Large actual baud rate variation when using Baud921600 setting (actual rate: 941176), even with HFCLK derived from using external clock source 32MHz +/-10ppm accuracy

[SyedSajid]

Hi 

We are seeing board to board variations in actual baud rate from NRF5340 for Baud921600 setting, bit time is varying from ~938k to ~952k
the HFCLK is derived from external clock 32MHz +/-10ppm accuracy (Epson Q22FA12800025) 

1. what is the tolerance expected on the baud rate when using such external clock?
2. In what ways we can tighten the tolerance deviation?
   
   

[Håkon Alseth]

Hi,

 

It should be quite stable once the HFCLK source is set to HFXO and its state is running. The actual baudrate is a bit off, around 20kHz, but it shall follow the tolerance of the HFCLK source.

Based on your description, it sounds like you are running on HFINT (internal RC oscillator). Could you ensure that the HFXO is running by checking this register?

https://infocenter.nordicsemi.com/topic/ps_nrf5340/chapters/clock/doc/clock.html?cp=3_0_0_3_10_2_33#register.HFCLKSTAT

 

Kind regards,

Håkon

[James Lim]

Thanks Håkon.

By manually triggering HFCLK using TASKS_HFCLKSTART,  we are able to get the 941k frequency.

However this would mean that the HFCLK will always be on.

Do we have to manually trigger the HFCLK start stop for UART TXRX operation?

[Håkon Alseth]

Hi,

 

James Lim said:

However this would mean that the HFCLK will always be on.

Do we have to manually trigger the HFCLK start stop for UART TXRX operation?

That is correct. The HFXO must be active when UART RX is active. If you have a added ready signal, this must then not only start the RXD receiver, but also start up the HFXO (and account for the time for settling: https://infocenter.nordicsemi.com/topic/ps_nrf5340/chapters/oscillators/doc/oscillators.html?cp=3_0_0_3_11_4_0#unique_817147230)

 

Kind regards,

Håkon