If there's a timer or two available for an ISR, couldn't that also serve to create the required baud rates (the O/P hasn't said whether that's even called-for) and then allow the core to be clocked at some rate that's a harmonic of the desired counter frequencies? If one 16-bit counter can create the interrupt that software counts, having it at an integral multiple of the desired interval would be useful, wouldn't it?

Locking one's timing to a specific baud-rate generator crystal can prove a serious disadvantage under some circumstances, e.g. this case. It is not unsolvable with the "old, tried-and-true" crystal but if it's convenient, why not try, say, 24 MHz? You can get "close enough" to standard baud rates with that. Aside from that, dividing 11059200 by 49152 in a 16-bit counter leaves a divide-by-225 to generate 1 second.

RE