More seriously

I have :

[RADE] sidecar starting at Fri, 10 Apr 2026 15:41:13 +0200
[RADE] helper starting on ws://0.0.0.0:8074
[RADE] radae_rx.py    : /home/f1nsk/radae/radae_rxe.py
[RADE] model          : /home/f1nsk/radae/model19_check3/checkpoints/checkpoint_epoch_100.pth
[RADE] lpcnet_demo    : /home/f1nsk/radae/build/src/lpcnet_demo
[RADE] auxdata        : ON (default)
[RADE] torch threads  : 1 per instance (RADE_TORCH_THREADS to override)
[RADE] architecture   : per-connection (each user tunes independently)
[RADE] listening — waiting for PhantomSDR-Plus clients

But don't work 

The most important, we solve the audio cut !

In a few time all my friends will be on W11 or Linux so i will switch on Opus.
Better sound and less latency.

Have a nice day Doctor ! 
 

I will try a full fresh install on an other server.

At this time, the ambulance is front the portal to bring me at the Psychiatric Hospital ....    

Let us try to reinstall manually:

sudo apt update
sudo apt install -y \
    build-essential cmake git \
    python3 python3-pip \
    nodejs npm \
    aplay alsa-utils

Python packages:    
pip3 install websockets matplotlib torch numpy scipy

Verify Python packages:
python3 -c "import websockets, matplotlib, torch, numpy, scipy; print('All OK')"

Clone and Build the radae Repository (delete the previous ~radae -> rm -rf radae):
sudo apt install build-essential cmake git python3-pip
git clone https://github.com/drowe67/radae.git ~/radae
cd ~/radae
mkdir build && cd build
cmake ..
make -j$(nproc)

Verify lpcnet_demo was built:
ls -la ~/radae/build/src/lpcnet_demo

Verify model19_check3 weights are present (this is the correct model):
ls ~/radae/model19_check3/checkpoints/checkpoint_epoch_100.pth

Install Python Dependencies:
pip3 install websockets matplotlib torch numpy
pip3 install scipy

Verify the Decode Pipeline:
cd ~/radae
./inference.sh model19_check3/checkpoints/checkpoint_epoch_100.pth \
    wav/brian_g8sez.wav /dev/null \
    --rate_Fs --pilots --pilot_eq --eq_ls --cp 0.004 \
    --write_rx rx.f32 --auxdata
 
Wait for it to finish. It prints stats ending with:
loss: 0.741 Auxdata BER: 0.012

Then decode and play:
cat rx.f32 \
    | python3 radae_rxe.py --model_name model19_check3/checkpoints/checkpoint_epoch_100.pth \
    | ./build/src/lpcnet_demo -fargan-synthesis - - \
    | aplay -f S16_LE -r 16000


You should hear a voice. The output shows sync acquisition. The `underrun!!!` messages during this offline test are expected:
  1 state: search     ...
  5 state: sync       ... SNRdB:  4.03 uw_err: 0
Playing raw data 'stdin' : Signed 16 bit Little Endian, Rate 16000 Hz, Mono


Verify the sidecar starts correctly:
python3 ~/PhantomSDR-Plus/rade_helper.py

Expected output (no WARNING lines):
[RADE] helper starting on ws://0.0.0.0:8074
[RADE] radae_rx.py    : /home/sv1btl/radae/radae_rxe.py
[RADE] model          : /home/sv1btl/radae/model19_check3/checkpoints/checkpoint_epoch_100.pth
[RADE] lpcnet_demo    : /home/sv1btl/radae/build/src/lpcnet_demo
[RADE] auxdata        : ON (default)
[RADE] torch threads  : 1 per instance (RADE_TORCH_THREADS to override)
[RADE] architecture   : per-connection (each user tunes independently)
[RADE] listening — waiting for PhantomSDR-Plus clients

My working files are these:

PhantomSDR-Plus
rade.sh
rade_helper.py

frontend/src
audio.js

Please download them, recompile frontend and retry without restarting

everything seems to be o.k. Does the window turns to green?

P.S. No it does not, as I can see...

Can you try via localhost, directly from the pc that server is installed? Open a browser and type http://localhost:8080

You can also try this audio.js

The result


[RADE] Running
f1nsk      80318  0.0  0.0   9976  3540 pts/4    S    13:25   0:00 bash -c          while true; do             echo "[RADE] sidecar starting at $(date --rfc-email)" >> "/home/f1nsk/PhantomSDR-Plus/rade.log"             python3 "/home/f1nsk/PhantomSDR-Plus/rade_helper.py" >> "/home/f1nsk/PhantomSDR-Plus/rade.log" 2>&1             echo "[RADE] sidecar exited at $(date --rfc-email) — restarting in 3s" >> "/home/f1nsk/PhantomSDR-Plus/rade.log"             sleep 3         done     
f1nsk      80321  0.1  0.1  32432 21552 pts/4    S    13:25   0:00 python3 /home/f1nsk/PhantomSDR-Plus/rade_helper.py
and the log file
 [RADE] sidecar starting at Fri, 10 Apr 2026 13:25:12 +0200
[RADE] helper starting on ws://0.0.0.0:8074
[RADE] radae_rx.py    : /home/f1nsk/radae/radae_rxe.py
[RADE] model          : /home/f1nsk/radae/model19_check3/checkpoints/checkpoint_epoch_100.pth
[RADE] lpcnet_demo    : /home/f1nsk/radae/build/src/lpcnet_demo
[RADE] auxdata        : ON (default)
[RADE] torch threads  : 1 per instance (RADE_TORCH_THREADS to override)
[RADE] architecture   : per-connection (each user tunes independently)
[RADE] listening — waiting for PhantomSDR-Plus clients

Quote from: F1NSK on Apr 10, 2026, 12:45 PMAt the line 264, I have changed the buffer threshold to 0.2 ( 200ms ?)

Wonderful !!!!

Quote from: F1NSK on Apr 10, 2026, 12:45 PMThe install run ok. The server start ok but no sidecar.
I investigate on the port 8074 without better results.

What about phantomSDR-Plus/rade.sh?
Please do ./rade.sh restart and then ./rade.sh status and paste me the results.

Hello Emmanuel

Sorry for Paul but i have deleted the link in the toml and now no problem.

An important fact about audio cuts

I benefit of a very fast internet connection, so i never use audio buffer.

I have tried to put the buffer at X3 and remark no cuts !

So i have shaked my brain    and edit the last release of audio.js

At the line 264, I have changed the buffer threshold to 0.2 ( 200ms ?)

After server restart, absolutly no cuts, even with buffer slider at 0

I've never had a such stability and now SSTV work perfecty, the auto-mode decoding very fast between Martin or Scottie !   

At this time i work on RADE...
The install run ok. The server start ok but no sidecar.
I investigate on the port 8074 without better results.

73's

Quote from: F1NSK on Apr 10, 2026, 08:27 AMI can read a lot of identical messages

" curl_easy_perform() failed for http://list.phantomsdr.fun/api/update_websdr: Timeout was reached "

Well, this means that the server faces difficulties to communicate with http://list.phantomsdr.fun/api/update_websdr. Try to join to http://list.novasdr.fun/api/update_websdr instead, so in .toml:

[websdr]
register_online=true # Enable directory registration updates
register_urls=[
  "https://sdr-list.xyz/api/update_websdr",
  "http://list.novasdr.fun/api/update_websdr"] # One or more directory endpoints that accept the same JSON payload

Hello Emmanuel

When i leave terminal open on the server after the restart, behind the end of the normal sequence

" waterfall is sent every 1 fft "

I can read a lot of identical messages

" curl_easy_perform() failed for http://list.phantomsdr.fun/api/update_websdr: Timeout was reached "

Perhaps a way for rare but always present short freezes of the server ..

73's