Thanks everyone for offering me the very considerable benefit of your knowledge and experience. I'm very grateful to have access to such a resource.

I'm just having some fun with a 'Supported Baud-Rate Calculator' to go along with a polled serial io module that I've been fiddling with recently. For my own hacky purposes it seems that 2-2.5% error might be a good figure for personal experimentation. As the coded io module will be for general distribution however, then a tighter tolerance might be advised. I won't ever know the circumstance in which the code will be used, so I thought I might stick with <1% baud-rate error as a point to flag crystal/ baud-rate combinations as being 'possibly' useful. That's a figure I've seen on a few websites, including Keil's. Following from the informative responses in this thread though, I've decided to adopt an even tighter tolerance of <0.5% baud-rate error as the threshold for flagging a crystal/ baud-rate combination as being 'supported'. It appears that even this figure may be optimistic for some user's purposes, so I thought I might also quote the respective % errors, so that the individual user may decide for themselves.

Thanks also to whoever mentioned using higher numbered timers or the internal baud-rate generator as a UART clock source. For my code to work for the widest possible range of 8051 derivatives, I've been restricted to the use of T1 as the UART's clock source. Thankfully though, I now have a better understanding of why I might wish to employ alternative UART clock sources in personal versions of the code.

I could not have been more pleased with the very helpful responses to my initial query. Thank you all again.

Regards,
Murray R. Van Luyn.