How I finally got LE Audio running on the ESP32-C6, steps that actually worked

I got something similar running last month with a pair of cheap JBL earbuds and a C6 dev board on my desk, single chip setup. The audio works but I would echo the timing warning hard. My first attempt had constant little dropouts every few seconds until I bumped the sync interval up, then it settled down. I never got around to testing the two chip split because the single chip version was stable enough for my bench tests. The silent failure on codec mismatch burned me too, spent an afternoon assuming my radio config was broken when really the earbuds wanted a diferent frame duration than what I was sending. Once I logged what the buds actually advertised and matched it, sound came through on the first try. Reconnect handling I did differently, I just wrote the bond info to NVS right after pairing and that seems to hold across reboots so far. Curious what latency numbers other people are seeing, mine feels around 200ms which is fine for music but useless for anything interactive.

Tried the single chip build on a dev board with some old buds last week. Mine connected fine but I never got audio, so the codec mismatch thing might be my whole problem. Question though, when you log what the buds advertise, does that show up in a normal HCI trace or do you need to dig into the firmware side? Also wondering if the latency varies by codec frame duration, since mine feels sluggish even before sound worked, if that makes sense.

This lines up with what I've seen on my day job testing radio firmware. We ran a two chip split on an ESP32-C6 dev kit with a separate board doing the host side, and the HCI link over UART at 3 Mbps was the piece that needed the most babysitting. Flow control off and it'd silently drop command packets under load. Also the C6 controller build has to have the ISO features compiled in, the prebuilt binaries don't all include them, so check that before you blame your code.