I'm having a strange issue on IMX8MP with the I2C bus.
- I2C used is I2C5 (muxed out from HDMI)
- The bus uses a TXS0102 1.8 to 3.3 shifter
- 10k pullups on the 3.3 side (+ internal 10k pullups on both that are bypssed by one-shot)
- Linux FSLC kernel
- 100 kHz clock (although the scope would lead me to believe it's 200 kHz?)
I've been having random bus glitches particularly with multple devices connected so I isolated the situation to a single slave connected with a short ~5 cm cable. The slave reacts correctly to enumeration, but after the 9th clock cycle the clock goes low and stays low for a while (I've seen values from less than 50 µs to 600+ µs). There's also a repeatable glitch on SDA during the low phase of the 8th clock cycle which may or may not be related to anything. Scope session probed at the 3.3v side:
After a while SDA is pulled low again (by who?), SCL is released high and eventually SDA is relesed too. Checked this with:
- Two different slaves
- Addressing a non-existing slave
- No slaves bus (just an empty connector)
I'll study it further, but if anyone knows what's up with this out of their head, would be greatly appreciated.
FET-shifted I²C behaves similarly too (CSI bus):
