Tx drive levels

[drowe67]

As per the standard we established with FreeDV waveforms, the signal level of the RADE Tx signal passing through freedv-gui and being sent to the sound card should be approximately +/- 16383 (leaving 1 bit of dynamic range free for overshoot issues caused by further signal processing, and the effect of filtering).

The tx level should never be anywhere near 32767 (16 bit clipping point) at any point in the freedv-gui Tx signal processing chain. This is because other signal processing (e.g. filtering/resampling in the drivers or sound card) can have overshoot that would push it beyond the clipping point, distorting the waveform, lowing performance, and causing splatter.

The example in the radae repo show scaling to approximately this level, e.g. in ota_test.sh

cat ${tx_radae}.f32 | python3 f32toint16.py --real --scale 16383 > ${tx_radae}.raw

In the sample from drowe67/radae#36 the Tx sample is clipped. Is this the same signal being sent to the soundcard D/A? This would cause distortion of the RADE waveform. I have plotted tx.wav (sample from the radae repo, scaled correctly) below for comparison.

Also using octave:

octave:21> s=load_raw("~/Downloads/FreeDV_FromRadio_20241201-191336sample.wav");
octave:22> plot(s)
octave:23> max(s)
ans =  32767

[tmiw]

This was partially fixed in #776 (https://github.com/drowe67/freedv-gui/pull/776/files#diff-96063afc5af5ad68d26aeb3c004e4ad8913e233ebc8658a2c79f9cc50042c5bbR126). EOO TX will be updated as well in #780 prior to the merge of that PR.

That said, the TX Attenuation slider would cause that signal to be scaled down prior to TX but wouldn't be reflected in any recordings done by users.

[tmiw]

EOO scaling fixed in 864640e.

[tmiw]

Since #780 is now merged, TX drive levels should be correct for all of RADE (even EOO). Closing.