Skip to content

Python tool for aircraft Doppler audio geolocation using historical ADS-B flight data (Source-available)

Document DLR real-flight scanner miss (#204)

Cinna Davis9c043753fcd0
diff --git a/benchmarks/dlr-record139-scanner-check-20261103.md b/benchmarks/dlr-record139-scanner-check-20261103.md
new file mode 100644
index 0000000..b278aac
--- /dev/null
+++ b/benchmarks/dlr-record139-scanner-check-20261103.md
@@ -0,0 +1,49 @@
+# DLR RECORD 139 real-flight scanner check
+
+## Source and scope
+
+- Source: [RECORD 139, Benchmark for Acoustic Source Localization of Flyover Microphone Array Measurements](https://zenodo.org/records/20494053), by DLR researchers.
+- The publisher describes one 2016 flyover recording of the DLR ATRA Airbus A320-232, with calibrated pressure signals from a 231-microphone array and an 88 m/s aircraft velocity. The downloaded HDF5 records 48,192 Hz audio and trajectory references in sample-relative local coordinates.
+- License: CC BY 4.0, verified from the Zenodo record API metadata. The raw 311,789,336-byte HDF5 and the derived WAV are kept under `/home/cinna/.local/state/dopplergeist-dlr-record139-20261103/`; neither is committed here.
+- This was scanner-only. No flight matching, aircraft association or geolocation ran.
+
+## Scanned excerpt
+
+The raw HDF5 member `REC139.h5` has MD5 `1fb08d67965550d26d5b35abaf71f3f3` and SHA-256 `826fb22c2daa5c09ea378f4f1104428ab064c00373cff94fa4fc957822084fa4`. It contains 231 channels by 337,344 samples, or seven seconds. The scanned WAV is channel 80, the microphone nearest the array origin: `[0.125, -0.043, -0.079]` m in the array's local coordinate frame. A single constant gain of `0.3188036516506252` scales the channel peak to 0.95; this does not alter its time or frequency structure. The resulting float WAV SHA-256 is `469c4257f4d230ac02adbc435ebdeae457dfefe4ceae3275a733b0543d4d7ae1`.
+
+The HDF5 includes three trajectory reference groups aligned to sample indices, but the inspected file has no absolute UTC timestamp or geodetic microphone coordinate. This prevents using this recording as a standard location-recovery benchmark without outside source evidence.
+
+## Scanner result
+
+The current-main scanner at commit `4dadf3903551a82d526163157c300b35814e1b5b` returned **no proposals**: zero ridges and zero harmonic groups. The scanner reported 4.278 seconds of runtime. This records a real publisher-identified aircraft excerpt that the current scanner did not detect; it does not establish absence of Doppler structure in other microphones or in the source's full data.
+
+Recreate the WAV from the downloaded HDF5 with `h5py`, NumPy and SciPy:
+
+```python
+import h5py
+import numpy as np
+from scipy.io import wavfile
+
+with h5py.File("REC139.h5", "r") as source:
+ positions = source["MetaData/ArrayAttributes/microphonePositionsM"][:]
+ channel = int(np.argmin(np.linalg.norm(positions, axis=1)))
+ pressure = source["MicrophoneData/microphoneDataPa"]
+ sample_rate = int(round(float(pressure.attrs["sampleRateHz"])))
+ samples = pressure[channel, :].astype(np.float32)
+ samples *= 0.95 / np.max(np.abs(samples))
+ wavfile.write("REC139-center-mic.wav", sample_rate, samples)
+```
+
+Reproduce the scanner stage after placing that WAV at the state path above:
+
+```bash
+PYTHONPATH=src .venv/bin/python scripts/collect_audio_corpus.py scan-one \
+ /home/cinna/.local/state/dopplergeist-dlr-record139-20261103/REC139-center-mic.wav \
+ --output /home/cinna/.local/state/dopplergeist-dlr-record139-20261103/scan.json
+```
+
+The conversion selects the microphone nearest `[0, 0, 0]` in
+`/MetaData/ArrayAttributes/microphonePositionsM`, reads that row from
+`/MicrophoneData/microphoneDataPa`, and writes the constant-gain-scaled float
+samples at the source sample rate. The upstream Zenodo record and its file
+checksums identify the source; the converted audio is not included in the repo.
diff --git a/docs/REAL_AUDIO_CORPUS.md b/docs/REAL_AUDIO_CORPUS.md
index 0d9bf43..83cce24 100644
--- a/docs/REAL_AUDIO_CORPUS.md
+++ b/docs/REAL_AUDIO_CORPUS.md
@@ -55,6 +55,18 @@ CC BY-NC 4.0; no media is copied into this repository. Its public coordinates
refer to nearby public places rather than exact microphones, so these clips do
not establish location truth for geolocation evaluation.
+## Publisher-documented aircraft scanner control
+
+The [DLR RECORD 139 scanner check](../benchmarks/dlr-record139-scanner-check-20261103.md)
+uses one center-microphone channel from a real, publisher-identified A320 flyover
+recording. The automatic scanner found no ridge proposals in this seven-second
+excerpt. That is a scanner miss on this excerpt, not proof that no Doppler
+structure exists in the full array recording. The source contains microphone and
+aircraft geometry only in a relative coordinate frame; the inspected HDF5 does
+not provide geodetic microphone coordinates or an absolute UTC clock. No flight
+matching or geolocation was run, and this control does not count toward the
+20-known plus five-unknown location benchmark target.
+
## Candidate index
The [25-row candidate index](../benchmarks/audio-candidate-index-20261103.csv)