Repository navigation
Releases: MehdiZD97/3gpp-tr-library
Release list
v0.2.0 - tr3gpp on PyPI
Adds the final planned TR 38.901 clause and makes the Python package installable
from PyPI.
New content
- TR 38.901 v19.4.0 §7.6: additional modelling components, completing the
channel-model chapter.
Full coverage is now TR 38.901 §7.4 (pathloss, LOS probability, O2I penetration),
§7.5 (fast fading), §7.6 (additional modeling components) and §7.9 (ISAC), plus
TR 36.777 v15.0.0 Annex B (aerial-UE channel model).
Now on PyPI
pip install tr3gppfrom tr3gpp import tr38901
entry = tr38901.section("7.4").pathloss(scenario="UMi-StreetCanyon", condition="NLOS")
entry.shadow_fading_std_dbThe install is self-contained, no clone required. A tr3gpp command ships with
it, and the API is self-describing, so sections and parameters can be discovered
without knowing their names:
tr3gpp list
tr3gpp describe tr38901 7.6
tr3gpp dump tr38901 7.5 channel_model_parameters --format csv > lsp.csvBreaking changes
- The package is renamed
tr_api→tr3gpp, which is now simultaneously the
PyPI name, the import name and the console command. The previoustr-api
command collided with an unrelated published project. Update imports from
from tr_api import ...tofrom tr3gpp import ....
Verification
Values marked verified are cross-checked against multiple independent source
formats. Where a source renders equations as images and no automated cross-check
is possible, content is PDF-visual-verified and recorded as single-source rather
than overstated. CSV↔YAML consistency, schema validity and formula cross-checks
run in CI.
v0.1.0 - TR 38.901 §7.4, §7.5, §7.9 and TR 36.777 Annex B
First public release of 3gpp-tr-library — hand-verified, section-level 3GPP
Technical Report content as queryable Markdown, CSV and YAML, with a typed
Python API and CLI.
What's included
| TR | Version | Section | Coverage |
|---|---|---|---|
| TR 38.901 | v19.4.0 | §7.4 | Pathloss, LOS probability, O2I (building/car) penetration loss, shadow-fading autocorrelation |
| TR 38.901 | v19.4.0 | §7.5 | Full 12-step fast-fading generation procedure (30 equations) + all 12 supporting tables, incl. the 49×16 channel-model-parameter table |
| TR 38.901 | v19.4.0 | §7.9 | Complete Rel-19 ISAC clause (7.9.0–7.9.6): sensing scenarios, RCS/XPR models, reference-channel mapping, target/background fast fading, additional components, calibration assumptions |
| TR 36.777 | v15.0.0 | Annex B | Aerial-UE (RMa-AV/UMa-AV/UMi-AV) channel model — height-dependent deltas to TR 38.901's terrestrial models |
Every section ships in three coordinated formats: Markdown for reading and RAG
ingestion, one CSV per real TR table number for tabular workflows, and YAML of
queryable parameters validated against Pydantic models.
Getting started
git clone https://github.com/MehdiZD97/3gpp-tr-library.git
pip install -e 3gpp-tr-library/tools/tr_apifrom tr_api import tr38901
entry = tr38901.section("7.4").pathloss(scenario="UMi-StreetCanyon", condition="NLOS")
entry.formula
entry.shadow_fading_std_dbDon't know the section or parameter names? The API is self-describing, and the
install adds a tr-api command:
tr-api list
tr-api describe tr38901 7.9
tr-api dump tr38901 7.5 channel_model_parameters --format csv > lsp.csvSee INDEX.md for the section-by-section status table and
tools/tr_api/README.md for the full API and CLI reference.
Verification
Values marked verified were cross-checked against multiple independent source
formats — docx structure, a rendered-PDF visual read, and the HTML export's
equation text — rather than batch-converted. Where a source document renders its
equations as images and no automated cross-check is possible, the content is
PDF-visual-verified and recorded as single-source rather than overstated. CSV↔YAML
consistency, schema validity and formula cross-checks are enforced in CI.
Scope
Coverage grows opportunistically as research needs demand, not front-to-back
through each document — depth and verification over breadth. Additional sections
and TRs will follow in later releases.
License
MIT. The 3GPP source documents are not redistributed; this repository contains
hand-verified structured extracts only, with links to the canonical documents on
3gpp.org.