TCA9539: the concern about /INT

Hello Muk,

Yusuke Mukuno said:

Question 1
Customer is worried the possibility of missing read the input signal at following 2 scenarios.

1. When P01 changes from H to L after changing P00 from H to L. And P01 changes from L to H before Read
⇒ Will TCA9539 miss the signal of P01?

It depends on if the I2C read of port 0 bit 1 was performed before or after P0b1 changed. I believe it is possible to miss this if the timing is bad as the ACK will reset the 'previous' remembered state of the inputs (Q output of the flip flop) to the state during the ACK.

2. When P00 change from L → H before Read after changing P00 from H to L.
⇒ Will TCA9539 miss the signal of P01?

P01 shouldn't be missed unless it was read and then immediately after the data changed before the ACK occurred.

Question 2
It is assumed that the product enters SLEEP as a condition of use of customer.
SCK does not supply during SLEEP because it does not access the GPIO expander.
In this case, customer think, the following signal is not changed, because Read Pulse is not supplied.
・ Register Data
・ INT
In other words, will customer not notice changes in input?

The device does not use the clock signal to flip the INT. You can see on the INT NOT in the picture below that one input goes to the current state (GPI) and the other goes to a flip flop (previous state). The INT will still trigger even if the clock is not being pulsed.
Is SCK a necessary condition?

No, I believe the customer may have misunderstood how the INT logic works. Please look at the XOR gate inputs from the INT logic in the image below again.

If customer is concerned about missing input changes, may I suggest TCA8418/TCA8418E. The device has an event FIFO to store up to 10 input changes.

-Bobby