Robin Abbott wrote:
>
> Thomas
>
> Your question raises some interesting points. In any asynchronous serial
> transfer such as RS232 the timing of an entire frame (byte) is derived from
> the falling edge of a start bit. You need to consider that a 10% error in
> clock might cause the first bit to be read 10% too soon - not a problem,
> however by the tenth bit a 10% error from the first bit is now an entire bit
> period, consider that both ends may use a clock with an error, and it's soon
> clear that accuracy needs to be at least 2-3% for such a transfer to be
> reliable with a safety margin.
>
> In your case a resonator will be fine I think.

Just to clarify better the idea, take a look at:
http://www.ustr.net/8051pc/starting.htm
from the middle of the page, "Synchronization"... I put some graphs and
text about this possible "clock slack".

Wagner