UART BAUDRATE register values

[Bengt Florin]

What is the formula for calculation of values for the UART BAUDRATE register from bit rates?

[Håkon Alseth]

Hi Bengt,

note: this formula is simplified and may differ from the actual baudrate measurement

The formula is: Baudrate = desired baudrate * 2^32 / 16000000

Example: Baudrate of 31250 should then be 8388608 decimal = 0x800000.

Note that you will have to round the number afterwards: rounded_value = (value + 0x800) & 0xFFFFF000

Since the baudrate generator will be sourced by the 16 M (Either RC or XOSC, depending on your configuration), then your error rate will be ~equal to the overall drift of the system clock and the accuracy of the baudrate generator towards your target baudrate.

Other way around: Check out this define in nrf51_bitfields.h line 5702:

UART_BAUDRATE_BAUDRATE_Baud115200 (0x01D7E000UL) /*!< 115200 baud. */

The actual baudrate set with the above define is:

baudrate_reg_val * 16M / 2^32 = 115203.86 baud.

Best regards Håkon

[Bengt Florin]

Tnx Håkon. I assume you put this in the next Reference Manual as well.

[Bengt Florin]

BTW, why the heavy rounding? 'nrf51_bitfields.h' have a value for 921600bps that is 0x0EBEDFA4UL. By rounding this gets 0x0EBEE000. Does this imply the the baud rate counter is not 32-bit but mere 20-bit?

[Bengt Florin]

This is the tested code to get the BAUDRATE value from an arbitrary bit rate. It gives the same values as 'nrf51_bitfields.h' except for 921600bps and for that we wait for an explanation from Håkon.

static uint32_t uartBaudrate(uint32_t bps) { return ((bps * 0x100000000ULL) / SYSCLK + 0x800) & 0xfffff000; }

[Håkon Alseth]

Hi Bengt,

That specific value got me a bit puzzled. I had a quick chat with a HW-designer, and he assured me that the baudrate generator for the UART is actually 20 bits, so rounding procedure is required.

As far as I see, the 921K600 baud definition in nrf51_bitfields.h is incorrect. I will input a bug regarding this item. I see that this define is removed from our reference manual, so it's likely a glitch that it's still defined in the header file.

I will also push for the calculation to be inserted into our documentation.

BR Håkon