As I documented in my previous log, decoding capabilities can be strongly improved.

Unfortunately, after working a couple of week in the attempt of porting ft8mon on ESP32-S3, I realized I'm not able to do it. In fact the so many FFTs and LDPC algorithms working on ft8mon cannot be shrunk easily into the poor ESP32-S3.  

So, I'm inspecting the chance to improve the actual decoder, already running into the ESP32-S3 and documented here https://github.com/guido57/ESP32_ft8_lib.

The main new feature would be to add a second decoding step where I run again the decoding process after having "subtracted" the already decoded FT8 messages from the original 15 seconds audio.

I.E.

1. receiving audio and creating the waterfall (one or more FFTs every 160 ms)
2. first step decoding
3. subtracting the FT8 decoded messages from the audio
4. re creating waterfall
5. second step decoding

As today, 1. and 2. must end just before the end of actual slot, so that the transceiver can manage/answer the already decoded messages during the next slot.

On the contrary,  3. 4. and 5. can also be accomplished by the end of the next slot.

Please note 3. 4. and 5. must run while 1. for the next slot is working.