On my M4 Max, the same loop with the same data runs 20% slower depending on which code ran on the core before it, even if that code ran in another process. Chasing why took me from a RustFFT PR to a covert channel through the App Sandbox.
2
4
344
Thanks for the previous work from @name99, @dougallj, @PradyumnaShome, the SLAP and FLOP authors and Asahi Linux. cc @ChipsandCheese9
77
SDRmm now decodes Benshi handhelds (UV-Pro, VR-N76): chat, positions and call requests straight onto the map.
Also new: the AM, NFM and SSB transmitters use a lookahead limiter instead of a hard clip. Hot audio stays in the channel, 30 dB less splatter.
sdrmm.com
Soon in the next Version
(And also a small Teaser for the new Design)
this is more how it should look. still synced at 3 dB SNR on a real PAL broadcast, 20 ms latency (half a frame), colors that actually match the source
Finished color decoding for my analog TV SDR receiver. Still needs a little bit of work and cleanup but it's already miles ahead of existing software. it's doing all of this in real time and barely uses 3% CPU on my desktop, and it can still get better...
youtu.be/EluYvN2pA6Q
SDRmm 2.4.4 is out.
Transmit: voice, APRS, PTT, VOX
DMR, P25, NXDN and POCSAG decryption with your keys
P25 trunking
Aaronia SPECTRAN V6
DAB 2.4x faster, LoRa up to 3x
macOS, Windows, Linux, Docker
github.com/newspicel/sdrmm/r…
sdrmm.com
Rewrote how sdrmm talks to the browser.
Network binds serve HTTPS by default. Chromium connects over WebTransport: events on one stream, audio as datagrams, one stream per lane. A lost packet now costs a few ms of audio instead of stalling everything queued behind it.
Video goes out as JPEG, IQ as 16-bit, and two clients asking for the same spectrum share one reduction. The workspace bar tells you when your link is lossy or falling behind.
1
6
324
Every Speaker in sdrmm can pick its own output device now. One SDR, FT8 into WSJT-X, RTTY into fldigi, the rest on your headphones.
FT8 and FT4 also show DT now, and weak stations get confirmed when they repeat.
7
395
sdrmm can transmit now.
Microphone, Modulator and Transmitter nodes: voice, APRS messages, a channel's baseband or a whole lane, out of any radio that transmits. PTT key, VOX, timeout, and an occupied-width readout so you see how wide you actually are.
Arming is by hand and never saved, so opening a workspace will never key up your radio at 3 am.
HackRF takes turns with its receiver, SoapySDR radios get TX gain, and plugins can feed the Modulator.
With the nightly or 2.5.0
The next sdrmm needs way less CPU:
DAB 2.4x faster
Stereo FM at half, the pilot loop runs at 1/8 rate now (it was very bored before)
Spectrum Monitor a third on busy bands
FIR 1.4 to 1.6x faster
Plus the RustFFT stuff. 2.5.0 soon.
Two weak signal things in the nightly:
802.11b got coherent detection and a CCK chip matched filter, 1 to 2 dB better at every rate. Yes, sdrmm decodes Wi-Fi. No, it will not replace your router.
Analog TV over FM now shows a picture down to 5 dB CNR, in proper PAL and NTSC colour. (The bars are synthetic, nobody transmits bars that clean.)
10
553
LoRa in sdrmm: 1.5 to 2 dB more sensitive and 2 to 3x faster.
Sync used to trust one window. Now it sums evidence over four and wants them to agree. A timing loop follows the $3 modules drifting past 100 ppm. And since the CRC barely covers the last block, that block gets re-encoded and checked against the symbol energies.
Meshtastic people, this one's for you. Nightly.
sdrmm decrypts DMR, P25 and NXDN now.
DMR Basic Privacy doesn't need a key, it leaks it in the silence frames. Calling that privacy is generous.
ARC4, AES, ADP and DES-OFB work with your own keys. Put them on the trunk node and it follows the calls, even when you join mid-call.
Nightly, coming in 2.5.0.
DMR in sdrmm used to fall off a cliff at 12 dB Eb/N0. Not a slope, a cliff. The first burst of a call needed clean timing, so headers just vanished.
Now the timing comes from the first 64 symbols of every burst. Cliff gone.
And BPTC(196,96) decodes as a turbo product code with soft decisions now: 2.8 dB at 1e-3. Curves are from the committed test baselines, not vibes.
Nightly, coming in 2.5.0.
1
179







