Personal HRTF capture — one speaker, a rotating listener

Measure your own head-related transfer functions with blocked-canal microphones in both ears, one fixed loudspeaker and a chair you rotate about a marked head-centre point. A reference sweep with the microphones at the head-centre and your head absent cancels the speaker and microphones out of the result. Each ear pair is captured simultaneously and gated to about 5 ms, so the room drops out and interaural timing is preserved to microseconds. Finish with an optional headphone sweep to compare what the headphone delivers to your ears against your own frontal HRTF.

Each step folds away once it is done; click a heading to reopen it.

1. Microphones, speaker and headphone

Binaural input — a two-channel interface with both blocked-canal microphones (one clock: both ears must be sampled together) Loudspeaker output Headphone output (for the optional comparison step)

Ear mapping Speaker plays through

Audio devices have not been scanned.

One sample is 20.8 µs at 48 kHz, 10.4 µs at 96 kHz. Interaural timing is refined below one sample, but 96 kHz or higher on the interface is the comfortable choice. The rate in use is reported after the first capture.

2. Geometry and headroom

Listener name / session Speaker to head-centre distance (m) Ear height above floor (m) Ceiling height (m) Nearest wall or large object (m) Head-related window (ms) Reference repeats Repeats at 0° (repeatability check)

Sweep duration (s) Output level (dBFS) Start frequency (Hz) End frequency (Hz) Wait before sweep (ms) Hold HRTF constant below (Hz)

Headroom: pinna resonances can sit 20 dB above the reference, so the reference sweep should peak near −25 to −30 dBFS. Set the interface gain during the reference and leave it alone for every ear capture.

3. Measurements

Graph

AngleEarSmoothingFrequency rangedB span Show repeats

4. Export

Load a project (.json)

The personal frontal target is the average of both ears at 0°, 1/6-octave smoothed: the curve a headphone should produce at your blocked canal to sound like the frontal speaker. Drop it into any EQ tool as a target.
The headphone correction is frontal HRTF minus the headphone's response at your ears, smoothed per region as recommended (tight through the ear gain, broad above 4 kHz).
HRIR pairs are gated stereo impulse responses (32-bit float WAV, left ear in the left channel) per angle, at the capture sample rate, for convolution engines.
Nothing is kept in the browser between visits: export the project before leaving, and load it again to continue — its captures are added to this session's, so a second sitting can fill in angles or add the headphone step.