Check what baud rates your clock can actually hit — divider, real baud and error %.

Where baud rate error comes from

A UART derives its bit clock by dividing the peripheral clock:

baud = \frac{f _{clk}}{O S \cdot D I V}

where OS is the oversampling factor (usually 16) and DIV an integer (or fractional) divider. When $f_{clk} / (O S \cdot baud)$ is not an integer, the divider is rounded and the real baud rate shifts. Keep the combined TX+RX error under ≈4.7% for 8N1 framing — in practice each side should stay below about 2.5%. Crystal choices like 48 MHz hit most rates well; 115200 from some clocks (e.g. 16 MHz at ×16) lands at 2.1% error, which is why higher rates sometimes fail.

UART Baud Rate Calculator

Where baud rate error comes from