CARCELL

Abstract

CARCELL presents Luxembourg's electronic network coverage data — mobile (2G, 4G, 5G) and fixed broadband (VHCN, DOCSIS, Fibre) — in a structured, searchable format. The project aggregates publicly available regulatory data from the Institut Luxembourgeois de Régulation (ILR) to provide a clear picture of network infrastructure across all 100 communes. It is aimed at anyone interested in Luxembourg's digital infrastructure coverage without navigating raw regulatory datasets.

Data Sources

Data freshness

ILR coverage data is published semiannually. The current dataset reflects Semester 2 of 2025 (measurement period: April 2025). The CACLR registry is updated continuously but downloaded as a point-in-time snapshot.

Known limitations

• Mobile coverage CSV files contain grid cell IDs (IdGrid) without geographic coordinates. The ILR grid geometry is not part of the public dataset, preventing map-based visualisation of mobile coverage.
• Speeds are theoretical maximums per operator per grid cell — not measured throughput.
• Broadband percentages represent household coverage, not actual subscriptions.
• CACLR data is archived but not currently displayed in the interface.

Methodology

1. Mobile coverage aggregation

Three CSV files (2G, 4G, 5G) are analysed offline. Each file contains one row per grid cell per operator, with the maximum theoretical speed for that cell. The grid has approximately 259,000 cells covering Luxembourg at ~100m × 100m resolution. Statistics are computed by counting cells where 1, 2, or 3 operators provide coverage, and by grouping cells by speed tier. These aggregates are presented as summary statistics in the page.

Periode;Technologie;IdGrid;Maxvitesse;operateur
202504;4G;8616;30;Proximus Luxembourg S.A.,
202504;5G;9518;300;Orange Communications Luxembourg S.A., Proximus Luxembourg S.A.,

2. Broadband coverage extraction

The broadband CSV contains one row per commune (100 total) with VHCN, DOCSIS, and Fibre coverage as percentages. These values use decimal commas and percent signs in the source (European format). The data is parsed, converted to numeric values, and embedded directly into the page as a JavaScript array literal. This avoids the need for runtime CSV parsing.

// Embedded data structure (100 communes)
var DATA = [
  ["Beaufort", 58.9, 0.0, 58.9],
  ["Luxembourg", 97.2, 94.6, 87.2],
  // ... 98 more
];

3. Tabular display with sort and search

The broadband table renders from the embedded array with three interactive features: column sorting (click any header to toggle ascending/descending), free-text search (case-insensitive substring on commune name), and numeric formatting (one decimal place). Sort direction is indicated with triangle arrows. The implementation is ~60 lines of vanilla JavaScript with no dependencies.

4. Decision: table over map

A map-based visualisation was considered but rejected for two reasons. First, the mobile coverage grid cells lack public geographic coordinates, making spatial display impossible without reverse-engineering the ILR grid. Second, the broadband data is at commune level (100 records) — a sortable table with search is more practical for comparison than a choropleth for this dataset size. A choropleth of broadband coverage could be added if commune boundary GeoJSON is integrated.

5. CACLR registry archival

The CACLR (Registre national des localités et des rues) is Luxembourg's official address registry. It uses a fixed-width format with 13 tables covering districts, cantons, communes, localities, streets, buildings, and postal codes. The data is archived in data/caclr/ for potential future use (e.g. street-level analysis or geographic enrichment) but is not displayed in the current interface.

// CACLR fixed-width format (LOCALITE table)
00001Luxembourg                              LUXEMBOURG
00002Esch-sur-Alzette                        ESCH-SUR-ALZETTE

Stack & Tooling

• Languages: HTML5, CSS3 (custom properties, grid), vanilla JavaScript
• Libraries: none — zero external dependencies
• Fonts: Barlow Condensed (condensed sans-serif), JetBrains Mono (monospace) self-hosted WOFF2
• Data: embedded in page (broadband) + archived CSV and fixed-width files (mobile, CACLR)
• Hosting: static file server
• Build tools: none

Key Observations & Limitations

What works

• 2G coverage is near-universal: 96.9% of grid cells have all 3 operators.
• 5G is still emerging: only 14.9% of cells have 3-operator coverage, with speeds ranging from 30 to 1000 Mbps.
• 7 communes have zero DOCSIS coverage (likely no cable infrastructure), all in northern/rural Luxembourg.
• Embedding 100 commune records directly in the page eliminates all network latency for the interactive table.

Known gaps

• No geographic visualisation of mobile coverage due to missing grid coordinates.
• No historical comparison — only the current semester is displayed.
• Operator-level breakdown is available in the raw data but not exposed in the interface.
• No correlation analysis between mobile and broadband coverage.

Possible future directions

• Semester-over-semester trend comparison when additional data is published.
• Operator-level filtering for mobile coverage statistics.
• Integration with commune GeoJSON for a broadband choropleth map.
• Cross-referencing with CACLR for locality-level analysis.