After months of getting my butt kicked by the TL16C2752 DUART I finally stumbled upon this errata document from TI which kind of explains the behaviour I'm seeing and which indeed goes away when using 2 stop bits instead of one as suggested in the document.

My hardware ain't broken, and neither my code. It's the fucking chip!

https://www.ti.com/lit/er/sllz058a/sllz058a.pdf

I did a writeup about the exact phenomenon and what I did to pin down the problem. May it save somebody's time.

https://www.steckschwein.de/post/2026/02/double-trouble/

#uart #retrocomputing #serial #hardware

UART: Double Trouble - Steckschwein

@afachat helped out with a TL16C2552, and all problems are gone. Thank you very much Andre, I owe you!
@steckschwein_6502 glad I could help!
Looking forward to meeting you again at vcfe in Munich!