Download the whole database

Everything behind this site is available as a file. 402 headphones and 156 amplifiers, dongles, players and receivers, in CSV and JSON, free to use with attribution.

The reason this is worth downloading rather than scraping is the provenance columns. Every sensitivity figure carries the value and unit the manufacturer actually printed alongside the converted one, and a basis column saying which of the two you are looking at. Every amplifier power figure carries a basis per load: published at that exact load, derived from neighbouring ones, an upper bound from a voltage rating, or not yet traced to a document. No other dataset of this kind tells you that, because assembling it means admitting how much of the manufacturer data is incomplete.

What the basis columns mean

For headphones, basis takes one of these values:

Value ModelsMeaning
published-per-mW173 the maker printed this figure per milliwatt; nothing was done to it
converted-from-per-volt82 the maker printed it per volt; converted with the published impedance
unit-not-stated-by-maker99 the maker printed a bare decibel figure with no reference; the value is as printed and may be either reading
not-published-by-maker47 no sensitivity figure exists in the maker's documentation; the value carried over from the material this database was first built from

For sources, each load has its own basis: published, interpolated, extrapolated, source-impedance, voltage-bound or unverified. Across the four standard loads that works out to 189 published figures and 391 derived ones. The method page explains each.

Licence and citation

Creative Commons Attribution 4.0. Use it in a spreadsheet, a calculator, an app, a video, a paper — commercial use included. The only condition is that you say where it came from.

The attribution line to use:
Data: The Phone Path (thephonepath.com), CC BY 4.0

Two honest warnings before you build on it. First, these are manufacturer claims, not measurements — a maker's own figure is a claim, and independent measurement is better than any of this. Second, the figures marked unit-not-stated-by-maker are genuinely ambiguous: if your use is sensitive to a 15 dB error, filter them out rather than trusting them.

If you find a figure that disagrees with a manufacturer document, the document wins and I want to know — the address is in the footer. The files are rebuilt whenever the database changes.