How this dataset is built
EMF information fails in a predictable way: numbers circulate without their unit, their averaging time, or the difference between a legal limit and someone's precautionary target. This dataset exists to make that context inseparable from the number.
Editorial rules
- A value without a source is not published. Every record lists the primary documents it rests on and the date each link was last checked. Where a value is an estimate, the formula and its assumptions are published with it.
- Regulatory, guideline and precautionary values are labelled and never merged. ICNIRP reference levels, FCC rules, Swiss installation values and building-biology grading bands answer different questions; presenting them as one scale is the single most common distortion in this field.
- Association is not causation. Where epidemiology finds an association — as pooled analyses do for childhood leukaemia above roughly 0.3–0.4 µT — the record says association, states the study design, and states what the design cannot establish.
- Uncertainty is published, not resolved. Where the literature disagrees, both findings are recorded with their weight in the evidence base. Nothing here is written to make either camp comfortable.
- No health claims, no product endorsements. Instruments and shielding materials are described as classes, with what each cannot do and how each fails in practice. Named models illustrate a class; they are not recommendations.
- Provisional records are marked. A record whose values are not yet fully cited is labelled provisional, excluded from the sitemap, and ranked below verified records.
One registry, many representations
Records are authored once as typed data. The HTML pages, Markdown twins, JSON API, CSV exports, JSON-LD, OpenAPI description and MCP tools are generated from that same registry, so a correction applied once propagates to every representation. That is also why the machine formats are not an afterthought: they are the same content, not a summary of it. See API, MCP and bulk data.
- Standards & frameworks — 17 records
- Exposure limits — 34 records
- Quantities & units — 7 records
- Sources & emitters — 28 records
- Instruments — 12 records
- Shielding & remediation — 10 records
- Research record — 10 records
- Organizations — 17 records
- Measurement protocols — 7 records
- Glossary — 27 records
- Device SAR — 56 records
- Direct answers — 30 records
Sources
56 sources are registered, weighted towards primary documents: 22 government, 14 regulation, 9 peer-reviewed, 4 guideline, 4 ngo, 3 standard. Regulations and standards are cited in their own words; secondary summaries are used only for context and are labelled as such.
Known limitations
- Computed field strengths are free-space main-beam upper bounds. They ignore antenna pattern, downtilt, absorption and reflection, and are invalid within roughly one wavelength of an antenna.
- Emitters are described as classes. A specific model's transmit power, duty cycle and antenna pattern can differ by an order of magnitude from the class figure.
- Regulatory texts are amended. Each record shows when its citation was last checked; the primary document remains authoritative.
- Coverage is currently strongest for the United States, the EU, Switzerland and international guidelines. Other national limits are being added.
- Practitioner listings are not published until a verification procedure exists that does not amount to repeating self-description.
Corrections
Send corrections to data@emfbase.com with the record id and the primary document. Because every value names its source, a correction can be checked rather than debated. Data is licensed CC BY 4.0; attribute as EMFBase (https://emfbase.com), dataset version 2026.08.1.
This is reference data, not medical or legal advice. If you have health concerns, consult a clinician; for a compliance determination, consult the cited regulation and a qualified professional.