If you don’t use compression (AGC) there is no reason that the background noise would grow up. It may well be the case otherwise if it passes the squelch (power threshold) level. The audio input is a bit picky you have to control it simultaneously with pulse audio control (pavucontrol is convenient for that).
If you checked out the latest commit on dev branch then yes I have made changes to the main samples buffer size calculation in the LimeSDR output plugin so that it will have much less latency, Above 192 kS/s in the baseband the FIFO has a 0.25s fixed delay size . 192 kS/s is the limit (hence 48000 samples) below which I can’t get along without chopping.
I’m also making changes to the other Tx plugins (BladeRF, HackRF) to get a fixed time sized FIFO when possible. This could also help at high rates where the fixed length FIFO was maybe in fact too short.
I also made changes to the audio plugins (AM, NFM, SSB, WFM) so that the audio FIFO is 100ms instead of 1s. They still work fine apparently and it doesn’t ruin the efforts made on the sample FIFO.
Best regard, Edouard.