Function LEUART_BaudrateCalc

bsp/efm32/Libraries/emlib/src/em_leuart.c:127–189

/ * @brief * Calculate baudrate for LEUART given reference frequency and clock division. * * @details * This function returns the baudrate that a LEUART module will use if * configured with the given frequency and clock divisor. Notice that * this function will not use actual HW configuration. It can be used * to determinate if a given configuration is sufficiently accurate for th

 *   Baudrate with given settings.
 ******************************************************************************/
uint32_t LEUART_BaudrateCalc(uint32_t refFreq, uint32_t clkdiv)
{
  uint32_t divisor;
  uint32_t remainder;
  uint32_t quotient;
  uint32_t br;

  /* Mask out unused bits */
  clkdiv &= _LEUART_CLKDIV_MASK;

  /* We want to use integer division to avoid forcing in float division */
  /* utils, and yet keep rounding effect errors to a minimum. */

  /*
   * Baudrate is given by:
   *
   * br = fLEUARTn/(1 + (CLKDIV / 256))
   *
   * which can be rewritten to
   *
   * br = (256 * fLEUARTn)/(256 + CLKDIV)
   *
   * Normally, with fLEUARTn appr 32768Hz, there is no problem with overflow
   * if using 32 bit arithmetic. However, since fLEUARTn may be derived from
   * HFCORECLK as well, we must consider overflow when using integer arithmetic.
   */

  /*
   * The basic problem with integer division in the above formula is that
   * the dividend (256 * fLEUARTn) may become higher than max 32 bit
   * integer. Yet we want to evaluate dividend first before dividing in
   * order to get as small rounding effects as possible. We do not want
   * to make too harsh restrictions on max fLEUARTn value either.
   *
   * For division a/b, we can write
   *
   * a = qb + r
   *
   * where q is the quotient and r is the remainder, both integers.
   *
   * The orignal baudrate formula can be rewritten as
   *
   * br = 256a / b = 256(qb + r)/b = 256q + 256r/b
   *
   * where a is 'refFreq' and b is 'divisor', referring to variable names.
   */

  divisor = 256 + clkdiv;

  quotient  = refFreq / divisor;
  remainder = refFreq % divisor;

  /* Since divisor >= 256, the below cannot exceed max 32 bit value. */
  br = 256 * quotient;

  /*
   * Remainder < (256 + clkdiv), which means dividend (256 * remainder) worst case is
   * 256*(256 + 0x7ff8) = 0x80F800.