diff options
Diffstat (limited to 'README.md')
| -rw-r--r-- | README.md | 103 |
1 files changed, 103 insertions, 0 deletions
@@ -10,3 +10,106 @@ somewhere safe. - https://pygobject.gnome.org/getting_started.html - https://code.nap.av.it.pt/mobility-networks/modemmanager/-/tree/1.18.12/examples/modem-watcher-python + + + + + +## Call answering machine (mmfwd-callam) +https://github.com/dxdxdt/mmfwd/raw/refs/heads/calls/doc/img/sim7600-rickroll.post1.720p.mp4 + +Usage: + +```sh +mmfwd-callam <modem audio char path> <output prefix> +``` + +Making audio samples: + +```sh +ffmpeg -i FILE -vn -ar 16000 -ac 1 -f s16le OUTFILE.pcm +``` + +or use Audacity: +- Go to `File` - `Export Audio` + - Format: "Other uncompressed files" + - Channel: Mono + - Sample Rate: 16000 Hz + - Header: RAW (header-less) + - Encoding: Signed 16-bit PCM +Audacity overrides the extension to .raw. Change it to .pcm. + + + +The audio streaming is handled by the child process `mmfwd-callam` since the +audio serial port is synchronous(CTS/RTS). Common tools like `aplay` and +`arecord` do not support nonblocking IO. + +[src/mmfwd-callam.cpp](src/mmfwd-callam.cpp): + +```c +void init_params (void) { + params.hello_sample_pat = "hello/*.pcm"; +``` + +Place PCM files in `hello/*.pcm`. Then the main mmfwd Python module can spawn +the process. + +[src/mmfwd/__init__.py](src/mmfwd/__init__.py): +```py +now = datetime.datetime.now(datetime.UTC) + +n_from = call.get_number() or "" + +dir = "rec/%02d-%02d" % (now.year, now.month) +filename = now.isoformat(timespec = 'milliseconds') + '_' + n_from +path = dir + '/' + filename + +os.makedirs(dir, exist_ok = True) + +exec = [ ud.instance.callam['exec'], "/dev/" + ud.audio_port, path ] +ud.instance.callam_proc = subprocess.Popen(exec) +# 5 minutes timeout +ud.instance.callam_timer = GLib.timeout_add_seconds( + 60 * 4, + self.on_call_timeout, + ud) +``` + +A pair of recorded audio files will be placed in +`rec/YYYY-MM/<ISOTIME in millis resolutiono>_<NUMBER>.<INOUT>` where + +- NUMBER is the phone number of the calling party +- INOUT is "in" for the audio from the calling party and "out" for from mmfwd + +To encode the recorded raw audio files: + +```sh +ffmpeg -ar 16000 -ac 1 -f s16le -i PCM_FILE -c:a libopus OUTFILE.ogg +``` + +## MM Patches +- https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/merge_requests/1293 +- https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/996 + +## Another idea (070 number) +Major telcos in South Korea(KT, SKT, LG U+) offer VoIP services. After +subscribing to such services, the SIP credentials can be extracted from the +device. AFAIK, the telcos' SIP servers don't do whitelist connections so the +credentials can be used pretty much from anywhere with any SIP+VoIP client +implementations. + +Sources: + +- https://blog.naver.com/rehearsal/80099885977 +- https://www.clien.net/service/board/use/1219343 +- https://koreapyj.dcmys.kr/152 + +ModemManager and LTE modems such as SIM7600 are usually optimised for mobile +date network connectivity on Linux world, not for voice calls. You might as well +get a 070 number(the standard VoIP prefix in Korea) and redirect the calls and +messages to the SIM number to the VoIP number to get high quality audio straight +from the telcos' IP core network. + +- Pro: Stable audio stream +- Con: Adding another point of failure to the system |