[Bug] Bosch BMP581 intermittent I2C read errors

[vincentpoont2]

Describe the bug

Bug report

Issue details

We are testing Bosch BMP581 as a replacement barometer on Pixhawk 6X due to ICP20100 and BMP388 being discontinued.

It seems like
1. BMP581 works reliably when connected on I2C2.
2. On I2C4, BMP581 intermittently produces corrupted samples, visible as large spikes in pressure or baro derived altitude.

baro0 (i2c4, BMP581 glitches) Baro1 (I2C2,BMP581, No issue)

Hardware type
Holybro PIxhawk 6X

Logs
https://logs.px4.io/plot_app?log=9f959489-aa7d-491e-909d-16570a3031b9

Flight Log / Additional Information

https://logs.px4.io/plot_app?log=9f959489-aa7d-491e-909d-16570a3031b9
Pixhawk 6X

[farhangnaderi]

@vincentpoont2 The issue is I2C4 bus contention between BMP581 (0x47), BMM150 magnetometer (0x10), and EEPROM (0x50). I2C4 is also configured as internal bus with different characteristics than I2C2.

If the I2C bus has timing issues during this 6-byte burst read:

A single corrupted byte causes massive spikes (e.g., if reg_data[5] is corrupted, pressure jumps by ~16 million Pa per bit error in MSB)

Can you try
bmm150 stop and see if it temporary solves? Then if yes, you might consider staying on I2C2 or we find another solution.

[jamming]

@vincentpoont2 The issue is I2C4 bus contention between BMP581 (0x47), BMM150 magnetometer (0x10), and EEPROM (0x50). I2C4 is also configured as internal bus with different characteristics than I2C2.

If the I2C bus has timing issues during this 6-byte burst read:

A single corrupted byte causes massive spikes (e.g., if reg_data[5] is corrupted, pressure jumps by ~16 million Pa per bit error in MSB)

Can you try bmm150 stop and see if it temporary solves? Then if yes, you might consider staying on I2C2 or we find another solution.

I just tried turning off BMM150 and reducing the ODR of BMP581 to 10hZ, which reduced errors (only one error occurs in over 20 minutes), but it was not completely eradicated.