and a translation key defined twice with different Spanish, where Python keeps the last and the railway wording was silently overridden. The three files are rebuilt from main plus exactly what the branch adds: the import block, the nineteen exported names, the plot renderer with its two constants and the seven translation entries. ruff sorts __all__, so the ordering convention is applied rather than guessed. * Regenerate the Spanish road figures for the single line-power label * Fold the oracle README back to four sections and say which correction applies to which term The rail and road branches each appended their own Source and authorship and Purpose and scope of use sections, so the oracle README had six headings where the convention is four, and the removal policy ended with two competing sentences. Merged, with the licence bullet stated once and a single closing sentence naming both oracles. The outdoor-sound section read as if every correction applied to both source terms. Checked against the signatures: temperature and studded tyres are rolling-only, gradient is propulsion-only, and pavement and junctions apply to both. Said that way now, in English and Spanish. * Put the road file rows back in the table and merge the exclusion notes Three table rows describing the road CSVs sat six paragraphs into the prose, so they rendered as literal pipe-delimited text, and the sentence they had displaced ended mid-clause. The rows are back in the table at the top and the sentence is whole again. The two near-identical exclusion paragraphs, one per workbook, are now one that names both.

2 months ago

road_emission_cases.csv

Add the CNOSSOS-EU road traffic source emission model (#446) * Add the CNOSSOS-EU road traffic source emission model * Test the road emission model against the published EC test set, and document it * Assert every Table F-4 speed range and the temperature coefficients * Add the CNOSSOS-EU road emission figures * Move the source-line legend clear of the data and correct the figure captions * Regenerate the conformance report, API reference and llms indexes * Range-check the studded-tyre share and season * Drop the em-dashes from the changelog entry * Say precisely how many surfaces Table F-4 carries * Share the committed CNOSSOS road oracle loaders between the suite and the report * Reject an incomplete substituted database and fix the guide snippet * Regenerate the llms indexes for the guide edits * Note that the zero rolling row of category 4 is Table F-1's own * Regenerate the API reference page * Import the shared reference data the way every other suite does * Address the review: restore the section markers, guard partial tables, sharpen the prose * Say the workbook oracle uses the superseded 2015 database, not the shipped one * Take the remaining static-analysis findings on the new code * Pin the gradient dead band on both sides of every breakpoint The correction is continuous, so it is exactly zero at every breakpoint and a test that samples the edges cannot tell a widened branch from the printed one: at s = 2 % the category 1 uphill branch returns (2 - 2)/1,5 = 0 whether it opens at 0 % or at 2 %. The workbook cases use -15, -5, 0, +5 and +15 %, none of which lands in a gap, so five breakpoint mutations survived the whole suite, each of them turning the correction negative on the gentle slopes that make up most of a real network, by up to 0,70 dB. Two tests close the class rather than the five instances: one samples strictly inside each dead band and just past each breakpoint, and one sweeps the whole gradient range asserting the correction never subtracts, which is what every branch of (2.2.13) to (2.2.15) guarantees by construction. The energy sum of (2.2.3) also lived in two places, and only one of them was pinned by an exact-equality test. It moves into a helper both call sites use; the outputs are unchanged bit for bit over a 7488-case sweep of both public entry points. * Hoist the array out of the raises block in the segment-length test * Share the 2015 oracle database and state what speed_range does not do The wrapping of the superseded Appendix F rows into the library's coefficient types existed twice, once in the test suite and once in the conformance report. Two copies is a drift hazard rather than an untidiness: if one gained a field the other did not, the report would validate against a different database from the tests while both still passed. It now lives in tests/cnossos_road_oracle.py, which is the only module there that imports the library, so reference_data.py stays free of the code it is used to check. The report regenerates byte for byte identical. speed_range is carried for the reader and never enforced, because Table F-4 prints those columns without attaching a consequence; the docstring now says so instead of leaving it to be inferred. * Rebuild the exports and the plot module structurally after the rebase Resolving the rebase by taking both sides of each conflict block broke three Python files: two conflicts landed inside a parenthesised import list and one inside a dict literal, so the union produced unbalanced brackets and an unterminated string. It also left the CNOSSOS road labels declared in __all__ but never imported, and a translation key defined twice with different Spanish, where Python keeps the last and the railway wording was silently overridden. The three files are rebuilt from main plus exactly what the branch adds: the import block, the nineteen exported names, the plot renderer with its two constants and the seven translation entries. ruff sorts __all__, so the ordering convention is applied rather than guessed. * Regenerate the Spanish road figures for the single line-power label * Fold the oracle README back to four sections and say which correction applies to which term The rail and road branches each appended their own Source and authorship and Purpose and scope of use sections, so the oracle README had six headings where the convention is four, and the removal policy ended with two competing sentences. Merged, with the licence bullet stated once and a single closing sentence naming both oracles. The outdoor-sound section read as if every correction applied to both source terms. Checked against the signatures: temperature and studded tyres are rolling-only, gradient is propulsion-only, and pavement and junctions apply to both. Said that way now, in English and Spanish. * Put the road file rows back in the table and merge the exclusion notes Three table rows describing the road CSVs sat six paragraphs into the prose, so they rendered as literal pipe-delimited text, and the sentence they had displaced ended mid-clause. The rows are back in the table at the top and the sentence is whole again. The two near-identical exclusion paragraphs, one per workbook, are now one that names both.

2 months ago

road_surfaces_2015.csv

Add the CNOSSOS-EU road traffic source emission model (#446) * Add the CNOSSOS-EU road traffic source emission model * Test the road emission model against the published EC test set, and document it * Assert every Table F-4 speed range and the temperature coefficients * Add the CNOSSOS-EU road emission figures * Move the source-line legend clear of the data and correct the figure captions * Regenerate the conformance report, API reference and llms indexes * Range-check the studded-tyre share and season * Drop the em-dashes from the changelog entry * Say precisely how many surfaces Table F-4 carries * Share the committed CNOSSOS road oracle loaders between the suite and the report * Reject an incomplete substituted database and fix the guide snippet * Regenerate the llms indexes for the guide edits * Note that the zero rolling row of category 4 is Table F-1's own * Regenerate the API reference page * Import the shared reference data the way every other suite does * Address the review: restore the section markers, guard partial tables, sharpen the prose * Say the workbook oracle uses the superseded 2015 database, not the shipped one * Take the remaining static-analysis findings on the new code * Pin the gradient dead band on both sides of every breakpoint The correction is continuous, so it is exactly zero at every breakpoint and a test that samples the edges cannot tell a widened branch from the printed one: at s = 2 % the category 1 uphill branch returns (2 - 2)/1,5 = 0 whether it opens at 0 % or at 2 %. The workbook cases use -15, -5, 0, +5 and +15 %, none of which lands in a gap, so five breakpoint mutations survived the whole suite, each of them turning the correction negative on the gentle slopes that make up most of a real network, by up to 0,70 dB. Two tests close the class rather than the five instances: one samples strictly inside each dead band and just past each breakpoint, and one sweeps the whole gradient range asserting the correction never subtracts, which is what every branch of (2.2.13) to (2.2.15) guarantees by construction. The energy sum of (2.2.3) also lived in two places, and only one of them was pinned by an exact-equality test. It moves into a helper both call sites use; the outputs are unchanged bit for bit over a 7488-case sweep of both public entry points. * Hoist the array out of the raises block in the segment-length test * Share the 2015 oracle database and state what speed_range does not do The wrapping of the superseded Appendix F rows into the library's coefficient types existed twice, once in the test suite and once in the conformance report. Two copies is a drift hazard rather than an untidiness: if one gained a field the other did not, the report would validate against a different database from the tests while both still passed. It now lives in tests/cnossos_road_oracle.py, which is the only module there that imports the library, so reference_data.py stays free of the code it is used to check. The report regenerates byte for byte identical. speed_range is carried for the reader and never enforced, because Table F-4 prints those columns without attaching a consequence; the docstring now says so instead of leaving it to be inferred. * Rebuild the exports and the plot module structurally after the rebase Resolving the rebase by taking both sides of each conflict block broke three Python files: two conflicts landed inside a parenthesised import list and one inside a dict literal, so the union produced unbalanced brackets and an unterminated string. It also left the CNOSSOS road labels declared in __all__ but never imported, and a translation key defined twice with different Spanish, where Python keeps the last and the railway wording was silently overridden. The three files are rebuilt from main plus exactly what the branch adds: the import block, the nineteen exported names, the plot renderer with its two constants and the seven translation entries. ruff sorts __all__, so the ordering convention is applied rather than guessed. * Regenerate the Spanish road figures for the single line-power label * Fold the oracle README back to four sections and say which correction applies to which term The rail and road branches each appended their own Source and authorship and Purpose and scope of use sections, so the oracle README had six headings where the convention is four, and the removal policy ended with two competing sentences. Merged, with the licence bullet stated once and a single closing sentence naming both oracles. The outdoor-sound section read as if every correction applied to both source terms. Checked against the signatures: temperature and studded tyres are rolling-only, gradient is propulsion-only, and pavement and junctions apply to both. Said that way now, in English and Spanish. * Put the road file rows back in the table and merge the exclusion notes Three table rows describing the road CSVs sat six paragraphs into the prose, so they rendered as literal pipe-delimited text, and the sentence they had displaced ended mid-clause. The rows are back in the table at the top and the sentence is whole again. The two near-identical exclusion paragraphs, one per workbook, are now one that names both.

2 months ago

README.md

What this folder contains#

File	Kind	Provenance
`rail_emission_cases.csv`	derived (extract of a published result set)	123 of the 19 872 cases of `CNOSSOS_RAIL_EMISSION_TEST.xlsx`, the railway emission test workbook published by the European Commission on CIRCABC with the CNOSSOS-EU source module. Each row carries the full case description read straight out of the workbook, together with the workbook's own per-octave-band and total line-power levels, copied verbatim to the two decimals it prints.
`rail_vehicles_2015.csv`	transcribed	The vehicle definitions of `CNOSSOS_Rail_Vehicles.xml` v1.1 (catalogue date 2015-04-09), the vehicle database of the same Commission source module: number of axles and the reference into each spectrum table.
`rail_wavelength_tables_2015.csv`	transcribed	The spectra of that catalogue and of `CNOSSOS_Rail_Track.xml` v1.0 (2014-04-27) that are given against wavelength: wheel roughness, contact filter, rail roughness and impact roughness, on the 32-step 1/3-octave grid from 1 000 mm to 0,8 mm.
`rail_frequency_tables_2015.csv`	transcribed	The spectra of the same two catalogues that are given against frequency, in the 24 1/3-octave bands from 50 Hz to 10 kHz: wheel, track and superstructure transfer functions, traction sound power for constant speed and for idling, and the aerodynamic reference spectra.
`road_emission_cases.csv`	derived (extract of a published result set)	60 of the 4 875 cases of `CNOSSOS_ROAD_EMISSION_TEST.xlsx`, the road emission test workbook published by the European Commission on CIRCABC with the CNOSSOS-EU source module. Each row carries the segment description recovered from the workbook (see below) and the workbook's own per-octave-band and total line-power levels, copied verbatim to the two decimals it prints.
`road_coefficients_2015.csv`	transcribed	Table F-1 of Appendix F of Annex II to Directive 2002/49/EC as published in Commission Directive (EU) 2015/996 (OJ L 168, 1.7.2015, p. 124), that is the version the workbook was computed with and which Commission Delegated Directive (EU) 2021/1226 later replaced. Machine-transcribed from the Official Journal text.
`road_surfaces_2015.csv`	transcribed	Table F-4 of the same 2015 text (OJ L 168, 1.7.2015, pp. 125-129), keyed by the `NLxx` surface identifiers the workbook uses. Also superseded by (EU) 2021/1226.

These are the 2015 coefficients, not the ones the library ships. The workbook was computed in 2014 with the database that Commission Directive (EU) 2015/996 went on to print in Appendix G of Annex II, before Commission Delegated Directive (EU) 2021/1226 replaced Tables G-1b, G-2, G-3a, G-4 and G-7. The coefficients of the current text live in tests/reference_data/, transcribed from the Official Journal, and pin the tables shipped in phonometry.environment.sources.cnossos_rail.

Only the running conditions the Directive models are transcribed. The catalogue carries four (constant, accelerating, decelerating, idling) where Annex II 2.3.2 models two, and states that L_W,0,const,i = L_W,0,idling,i; the accelerating and decelerating spectra are a database extension with no counterpart in the method, so neither they nor the workbook cases that use them are committed here.

Three of the workbook's twenty-three vehicles are left out as well, because their rows record defects of the reference program rather than of the method: vehicle 15 carries n instead of a number in its axle-count and length fields and produces no rolling noise at all, and vehicles 24 and 25 have no traction data, which makes visible an all-zero aerodynamic spectrum that the program energy-sums below 200 km/h. The remaining twenty vehicles are reproduced exactly: the shipped model reproduces all 34 560 published rows of those twenty vehicles, 17 280 cases at both source heights, to 0,0055 dB.

Source and authorship#

The workbook is CNOSSOS_RAIL_EMISSION_TEST.xlsx, delivered with the Testing of Emission DLL's for CNOSSOS-EU Road, Rail and Industry Noise Sources report (Stapelfeldt Ingenieurgesellschaft mbH, Dortmund, doc. rev. 1406-2, 7 July 2014) as part of the CNOSSOS-EU source module of the Joint Research Centre of the European Commission. Its results were produced twice, once by the reference emission library of DGMR and once by an independent implementation, and the two agree to 0,01 dB. The XML catalogues are delivered with the same module. Everything is distributed through the Commission's CIRCABC document repository (https://circabc.europa.eu/, CNOSSOS-EU interest group), retrieved 2026-07-29.
Official Journal texts are published by the Publications Office of the European Union; reuse is authorised under Commission Decision 2011/833/EU, provided the source is acknowledged, which the header of every file here and the comments in tests/reference_data/ do.
The workbook is CNOSSOS_ROAD_EMISSION_TEST.xlsx, delivered with the Testing of Emission DLL's for CNOSSOS-EU Road, Rail and Industry Noise Sources report (Stapelfeldt Ingenieurgesellschaft mbH, Dortmund, doc. rev. 1406-2, 7 July 2014) as part of the CNOSSOS-EU source module of the Joint Research Centre of the European Commission. It is distributed through the Commission's CIRCABC document repository (https://circabc.europa.eu/, CNOSSOS-EU interest group), retrieved 2026-07-29. Its results were produced twice, once by the reference emission library of DGMR and once by an independent implementation, and agree to 0,01 dB.
The coefficient tables are the text of Commission Directive (EU) 2015/996 (OJ L 168, 1.7.2015, p. 1). Official Journal texts are published by the Publications Office of the European Union; reuse is authorised under Commission Decision 2011/833/EU, provided the source is acknowledged, which the header of every file here does.
This repository's MIT licence does not cover these files. They are redistributed under the Commission's reuse policy as Commission publications.

Purpose and scope of use#

tests/environment/sources/test_cnossos_rail.py feeds the shipped equations of Annex II 2.3 the 2015 coefficient set and requires the workbook's levels back, band by band, at both source heights. That is the only oracle that pins the equations: the Directive itself prints no worked example anywhere in 2.3 or Appendix G. The 123 committed cases cover all twenty usable vehicles, both source heights, all three speeds, both running conditions, all nine track-parameter combinations, all four receiver-angle pairs and both flow rates, so the committed subset reaches every level of every factor the full set varies; what it does not reach is the sheer number of their combinations.

What the agreement can and cannot resolve is worth stating, because it is a property of the source model and not of the test selection. A coefficient that sits twenty decibels below the term beside it does not move the total, so the workbook constrains the loud rows of the database and not the quiet ones. Corrupting every committed coefficient one at a time (sign flip, decimal shift up and down, transposition with the neighbouring cell) and recomputing all cases kills 1 078 of the 1 116 mutations of the roughness and contact-filter tables, but only 3 010 of the 4 015 mutations of the transfer and traction tables. The survivors are concentrated in rows the published test set never brings to the surface: the all-zero and all-140 dB "min" and "max" placeholders of the IMAGINE catalogue, the source-A column of the constant-speed traction, which rolling noise always covers, and the two track transfer functions the workbook only ever pairs with the 140 dB superstructure placeholder. A mistranscription in one of those cells could not change a computed value, so it could not turn a wrong model into a passing run either; it would simply go unnoticed, and the coverage assertion in the test module keeps at least one unmasked case for every committed row.

The published levels themselves are pinned hard: 3 805 of the 3 806 mutations of the 984 committed band levels are caught. The single survivor transposes two levels of a saturated case that differ by 0,01 dB, the last digit the workbook prints.

Because the workbook predates the 2021 amendment, the cases are run with the 2015 vertical directivity, with the curve-squeal excess and the bridge constant of the 2015 text supplied as data columns, and with the roughness speed floor switched off, which is how the reference program behaves. Everything the 2021 amendment changed is therefore covered by the table transcriptions and by the closed-form assertions in the same test module, not by these cases.

The workbook tabulates the total line power of a segment carrying all five vehicle categories, while each of its rows names only the category whose flow and speed that row varies. The traffic composition behind each block was recovered by exhaustive search over the flow and speed grids of the workbook and then verified against all 4 875 rows, which it reproduces to 0,005 dB, i.e. inside the workbook's own rounding. road_emission_cases.csv therefore writes the five flows out explicitly, so the committed cases are readable without the reconstruction.

The coefficients of the current text, Tables F-1 to F-4 as they stand after (EU) 2021/1226, are not here: they live in tests/reference_data/, also machine-transcribed from the Official Journal, and pin the tables shipped in phonometry.environment.sources.cnossos_road.

tests/environment/sources/test_cnossos_road.py feeds the shipped equations of Annex II 2.2 the superseded 2015 coefficient set and requires the workbook's levels back, band by band. That is the only oracle that pins the equations: the Directive itself prints no worked example anywhere in 2.2 or Appendix F. The 60 committed cases span all thirteen road surfaces exercised by the workbook, all five vehicle categories, all five gradient and junction combinations, all four temperature and studded-tyre combinations and all twelve speeds, so the committed subset reaches everything the full set does except the sheer number of repetitions.

Deliberately excluded: the two workbooks themselves (10 MB for the railway set and 9 MB for the road one, almost all of it repeated cases), the CNOSSOS-EU reference source module and its libraries, and the Official Journal PDFs. The data are not part of the phonometry package and are not installed with it.

Removal policy#

If you represent the European Commission, the Joint Research Centre or Stapelfeldt Ingenieurgesellschaft and consider that committing this subset exceeds the intended reuse of the published material, please open an issue or contact the maintainer (see CITATION.cff) and it will be removed promptly. The suite would then lose its only oracle for the railway and road emission equations alike: the Appendix F and G transcriptions and the closed-form assertions would still run, but the end-to-end cases would have to go with them.