UART / USART baud rate

[aseris]

Posted on May 27, 2016 at 14:57

Is it possible to use baud rate divisor less than 1? i.e. 1/16 - USART_BRR = 0x01

I did not find any information about it in STM32F4 RFM.  #uart-usart-baud

[Tesla DeLorean]

Posted on May 27, 2016 at 15:16

It would be 1/16 or 1/8 depending on the overclock setting.

Typically in 16x mode, BRR = APBxCLK / Baud

[aseris]

Posted on May 27, 2016 at 15:27

I wrote wrong number it have to be 1/16. Oversampling8 = 0

But anyway. If my clock is. i.e. 10 MHz, then can I apply divisor  1/16? - to achieve 10mbps?I.e. is the closer to HW equation is  fclk/((uint16_t)divisor)?      (OVER8 =0)But I'm also confused how I can divide it by 0? There are no any restrictions in datasheet.

[mark239955_stm1]

Posted on May 28, 2016 at 03:17

The divisor is a fixed-point rational number; the mantissa is to the left of the binary point, and the fraction is to the right of the binary point, in the form mmmmmmmmmmmm.ffff (when OVER8 = 0).

In other words, if you set the mantissa to 0 and the fraction to 1, i.e. 000000000000.0001, the clock is divided by 16 and then divided by 1/16, which cancel out to give you a baud rate equal to the clock. You're correct in thinking that in effect this means that if you treat the BRR register's value as an integer, the baud rate is given by dividing the UART clock by the BRR value. Bear in mind that the UART clock is not HCLK; it is APBxCLK, depending on which UART you're using.

[Tesla DeLorean]

Posted on May 28, 2016 at 15:11

It can't divide by zero and it is not actually fractional. The BRR is a prescaler the division by 16 is fixed and occurs later

[Tesla DeLorean]

Posted on May 28, 2016 at 15:22

Divide by one wouldbe the unity case. I suspect they use a downcounter.

The silicon implementation is a lot simpler than the dogs breakfast presented by the techical writer. So either someone didn't explain it well or used a cell library they really didn't understand the internals for. Ie inherited it where the designer said it does the *equivalent* to this by doing that.

[aseris]

Posted on May 31, 2016 at 13:12

After additional deep study RFM and device datasheet, I figured out next items:

The divisor in fixed point interpretation can't be less than 1. It can be seen from maximum baud rates for UART in device datasheet.