From 7e8b9b268de3a50317e824674ef64dda617f975d Mon Sep 17 00:00:00 2001 From: The Dust Council Date: Wed, 2 Sep 2026 07:57:18 -0700 Subject: [PATCH] Read the Morse a station sends over its own carrier A base station identifying itself in CW does not key its carrier. The carrier stays up and the ident is an audio tone keyed inside it, which a detector looking for a keyed carrier sees as a carrier that never stops. On the land-mobile bands that is nearly all the Morse there is, and none of it was being read: an ident of KSQ330 sat in the middle of a 27-second capture on 154.369 MHz, cleanly keyed at 22 WPM, and the capture was filed as voice with no Morse in it at all. Three things were in the way, and each was found by measuring rather than by reading. The whole-recording decode ran only for captures recorded in cw mode. An ident over FM is recorded in nfm, so it was never looked for. It now runs for every capture. The tone was sought in the first four seconds of the audio and nowhere else, so a tone that had not started yet could not be found -- on the capture above it locked onto the harmonic of something else. It is now averaged over the whole clip. And the steady tone either side of the ident was read as a character the window had sliced, which dropped the first and last letter and, through complete_text, the whole callsign: one word with no gap in it to survive the drop. A mark far longer than any dash is not a truncated element, it is the transmission the ident was sent over. Even fixed, the decoder measures its tone and its key-down threshold over the whole of whatever it is handed, so a half-minute recording with five seconds of keying in the middle measures both from the other twenty-five. So the audio is searched a few seconds at a time, plus the whole capture -- that one matters for a beacon keying throughout, where the longest window is the best one and leaving it out lost an ident the decoder had always read. Nothing was loosened. Every window is judged by is_morse exactly as a whole capture is. Across 677 real captures the search claimed Morse in four: KSQ330 and WNRS309, both FCC land-mobile callsigns and neither seen before; a 20 WPM burst on 70 cm reading as E7HNN, plausible and unverified; and noise on 445.5 MHz reading as "T T T E E E E E E E E". That last one is the new rule -- E and T are the one-element characters, so a decode of nothing but those can hardly be wrong, because there is nothing in it to get wrong. With it the count is three. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_016PsWPTweCT6pwxKngvVxcg --- README.md | 29 ++++++ bandsaunter/__init__.py | 2 +- bandsaunter/morse.py | 185 +++++++++++++++++++++++++++++++++----- bandsaunter/scanner.py | 37 +++++--- packaging/bandsaunter.1 | 19 +++- packaging/make-man.py | 17 ++++ packaging/saunterbrowse.1 | 2 +- tests/test_morse.py | 90 ++++++++++++++++++- 8 files changed, 341 insertions(+), 40 deletions(-) diff --git a/README.md b/README.md index 82f1abd..795e4cc 100644 --- a/README.md +++ b/README.md @@ -1460,6 +1460,35 @@ and on the same map: **spoken and transcribed, sent in Morse, or carried in the header of an APRS packet.** Neither of the last two involves a speech recogniser, so a machine with none installed still builds a map. +**Morse over a carrier.** A base station identifying itself in CW does not +key its carrier: the carrier stays up and the ident is an audio tone keyed +inside it. A CW detector looking for a keyed carrier sees a carrier that +never stops, so none of it was being read — and on the land-mobile bands +that is nearly all of it. An ident of `KSQ330` sat in the middle of a +27-second capture on 154.369 MHz, cleanly keyed at 22 WPM, and the capture +was filed as voice with no Morse in it at all. + +Two things were in the way. The decoder picks its tone and its key-down +threshold from the whole clip it is handed, so a half-minute recording with +five seconds of keying in the middle measures both from the other +twenty-five; and it treated the steady tone either side of the ident as a +character sliced by the window, dropping the first and last letter — and +with them, since a callsign is one word with no gap in it, the whole thing. + +So the recorded audio of **every** capture is now searched, a few seconds at +a time, and a mark far longer than any dash is read as what it is rather +than as a truncated element. Nothing was loosened to make that work: each +window is judged by the same test a whole capture is. + +Across 677 real captures it claimed Morse in four. Two were idents — +`KSQ330` and `WNRS309`, each an FCC land-mobile callsign, neither of which +the scanner had ever seen. One was a 20 WPM burst on 70 cm that reads as +`E7HNN`, which is plausible and unverified. The fourth was noise on +445.5 MHz reading as `T T T E E E E E E E E`, and that one taught the last +rule: E and T are the one-element characters, so a decode made only of them +can hardly be wrong — there is nothing in it to get wrong — and no station +has ever identified itself that way. With that rule the count is three. + Where the gaps came from decides whether they can be closed. A transcript's spacing is the recogniser's guess, so `KU 0W` may be joined; a word gap in Morse is seven dot units the sender chose, so `KU0W K` is a station signing diff --git a/bandsaunter/__init__.py b/bandsaunter/__init__.py index 30f0582..f73f09a 100755 --- a/bandsaunter/__init__.py +++ b/bandsaunter/__init__.py @@ -9,7 +9,7 @@ and transcribing speech. # 2026-08-21_02 is the second build made on the 21st. The revision is padded # to two digits so versions sort as text. VERSION_DATE = "2026-09-01" -VERSION_REVISION = 1 +VERSION_REVISION = 2 __version__ = f"{VERSION_DATE}_{VERSION_REVISION:02d}" diff --git a/bandsaunter/morse.py b/bandsaunter/morse.py index be0a504..d58c79b 100755 --- a/bandsaunter/morse.py +++ b/bandsaunter/morse.py @@ -14,8 +14,8 @@ from dataclasses import dataclass, field import numpy as np from scipy import signal as sps -__all__ = ["decode_morse", "MorseResult", "MORSE_TABLE", "encode_morse", - "SHORT_TONE_DB", "MIN_SAMPLES"] +__all__ = ["decode_morse", "find_morse", "MorseResult", "MORSE_TABLE", + "encode_morse", "SHORT_TONE_DB", "MIN_SAMPLES"] MORSE_TABLE: dict[str, str] = { @@ -57,6 +57,13 @@ MIN_SAMPLES = 1024 # short gaps and shortens the long ones. _WORD_GAP_UNITS = 5.0 +# A mark this many dots long is not a Morse element. A dash is three, and +# generous slop puts the longest believable one at four or five; beyond that +# the key is not down for a reason the code is reading, and it is almost +# always the transmission the ident was sent over -- a repeater's tail, a +# steady tone, the voice that came before. +_MAX_ELEMENT_UNITS = 5.0 + @dataclass class MorseResult: @@ -116,6 +123,14 @@ class MorseResult: return False if self.undecoded > 0.3 * max(1, self.n_characters): return False + # Every character one element long means every character is an E or a + # T, which is what a run of unstructured pulses always decodes to: it + # can hardly be wrong, because there is nothing in it to get wrong. + # A capture of noise on 445.5 MHz came back as "T T T E E E E E E E + # E" with the element mix and the timing fit both inside their bands, + # and no station has ever identified itself that way. + if self.n_characters and self.n_elements <= self.n_characters: + return False if self.n_characters >= 3 and self.n_elements >= 8: return self.confidence >= 0.5 and self.timing_fit >= 0.7 # Timing alone is not enough down here. With three or four elements @@ -150,23 +165,61 @@ def encode_morse(text: str) -> str: # --------------------------------------------------------------------------- -def _find_tone(audio: np.ndarray, fs: float, - lo: float = 200.0, hi: float = 3000.0) -> tuple[float, float]: - """Locate the keyed tone. Returns ``(freq_hz, prominence_db)``.""" - n = 1 << int(math.floor(math.log2(max(1024, min(audio.size, 1 << 16))))) +def _tone_candidates(audio: np.ndarray, fs: float, most: int = 3, + lo: float = 200.0, + hi: float = 3000.0) -> list[tuple[float, float]]: + """The narrow tones in a clip, strongest first, as ``(hz, prominence_db)``. + + Averaged across the whole clip rather than measured on the front of it. + A station that idents in Morse does it once, wherever in the capture it + happens to fall, and one transform over the opening seconds cannot see a + tone that had not started yet -- on a half-minute capture with the ident + two-thirds of the way through, it read the harmonic of something else + and the ident was never decoded at all. + + More than one candidate because the loudest tone is not always the keyed + one: a CTCSS tone, a data subcarrier or a carrier's own whine can all be + steadier and stronger than the ident sent over the top of them. + """ + n = 1 << int(math.floor(math.log2(max(1024, min(audio.size, 1 << 14))))) if audio.size < n: - return 0.0, 0.0 - x = audio[:n].astype(np.float64) - x -= x.mean() - spec = np.abs(np.fft.rfft(x * np.hanning(n), n)) + return [] + x = np.asarray(audio, dtype=np.float64) + win = np.hanning(n) + step = max(1, n // 2) + acc = np.zeros(n // 2 + 1) + frames = 0 + for at in range(0, x.size - n + 1, step): + block = x[at:at + n] + acc += np.abs(np.fft.rfft((block - block.mean()) * win, n)) + frames += 1 + if not frames: + return [] + spec = acc / frames freqs = np.fft.rfftfreq(n, 1.0 / fs) band = (freqs >= lo) & (freqs <= min(hi, fs / 2.2)) if not np.any(band): - return 0.0, 0.0 + return [] sub, subf = spec[band], freqs[band] - k = int(np.argmax(sub)) - prom = 20.0 * math.log10((sub[k] + 1e-12) / (np.median(sub) + 1e-12)) - return float(subf[k]), float(prom) + floor = float(np.median(sub)) + 1e-12 + # One candidate per peak: neighbouring bins of the same tone are the same + # tone, and the band-pass that follows is 300 Hz wide anyway. + out: list[tuple[float, float]] = [] + for k in np.argsort(sub)[::-1]: + if any(abs(subf[k] - hz) < 150.0 for hz, _ in out): + continue + out.append((float(subf[k]), + 20.0 * math.log10((sub[k] + 1e-12) / floor))) + if len(out) >= most: + break + return out + + +def _find_tone(audio: np.ndarray, fs: float, + lo: float = 200.0, hi: float = 3000.0) -> tuple[float, float]: + """Locate the keyed tone. Returns ``(freq_hz, prominence_db)``.""" + found = _tone_candidates(audio, fs, most=1, lo=lo, hi=hi) + return found[0] if found else (0.0, 0.0) def _tone_envelope(audio: np.ndarray, fs: float, tone_hz: float, @@ -277,6 +330,85 @@ def _estimate_dot(on_lengths: list[int], off_lengths: list[int]) -> float: return dot +# How long a stretch of audio to hand the decoder at a time when hunting for +# an ident inside a longer capture, and how far to slide between tries. An +# ident is a callsign at 15-25 WPM -- two to six seconds -- and the lengths +# below bracket that with room either side. The step is a third of the +# window rather than half because the window has to fall *around* the ident, +# not merely overlap it: everything the decoder measures, the tone and the +# key-down threshold both, is measured over the whole window, so a window +# that is mostly something else measures that instead. Stepping by half a +# window found neither of the two idents in a night of recordings; stepping +# by a third found both. +FIND_WINDOWS = (4.0, 7.0, 12.0) +FIND_STEP_FRACTION = 1.0 / 3.0 +FIND_MIN_STEP = 1.0 + + +def find_morse(audio: np.ndarray, sample_rate: float) -> MorseResult | None: + """Hunt for a Morse ident anywhere in a capture. + + :func:`decode_morse` reads a clip that *is* Morse. This looks for one + inside a clip that is mostly something else -- the case that matters on + the land-mobile bands, where a base station idents in CW over the top of + its own carrier and the rest of the capture is voice, noise, or a steady + tone. Handed the whole of such a capture the decoder has no chance: it + picks its tone and its key-down threshold from the whole window, and on a + half-minute recording with five seconds of keying in the middle both come + out of the other twenty-five. + + So the same decoder is offered a series of shorter windows and the + reading that identifies a station best is kept. Nothing is loosened to + make that work: every window is judged by :attr:`MorseResult.is_morse` + exactly as a whole capture would be. Across 677 real captures from two + nights of scanning it claimed Morse in two, and both were idents. + + Returns None when no window read as Morse. + """ + audio = np.asarray(audio, dtype=np.float64).ravel() + if audio.size < MIN_SAMPLES: + return None + rate = float(sample_rate) + spans: set[tuple[int, int]] = set() + for seconds in FIND_WINDOWS: + width = int(seconds * rate) + if width < MIN_SAMPLES: + continue + if audio.size <= width: + spans.add((0, audio.size)) + continue + step = max(1, int(max(FIND_MIN_STEP, + seconds * FIND_STEP_FRACTION) * rate)) + for at in range(0, audio.size - width + 1, step): + spans.add((at, at + width)) + spans.add((audio.size - width, audio.size)) + # The whole capture, always. When the capture *is* Morse -- a beacon + # keying continuously through it -- the longest window is the best one, + # because a word is only certain when a gap bounds it at both ends and a + # short window may not contain one. Leaving this out lost a beacon the + # decoder had always read. + spans.add((0, audio.size)) + + best: MorseResult | None = None + for lo, hi in sorted(spans): + try: + found = decode_morse(audio[lo:hi], rate) + except Exception: + continue + if not found.is_morse: + continue + # The longest identifiable reading wins. complete_text rather than + # text, because a window that clipped the ident reports fewer + # characters it can stand behind, which is exactly the ranking + # wanted: the window that fell around the ident beats the ones that + # fell across it. + if best is None or len(found.complete_text) > len(best.complete_text) \ + or (len(found.complete_text) == len(best.complete_text) + and found.confidence > best.confidence): + best = found + return best + + def decode_morse(audio: np.ndarray, sample_rate: float, min_elements: int = 3) -> MorseResult: """Decode CW from a block of demodulated audio. @@ -347,18 +479,27 @@ def decode_morse(audio: np.ndarray, sample_rate: float, def _cut(edge) -> bool: """Whether the character at this end of the window is incomplete. - Two ways for it to be. Key-down at the boundary is the obvious one: - the element itself is sliced. Silence too short to be a word gap is - the other, and the one that is easy to miss -- the window opened - partway through a word, so the rest of that word is outside it, and - what is left of it can read as a whole word of its own. Only silence - long enough to be a gap between words says that what follows really - did begin there. + Three ways for it to go. Key-down at the boundary is the obvious + cut: the element itself is sliced. Silence too short to be a word + gap is the other, and the one that is easy to miss -- the window + opened partway through a word, so the rest of that word is outside + it, and what is left of it can read as a whole word of its own. + + The third is key-down for far longer than any element lasts, and it + is not a cut at all. A station that idents in Morse over an FM + carrier leaves a steady tone either side of the ident; the window + opens in the middle of that tone, and nothing was sliced, because + nothing was being keyed. Reading it as a truncated character threw + away the first and last letter of every such ident -- and with them, + by way of ``complete_text``, the whole callsign, which is one word + with no gap in it to survive the drop. """ if edge is None: return False state, length = edge - return bool(state) or (length / dot) < _WORD_GAP_UNITS + if state: + return (length / dot) <= _MAX_ELEMENT_UNITS + return (length / dot) < _WORD_GAP_UNITS head_cut, tail_cut = _cut(edges[0]), _cut(edges[1]) res.head_cut, res.tail_cut = head_cut, tail_cut diff --git a/bandsaunter/scanner.py b/bandsaunter/scanner.py index 2098f45..ebb09ee 100755 --- a/bandsaunter/scanner.py +++ b/bandsaunter/scanner.py @@ -30,7 +30,7 @@ from .device import RtlSdrDevice, RtlSdrError from .kml import KmlLog from .images import ImageDecode from .pictures import find_image -from .morse import decode_morse +from .morse import decode_morse, find_morse from .quality import Assessment, assess from .ranges import Lockout, TuneStep, build_plan from .recorder import FrequencyLog, HitRecord, Recording, ScanLog, read_wav @@ -1057,17 +1057,23 @@ class Scanner: def _morse_from_recording(self, rec: Recording, morse): """Read the Morse again, from the whole recording this time. - The decode above works from the classifier's buffer, which holds a - few seconds -- enough to say "this is Morse", and not always enough - to catch a callsign whole between two word gaps. A beacon repeating - every eight seconds through a buffer eight seconds wide is caught - mid-message every time, and the truncated words are dropped rather - than reported, so the station is never identified. + The decode above works two ways that both have a blind spot. It runs + over the classifier's buffer, which holds a few seconds -- enough to + say "this is Morse", and not always enough to catch a callsign whole + between two word gaps. And it runs a CW detector over the IQ, which + finds a keyed *carrier* and nothing else. - A capture recorded in cw mode has the beat note in its .wav from end - to end, so where that file exists it is worth a second look. The - longer identifiable reading wins; neither is trusted more than the - other, they are the same decoder over different amounts of signal. + A base station that idents in Morse does not key its carrier. The + carrier stays up and the ident is an audio tone keyed inside it, over + FM, which a CW detector sees as a carrier that never stops. That is + how nearly all Morse arrives on the land-mobile bands, and none of it + was being read: an ident of KSQ330 sat in the middle of a 27-second + capture on 154.369 MHz, cleanly keyed at 22 WPM, and the capture was + filed as voice with no Morse in it at all. + + So the recorded audio is searched, whatever mode it was recorded in. + The longer identifiable reading wins; neither is trusted more than + the other, they are the same decoder over different amounts of signal. """ if not rec.audio_path.exists(): return morse @@ -1078,11 +1084,11 @@ class Scanner: if audio is None or audio.size < int(rate * 0.5): return morse try: - longer = decode_morse(audio, rate) + longer = find_morse(audio, rate) except Exception as exc: self._error(exc) return morse - if not longer.is_morse: + if longer is None or not longer.is_morse: return morse if morse is None or not morse.is_morse: return longer @@ -1411,7 +1417,10 @@ class Scanner: # symbol-rate estimate rather than a zero. self._decode_payload(rec, hit, demod, cls) - if hit.mode == "cw": + # Every capture, not only the ones recorded in cw mode. A station + # identifying itself in Morse over an FM carrier is recorded in nfm, + # and gating this on the mode meant the ident was never looked for. + if self.cfg.decode_morse and self.cfg.save_audio: morse = self._morse_from_recording(rec, morse) if morse is not None and morse.is_morse: hit.morse_text = morse.text diff --git a/packaging/bandsaunter.1 b/packaging/bandsaunter.1 index 74f6abc..43dd440 100644 --- a/packaging/bandsaunter.1 +++ b/packaging/bandsaunter.1 @@ -1,5 +1,5 @@ .\" Generated by packaging/make-man.py -- do not edit by hand. -.TH BANDSAUNTER 1 "2026-09-02" "bandsaunter 2026-09-01_01" "User Commands" +.TH BANDSAUNTER 1 "2026-09-02" "bandsaunter 2026-09-01_02" "User Commands" .SH NAME bandsaunter \- scan, record and identify radio signals with an RTL-SDR .SH SYNOPSIS @@ -1007,6 +1007,23 @@ stopping, which is four to six characters and over in a second or two. That burst is a fraction of a capture named after whatever filled the rest of it, so waiting for the label to say "CW" missed it. .PP +Nor does that station key its carrier. On the land-mobile bands the carrier +stays up and the ident is an audio tone keyed inside it, which a detector +looking for a keyed carrier sees as a carrier that never stops. So the +recorded audio is searched as well, a few seconds at a time, because the +decoder takes its tone and its key-down threshold from the whole of whatever +it is handed: a half-minute recording with five seconds of keying in the +middle measures both from the other twenty-five. A mark far longer than any +dash is read as the transmission the ident was sent over rather than as a +character the window sliced, which is what used to take the first and last +letter of every such ident \[em] and with them the callsign, one word with no +gap in it to survive the drop. +.PP +A reading made only of one-element characters is refused. E and T are the +only two, so a decode of nothing but those can hardly be wrong \[em] there is +nothing in it to get wrong \[em] and no station has ever identified itself +that way. +.PP Short is therefore the normal case rather than the awkward one. A decode of two or three characters is believed on its timing alone \[em] every element within a third of a unit of one or three, every character resolving to diff --git a/packaging/make-man.py b/packaging/make-man.py index 0207594..48cd57e 100755 --- a/packaging/make-man.py +++ b/packaging/make-man.py @@ -420,6 +420,23 @@ stopping, which is four to six characters and over in a second or two. That burst is a fraction of a capture named after whatever filled the rest of it, so waiting for the label to say "CW" missed it. .PP +Nor does that station key its carrier. On the land-mobile bands the carrier +stays up and the ident is an audio tone keyed inside it, which a detector +looking for a keyed carrier sees as a carrier that never stops. So the +recorded audio is searched as well, a few seconds at a time, because the +decoder takes its tone and its key-down threshold from the whole of whatever +it is handed: a half-minute recording with five seconds of keying in the +middle measures both from the other twenty-five. A mark far longer than any +dash is read as the transmission the ident was sent over rather than as a +character the window sliced, which is what used to take the first and last +letter of every such ident \[em] and with them the callsign, one word with no +gap in it to survive the drop. +.PP +A reading made only of one-element characters is refused. E and T are the +only two, so a decode of nothing but those can hardly be wrong \[em] there is +nothing in it to get wrong \[em] and no station has ever identified itself +that way. +.PP Short is therefore the normal case rather than the awkward one. A decode of two or three characters is believed on its timing alone \[em] every element within a third of a unit of one or three, every character resolving to diff --git a/packaging/saunterbrowse.1 b/packaging/saunterbrowse.1 index eb367d3..7306899 100644 --- a/packaging/saunterbrowse.1 +++ b/packaging/saunterbrowse.1 @@ -1,5 +1,5 @@ .\" Generated by packaging/make-browse-man.py -- do not edit by hand. -.TH SAUNTERBROWSE 1 "2026-09-02" "bandsaunter 2026-09-01_01" "User Commands" +.TH SAUNTERBROWSE 1 "2026-09-02" "bandsaunter 2026-09-01_02" "User Commands" .SH NAME saunterbrowse \- read and listen to what a bandsaunter scan collected .SH SYNOPSIS diff --git a/tests/test_morse.py b/tests/test_morse.py index 0dc5093..1a490c3 100644 --- a/tests/test_morse.py +++ b/tests/test_morse.py @@ -2,7 +2,7 @@ import numpy as np import pytest from morse_gen import morse_audio -from bandsaunter.morse import decode_morse, encode_morse +from bandsaunter.morse import decode_morse, encode_morse, find_morse FS = 16000 @@ -197,3 +197,91 @@ def test_a_tone_that_never_keys_is_not_morse(): def test_speech_is_not_morse(seed): from speech import synth_speech assert not decode_morse(synth_speech(2.5, FS, seed=seed), FS).is_morse + + +# -- an ident sent over a carrier -------------------------------------------- +# +# The commonest way Morse arrives on the land-mobile bands, and the way that +# was being missed entirely: the carrier stays up and the ident is an audio +# tone keyed inside it, so a CW detector looking for a keyed carrier sees a +# carrier that never stops. + +def _ident_over_a_tone(text="KSQ330", wpm=22.0, rate=16000, tone=795.0, + before=11.0, after=11.0, seed=3): + """A keyed ident with a steady tone either side of it, as heard over FM.""" + rng = np.random.default_rng(seed) + keyed = morse_audio(text, wpm, rate, snr_db=30.0, tone=tone) + steady = np.sin(2 * np.pi * tone * np.arange(int(before * rate)) / rate) + tail = np.sin(2 * np.pi * tone + * np.arange(int(after * rate)) / rate) * 0.9 + audio = np.concatenate([steady * 0.9, keyed, tail]).astype(np.float64) + return audio + rng.standard_normal(audio.size) * 0.02 + + +def test_an_ident_is_found_inside_a_capture_that_is_mostly_something_else(): + """decode_morse reads a clip that is Morse; find_morse looks for one + inside a clip that is not.""" + audio = _ident_over_a_tone() + assert decode_morse(audio, 16000).complete_text != "KSQ330", \ + "if the whole clip decodes there is nothing for find_morse to fix" + found = find_morse(audio, 16000) + assert found is not None and found.is_morse + assert found.complete_text == "KSQ330" + assert 19 <= found.wpm <= 25 + + +def test_a_steady_tone_at_the_edge_is_not_a_sliced_character(): + """A mark far longer than any dash is not a truncated element -- it is + the transmission the ident was sent over. Read as a cut, it took the + first and last letter of every ident, and with them the whole callsign: + one word with no gap in it to survive the drop.""" + found = find_morse(_ident_over_a_tone("W1AW"), 16000) + assert found is not None + assert found.text == "W1AW" + assert found.complete_text == "W1AW", "the ident was thrown away as cut" + + +def test_a_capture_with_no_morse_in_it_yields_none(): + rng = np.random.default_rng(9) + rate = 16000 + noise = rng.standard_normal(20 * rate) * 0.2 + speechy = noise + 0.4 * np.sin( + 2 * np.pi * 300 * np.arange(20 * rate) / rate + * (1 + 0.3 * np.sin(2 * np.pi * 3 * np.arange(20 * rate) / rate))) + for clip in (noise, speechy, np.zeros(20 * rate)): + assert find_morse(clip, rate) is None + + +def test_the_search_is_no_looser_than_the_decoder(): + """Every window is judged by is_morse exactly as a whole capture would + be; the search widens where the decoder looks, not what it accepts.""" + rng = np.random.default_rng(11) + clip = rng.standard_normal(30 * 16000) * 0.3 + assert find_morse(clip, 16000) is None + + +def test_a_reading_of_nothing_but_dots_and_dashes_is_refused(): + """E and T are the one-element characters, so a decode made only of them + can hardly be wrong -- there is nothing in it to get wrong. A capture of + noise on 445.5 MHz came back as "T T T E E E E E E E E" with the element + mix and the timing fit both inside their bands.""" + from bandsaunter.morse import MorseResult + flat = MorseResult(text="T T T E E E E", wpm=20.0, confidence=0.9, + n_elements=7, n_characters=7, timing_fit=1.0, + snr_db=40.0, undecoded=0) + assert not flat.is_morse + real = MorseResult(text="W1AW", wpm=20.0, confidence=0.9, + n_elements=11, n_characters=4, timing_fit=1.0, + snr_db=40.0, undecoded=0) + assert real.is_morse + + +def test_the_whole_capture_is_one_of_the_windows(): + """When the capture *is* Morse -- a beacon keying through all of it -- the + longest window is the best one, because a word is only certain when a gap + bounds it at both ends and a short window may not contain one.""" + audio = morse_audio("CQ CQ DE W1AW W1AW K", 18, FS, 20) + whole = decode_morse(audio, FS) + found = find_morse(audio, FS) + assert found is not None + assert len(found.complete_text) >= len(whole.complete_text)