Research & detection

The goal: from testimony to measurement

Personal reports are where every investigation of the Hum begins — but they cannot end there. The question "what is making this sound?" will only be answered by instruments: recordings that show the tone on a screen, timestamps that can be compared between cities, spectrograms that prove the signal exists outside anyone's head.

This page collects what the community has learned about detecting low-frequency sound and infrasound, and describes the portal's flagship project: a network of synchronized measuring stations for locating the source by triangulation. If you have built something that works, share it in the Detection & hardware category — proven, reproducible setups get added here.

Why your phone (mostly) fails

A phone microphone is designed for speech. Below roughly 80–100 Hz its sensitivity collapses and its own electronic noise dominates — exactly the band where the Hum lives. Phone apps (Spectroid on Android, or any FFT analyzer) are still worth trying as a first look: if a stable spectral line around 30–80 Hz appears and disappears in sync with what you hear, that is already valuable evidence. But a silent spectrogram from a phone proves nothing. For real work you need better sensors.

Sensors that work in the Hum band

Low-noise measurement microphones. The key spec is equivalent self-noise at low frequencies and a frequency response reaching well below 20 Hz. Community favourites include electret capsules such as the Primo EM172/EM272, popular among nature recordists precisely for their very low self-noise, connected to a recorder that can disable its high-pass filter. MEMS measurement microphones (e.g. Infineon's low-noise IM7x series) are a compact alternative.

Geophones. A geophone (e.g. the ubiquitous SM-24 element, ~10 Hz natural frequency) senses ground vibration rather than air pressure. This distinction is powerful: if a geophone buried in the garden shows the same pattern as what you hear, the phenomenon has a seismic/structural component; if the microphone sees it and the geophone does not, it is airborne. Running both simultaneously is the single most informative experiment an individual can do.

Infrasound sensors and citizen seismographs. The Raspberry Shake & Boom combines a geophone with an infrasound pressure sensor on a Raspberry Pi, uploads data to a public worldwide network, and is the closest off-the-shelf product to what this project needs. Differential-pressure infrasound microphones (as used for monitoring volcanoes and wind farms) cover the band below 20 Hz.

The recording chain. Any decent USB audio interface sampling at 44.1 kHz or more is adequate — low frequencies are trivial for modern converters. What matters: DC-coupled or low cut-off input, gain set so the noise floor of the sensor, not the interface, dominates, and long recordings (whole nights), because the Hum's irregular breaks mean short samples miss it.

Analyzing what you captured

1. Spectrogram first. Audacity (free) or Raven Lite: window length 8–32 s for good low-frequency resolution. Look for a persistent narrow line between 30 and 80 Hz.
2. Correlate with your ears. Keep a simple log while recording: note times when you hear the tone and when it breaks off. The founder of this portal describes the pattern to look for: a steady deep tone with irregular, unsynchronized interruptions — sometimes minutes of droning, sometimes a short burst. If your logged on/off times match line segments in the spectrogram, you have evidence.
3. Rule out the house. Repeat with the main breaker off (battery-powered recorder), and outdoors far from buildings. A signal that survives both is not your refrigerator.
4. Compare locations. The same night, same hardware, different town — differences in level or presence are data.

The flagship: synchronized triangulation network

One station proves a signal exists. Several stations with synchronized clocks can locate its source. The principle — time difference of arrival (TDOA) — is how seismologists pinpoint earthquakes and how infrasound arrays detect distant explosions:

1. Identical stations (sensor + single-board computer) record continuously.
2. Every sample is timestamped via NTP, or better GPS-disciplined time (millisecond precision or better; sound travels ~343 m per ms in air, ~3–6 km/s in ground).
3. When the same signal appears at multiple stations, cross-correlation yields the relative delays.
4. Delays from three and more stations constrain the direction and distance of the source.

Stations in different countries — Slovakia, Czechia, Hungary, Greece and beyond — would answer the biggest open question directly: is the Hum one source, several regional sources, or something that has no propagating source at all? Every outcome, including the null result, is scientifically valuable.

This is what the donations fund: sensors, boards, enclosures and the time-sync hardware for stations hosted by volunteer members. All designs, configurations and collected data will be published openly on this portal.

Contribute

• You have recordings or a working setup? Post the details — sensor model, interface, settings, spectrograms — in the community. Reproducibility is everything.
• You can host a station? A quiet location, a few watts of power and a network connection is all a station needs. Register and mention it in your profile or a post.
• You have skills? Signal processing, electronics, embedded Linux — the analysis pipeline is being built in the open.

The Hum has survived fifty years of being dismissed. It will not survive a network of synchronized microphones.