I2C_1 issue with nRF5340

[GiulianoFranchetto]

Hi,

We just received a custom board featuring a nRF5340.

We have some strange behaviour on I2C1 (twim1). We have multiple sensors on the bus, and only some of them answer the master packets.

The i2c_xxx function of Zephyr returns error "address not acknowledged" (nrfx driver error), meaning that some sensors are not detected.

We are using a 100kbps I2C bus.

We know that there are several errata related to TWI for the nRF5340, and I would like to know if they could be the cause of some sensors responoding and some not responding on the same I2C bus.

We have checked the hardware, everyone is well soldered, no inversion during assembly, no voltage drop or strange leakage current.

None of the sensors have duplicate addresses.

Moreover, as all the sensors are on the same bus, if one of them would have locked the bus, the other should not work at all.

Do you have any idea where this issue could come from? Any workarounds?

Regards

Giuliano

[haakonsh]

TWIM: First clock pulse after clock stretching may be too long or too short Might be a culprit. 

Have you scoped the bus with a logic analyzer? I suggest you look for these clock stretching violations, and any other anomalies. 

[GiulianoFranchetto]

We don't have small-enough probe for now. We are wiating for them to analyze the I2C probe

[GiulianoFranchetto]

I think, we have found the issue.

For an unknown reason, the function i2c_burst_write() in Zephyr does not work with the nRF53.

It worked on the nRF52840.

Replacing it by N i2c_reg_write_byte() functions makes it work.

Is there any reason why this function is not working properly with nrf53?

[haakonsh]

I don't know yet. Either way I suggest you try the nrfx drivers for peripherals on the nRF53 as for now. They are more mature.