I’ve had this radio for a few years now and always wanted to use it for M17. Today is the day! The app supports AGC known from DroidStar and also adds FARGAN neural synthesizer for Codec2 payload. If you own an Atom XL, give it a try. This is not a Zello-type app, it works with RF only.
Repository: https://github.com/sp5wwp/AtoM17
I successfully transmitted an encrypted and signed M17 stream carrying Codec2 speech payload with a CS7000. The stream was successfully deciphered using our fresh SDR# M17 plugin. The ECDSA signature was successfully verified as well.
Splitting the signing process into “precompute” and “finish” was a great idea. The “precompute” step picks a random nonce, does the expensive point multiplication and computes the nonce’s inverse (using the elliptic curve). This happens when the PTT is pressed and takes roughly 30ms. When PTT is released, the “finish” function does the rest, and that part is shorter – about 2ms.
Vlastimil OK5VAS (one of the guys behind LinHT!) is preparing an awesome presentation Hackněte si éter: radioamatérství 21. století (Eng.: Hack the radio waves: 21st century amateur radio) for this year’s OpenAlt in Brno. His presentation is scheduled for Sunday, November 8, at 11:00, room D0207.
It is very likely that the event will be livestreamed. Follow OpenAlt’s YouTube channel for updates.
I prepared an M17 plugin for SDR#. It supports streams and packets. It also verifies ECDSA signatures and can decrypt streams .
Repository: https://github.com/sp5wwp/m17_sdrsharp
Am I crazy? Without a doubt, but at least I focus my craziness on a good purpose. My last idea was to find out if all of the following assumptions are viable:
The answer is… yes to all.
Let’s analyze it step by step.
I wrapped it all up on top of turnrye’s OpenRTX fork (this one is focused on M17 and uses the DSP firmware part setting the baseband sample rate to 48kHz). My fork is here. The build process pulls ETSI’s codebooks for their ACELP vocoder, so make sure you do not distribute the binaries (as they are embedded). TETRA is available through the quick menu key, the [5] key switches between M17, FM, and TETRA.
Below, I attach a video showing a successful TETRA decode using a TMO transmission sample from the HamTETRA group. Big shout out to Manoel ON6RF for granting me access to their samples database.
Have fun!
OpenACELP vocoder has been updated to pull ETSI EN 300 395-2 tables at build-time. It is now a clean-room implementation of the TETRA voice coder.
The build process offers two solutions:
When using ETSI’s codebooks, make sure you do not distribute the compiled binary, as distribution of the codebook tables is prohibited. More details in the repository’s readme files.
Our experimental fork of DroidStar can now decode M17 voice with a neural vocoder. The new decoder comes from Codec2-mod-neural, which uses the FARGAN network from the Opus project. It takes the same Codec2 3200 frames every M17 radio already sends and turns them into wideband 16 kHz speech instead of the classic 8 kHz sinusoidal synthesis. Nothing changes on the transmitting side, the bitstream is untouched, so you get the improvement on every 3200 bps M17 stream you receive, no matter who is talking.
Installing
To enable the neural synthesizer, go to Settings and scroll down to “Neural Codec2 decoder (M17 3200)” – tick it. Connect to a reflector afterwards, as usual.
Tell us how it sounds and which phone you use. Waiting for comments!
Codec2 3200 carries M17 voice, and its classic decoder is the weak spot. It’s limited to 8 kHz and has that recognizable buzzy, synthetic sound. Instead of designing a new codec, I wanted to know how far a better receiver could go with exactly the same bits on the air.
The answer is FARGAN, the low-complexity neural vocoder from Xiph’s Opus codec (Valin et al.). The idea is simple. The Codec2-mod decoder unpacks each 8-byte frame into its usual parameters: pitch, voicing, energy and LSPs. Instead of feeding those to the sinusoidal synthesizer, we map them to 20 features per 10 ms and let FARGAN generate 16 kHz speech.
Data: the Opus speech corpus, processed with Opus’s own augmentation recipe, but with features computed by the real Codec2-mod encoder and decoder. An important early lesson: never pair noisy features with a clean target. The network can’t denoise, and my first attempts sounded like speech through a pipe.
Training: Opus’ two stage recipe on rented GPUs1. Stage 1 was 400 epochs (about 14.5 hours), and stage 2 was 50 adversarial epochs (about 37 hours on an RTX 5090). The adversarial stage made a clearly audible difference from its first few epochs.
C library: the model uses about 818,000 parameters, runs on Opus’ own C inference code, and takes about 1.6 ms of CPU time per 20 ms frame on a single x86 core. The ARM64 NEON build compiles and produces matching output, real hardware timing test will follow.
The result is Codec2-mod-neural, a drop-in alternative to codec2_decode(): 8 bytes in, 320 samples at 16 kHz out. Nothing changes for transmitters, and every existing Codec2 decoder still works on the same stream.
Code, training scripts, and the model: https://github.com/M17-Project/Codec2-mod-neural
Next up: integration into DroidStar and mvoice, and later a look at running it on LinHT. ViSQOL benchmarks will also be ran along with synthetic BER tests to assess synthesizer’s noise immunity.
Vlastimil OK5VAS just sent me a nice set of photos (click to enlarge).
Standby for test results!
The LinHT project is proudly supported by the M17 Foundation and their generous donors.
Two recent commits make libm17 (versions 1.4.0 and 1.4.1) a better fit for small microcontrollers, such as the Cortex-M4-based radios OpenRTX runs on.
The first, 5214d99, reduces memory and code size without changing any output. The receive front-end and the Viterbi depuncturing now work in a single pass, and the four convolutional encoders became one. On a Cortex-M4F this reduced the Viterbi decoder context from 1,596 to 620 bytes, decoder stack from about 1.5 KB to 800 bytes, and flash by about 840 bytes. The decoder context changed size, so this is a slight ABI change.
The second, 0c92621, replaces the soft Golay(24,12) decoder used for the LICH with a Chase-II decoder that rejects ambiguous codewords. In simulation, LICH sensitivity improved by about 1 dB, and wrong LICH chunks are accepted 20–80x less often. As a result, decode_str_frame() will report a failed LICH (lich_cnt = 0xFF) more often at low signal levels. On a Cortex-M4, decoding a stream frame takes roughly 30-40% more CPU than before, which is still less than OpenRTX’s own decoder. The same commit uses a hardware popcount where the target has one and makes the unit tests compile without warnings.
LSF, payload and transmit output are unchanged, and the public API is the same. In simulated tests against OpenRTX’s hard-decision frame decoder, libm17 needed about 3 dB less SNR for a successful decode.