PCA9517: about B side SCL abnormal waveform when A side power is on

Joyce Li said:

Part Number: PCA9517

Hi dear expert,

My customer has used PCA9517 in their system and met below issue, all boards see the same phenomenon: 

Blue curve is B side SCL, pink is A side SCL. SCL signal is sending from B to A direction. B side is master, 3.3V, A side is slave, 1.8V.

Have tried to change pull up resistor, didn't help.

If A side 1.8V is not on, B side SCL wave form is normal. If A side 1.8V is on, B side SCL wave form will become below abnormal one.

Customer also says if signal direction is A to B, no problem found. 

Could you please help look at this issue?

[Bobby] So it looks like this is caused by the "unwrapping" effect of the 9517 (occurring when B side releases low and A side is not driving low). A side needs to go above a certain threshold (I believe ~40% of VccA) before B side will fully release(otherwise A side will temporarily drive B side to ~540mV) I suspect you can minimize this overshoot be making B side's pull up resistor weaker/larger and making A side's pull up resistor stronger/lower. If the customer can modify their resistors to minimize the glitch to be less than 50ns, then there shouldn't be a problem (I2C spec dictates all slave inputs must have a 50ns deglitch filter built in.)

Since PCA9517 is not recommended in new design, can we promote TCA9617 to replace? Customer also has TCA9617 AVL code.

[Bobby] Does the customer have TCA9517 or TCA9517A qualified for them to replace the device? Which TCA9617 do they have qualified (A version or B version?) I believe the 9617 does have this unwrapping as well but the prop delays should be faster but the device has quite a few differences compared to the 9517.

Below is the schematic:

Thanks a lot!

Joyce Li