AM62A7-Q1: I2C controller occasional clock offset.

YuWei Wu

Part Number: AM62A7-Q1

Hi,

We encountered an issue during our long-term testing. We have an I2C slave device, which is a deserializer device. During repeated restarts, we found that sometimes the clock and data sent on the I2C bus are abnormal.

The waveform of the I2C data we send under normal conditions on the I2C bus is as follows.

After repeatedly restarting the device, the following abnormal waveforms will appear.

It can be clearly seen that the clock corresponding to the ACK bit has experienced a severe offset.

We need TI's assistance to resolve this issue. The SDK version we are using is 10.01.

Thanks,

Yuwei

4 months ago

0 Stanislav Stilyanov 4 months ago

TI__Mastermind 25080 points

Hi Yuwei,

To clarify your use-case, can you please let us know for the below items:

What is the part number of the de-serializer (slave) device? A datasheet would help if it's not widely available.
What's the I2C pull-up resistors value? At 1.8 or 3.3V? A schematic snapshot related to I2C would be helpful
If you can capture CLK and DATA with an oscilloscope, it would be helpful to review the actual waveforms

Thanks,

Stan

0 Stanislav Stilyanov 4 months ago in reply to Stanislav Stilyanov

TI__Mastermind 25080 points

4. What's the I2C clock speed?

Regards,

Stan

+1 Stanislav Stilyanov 4 months ago in reply to Stanislav Stilyanov

TI__Mastermind 25080 points

5. Are getting error messages?

It looks like the slave i2c device is stretches the clock to keep up with the received data. It doesn't make sense why i2c controller sends wrong data after clk release, however.

Regards,

Stan

0 YuWei Wu 4 months ago in reply to Stanislav Stilyanov

Intellectual 390 points

Hi Stan:

My slave device address is 0x28 (7-bit address). The issue I mentioned is a low-probability sporadic problem. I suspect there is a bug in TI's driver, and I would like to know if there is a manual reset method available for the Linux I2C controller.

Stanislav Stilyanov said:
What's the I2C clock speed?

The clock speed of I2C is 400k.

Stanislav Stilyanov said:
If you can capture CLK and DATA with an oscilloscope, it would be helpful to review the actual waveforms

Since I did not use an oscilloscope to record the data under the error condition, I am currently unable to provide it. If you urgently need this information, I will try to reproduce the issue and then provide you with the oscilloscope data.

I will have my colleague answer other hardware-related questions for you.
It is worth noting that this is an extremely low-probability issue, occurring approximately once every 10 hours. Currently, I would like to know what might be causing this problem.

Thanks，
Yuwei

0 Stanislav Stilyanov 4 months ago in reply to YuWei Wu

TI__Mastermind 25080 points

Hi Yuwei,

Ok, I'm looking forward for hardware details from your colleague.

I'm not suspecting software problem at this moment, since clocking is something done in I2C controller hardware.

Regards,

Stan

0 liu xiwen 4 months ago in reply to Stanislav Stilyanov

Prodigy 180 points

Hi Stan:

The following is the schematic diagram of our IIC interface, and the voltage is 3.3V.The pull-up resistor is 2k.

0 Stanislav Stilyanov 4 months ago in reply to liu xiwen

TI__Mastermind 25080 points

YuWei Wu said:
If you urgently need this information, I will try to reproduce the issue and then provide you with the oscilloscope data.

Yes, this will be helpful.

for the 2k resistors , they are a bit strong for 3.3V, but I don't think this can cause issues until we can check waveforms.

Thank you,

Stan

+1 YuWei Wu 4 months ago in reply to Stanislav Stilyanov

Intellectual 390 points

Hi Stanislav,

We attempted to reduce the I2C communication rate from 400K to 100K, and this issue was resolved. However, the root cause of the problem remains unidentified. We hope that TI can pay attention to this issue and address it in future versions.

Thanks,

Yuwei

Processors

Processors forum

AM62A7-Q1: I2C controller occasional clock offset.