Answer 1—An asynchronous receiver starts measuring time from the leading edge of the start bit associated with each byte transmitted. It then samples the data line at what it thinks is the center of the next 10 bits (start bit, 8 data bits, stop bit). An error will occur if the transmitted stop bit misses the tenth sampling point as defined by the receiver clock. This will happen if the transmitter clock is fast or slow relative to the receiver clock by plus or minus half a bit after 10 bit times (or ±5%).

Note that this is the total clock error in the link, measuring one clock with respect to the other. If the clocks are each individually calibrated against an absolute standard, they can each be off by only up to ±2.5% and still have the link be guaranteed to work.

If there are any sources of uncertainty on where the leading edge of the start pulse is (e.g., from noise in the link or the use of a 16× clock in the receiver), then a more conservative number (e.g., ±2%) should be used.

Contributor: David Tweed