Why TWI can't work at 100k freq?

[luohuabo]

Hi dear engineers,

I have a strange TWI problem on my production protype which programmed based on the official twi example with softdevice s130_nrf51_2.0.1 and SDK 12.3.0. The device worked under a 16M XTAL HF clock and 32.768k Rail-2-Rail LF Clock. Nrf51822 TWI bus connects to 2 slave sensors with 2.8k pull-up resistors (I tried 10k, 4.7k before) and no internal pull-up GPIO setting. It's almost 100mm between the Master and Slave chips, and the wire have to go through FPC by 2 fpc connectors in the route.   

TWI works not very good at 400 freq, but much better at 250k with occasional commucation error . When I changed the freq to 100k so as to obtain a perfectly normal result, it couldn't work entirely!  I don't know what happened and how to calculate and set to other feasible freq.  Or maybe it is a flaw in my hardware design.

The twi signal wave shape from oscilloscope is as below:

(fig1. wave when 400k freq with 2.8k pull-up)

(fig2. wave when 250k freq with 2.8k pull-up)

Thanks.

[FormerMember]

At which priority does TWI run?

[luohuabo]

Hi Kristin,

Thank you.

The Prioprity is 3 according to the default setting for nrf51 in the file <sdk_config.h>.

#ifndef TWI_DEFAULT_CONFIG_IRQ_PRIORITY
#define TWI_DEFAULT_CONFIG_IRQ_PRIORITY 3
#endif

[FormerMember]

Could you upload a similar plot from the oscilloscope when using 100k?

When setting the frequency to 100k, what happens? Is there any error codes? An assert or a hardfault? How do you see in the code that it doesn't work?