Methodology

1 · Measurement environment

Speaker measurements use gated sine sweeps in a quiet, lightly damped room sized to suppress reflections within the measurement window. Below the gate limit, low-frequency data is spliced from ground-plane measurements. Electrical measurements (amplifiers, sources) are taken at the bench with resistive loads and a calibrated analyser loop. Temperature and humidity are logged with every session because air absorption above 10 kHz is not a myth.

2 · Equipment

Each archive lists its own equipment, but the shared backbone is:

• A calibrated free-field measurement microphone with a current calibration file.
• An audio interface with loopback reference for amplitude calibration.
• A resistive load bank (non-inductive) for amplifier tests.
• Calibrated microphone stands and laser-verified distances.

When equipment changes, archives taken with the previous setup keep their original equipment list and are re-measured as capacity allows.

3 · Test flow

• Calibration first. Microphone and loopback calibration files are captured the same session as the data.
• Averaged sweeps. Multiple sine sweeps averaged to push noise below the signal of interest.
• Fixed geometry. Distance, height and angle are recorded to the centimetre and reproduced for comparisons.
• Raw files kept. Processed graphs always derive from stored raw captures; nothing is smoothed into existence.

4 · Smoothing

Display smoothing is stated on every chart — typically 1/12 octave for broadband views. Raw unsmoothed data is always available in the downloadable CSVs. Narrow-band anomalies wider than the smoothing window are reported even when display smoothing would hide them.

5 · Evidence levels

Every record carries exactly one of these labels:

• concept — direction and intent; no dataset.
• simulation — modelled behaviour (BEM, FEA, SPICE), not yet built or tested.
• sample — synthetic dataset representing the shape of future measurements. Banner + watermark + -sample filenames apply, and the page stays noindex.
• measured — captured in our lab with the equipment and flow above; page becomes indexable.
• independently-tested — reproduced by a third party we link to.

A record can only move forward through these levels, never silently backward; a downgrade is a published correction.

6 · Data versioning & revision policy

• Every dataset file is versioned by measurement date in its archive.
• Replacing a dataset never deletes the previous one — the archive gains a revision entry explaining what changed and why.
• Sample datasets are replaced wholesale when measured data arrives; the replacement is flagged in the archive header.
• Corrections of fact are published in the journal, not slipped in quietly.

7 · What we will not do

• No graphs without a stated environment, distance and smoothing.
• No conclusion language stronger than the evidence level allows.
• No synthetic data presented without its three markers.