It is not very clear to us from the comment that when the unusual delay happens and what the actual issue here is.
Is the primary issue the write failure or the delay after the address byte?
Is there any correlation that they found between the write failure and this delay? Or are both independent issues?
Sorry, if I did not state that clearly in my posts.
The primary issue is that i2c write fails. The observed delay after the address byte is just a symptom.
Regarding the correlation: When I analyzed some i2c traffic recording I got the following results:
- For successfull i2c transactions the duration varies between 105 us and 3.19 ms. The difference in the durations is always caused by delays after sending the address byte.
- All of the failed transactions have a delay of 12.9 to 22.1 ms after the address byte
So there is a very clear correlation between “delay after address byte” and “failed i2c write”.
What are the steps they follow to reproduce the issues?
As written in previous posts in this thread, the fact that the time between sending the address byte and the first data byte varies can be reproduces with a simple i2cset loop.
For reproducing the failed write: I was only able to reproduce that by running our software that communicates with the camera for 1-3 hours.
Could you dump the REG after address byte is acked? From I2C base address up to maybe 0x100?
What REG are you referring to?
(As our workaround to just retry the i2c write when it fails is working very well, I unfortunately cannot invest bigger amounts of time into this topic anymore.)