TWI clock stretching for nrf51822

[haruthakop]

Dear Nordic developers,

Does TWI supports bit level clock stretching? Means that after each bit (not byte) slave can hold down the clock and master should perform appropriate actions.

As I understand from the documentation, it supports only byte level clock stretching.

I'd really appreciate any advise.

Best regards, Harut

[MartinBL]

Hi. I tested it and it seems to work. I agree that we should have better documentation on this.

[haruthakop]

Hi, Thanks a lot for your help. How did you test it? I also want to test it. Currently as a I2C slave I'm using an EEPROM. Do you have any idea how can I test bit level stretching with EEPROM?

Thanks again

[MartinBL]

I guess you can test it using a device which only tolerates low clock speeds, e.g. 100 kHz. Then you can try to run the TWI on your nRF51 at 400 kHz and see how the slave behaves. I actually concluded it works based on an unrelated test I have done and a brief talk with the TWI designer (who hadn't tested bitwise stretching specifically). What I did was actually testing what would happened if I used too large pullup resistors on the TWI lines. More specifically, what would happen if you only used the nRF51's internal 12 kOhm pullup resistors. What happened was that the device was not able to drive the lines as fast as I wanted and reduced the clock speed from 400 to 200 kHz (basically stretching the clock signals) due to slow rise times on the TWI lines. Except from the reduced clock speed it worked fine, so I concluded that bitwise clock stretching works.

[haruthakop]

Hi thanks for your help.

Software controlled TWI Master driver supports bit and byte level clock stretching (It can be clearly seen from source code). I tested both (Software controlled and TWI hardware) implementations by delaying ACK bit, both were working.

Thanks again for your help.