Parity not used at all - just start, 8-data, stop (10 bit times in all).
The Stop bit time is precise - verified on scope. I can scope the ES interrupt occuring just after 9/16ths of the stop bit - but the following transition from stop to start states at 16/16ths is ignored - the next transition (in the same direction) is incorrectly interpreted as the next start bit...
Have read the print off the pages of the (various) manuals - no clues - it SHOULD work. Some days it does, some it doesn't - I assume that is down to my PC comms utility having a slightly longer idle time between transmitted characters because of varying background tasks.
The timing is so marginal that I am convinced if I could introduce just the tiniest fraction of a bit time idle period between characters, it would work ok. Adding an extra stop bit always works - but is obviously rather brutal...