Hi Ryuji San,
I double checked with our digital design team on the timing spec.
To put it plainly, the T_hold MAX means “when (master) reading, please don’t try to change the state of SDA pin for at least this amount of time, because the slave can hold its state for a MAX of this amount of time”; T_hold MIN means “when (maser) writing, please hold SDA in current state for at least this amount of time before changing its state”.
When the master writes to a slave device, the SDA can be held (almost) indefinitely after the clock falling edge, as long as it does not interfere with the next data bit. In this case the T_hold MIN applies and T_hold MAX does not. On the other hand, when master is reading back from the slave, the master must wait at least T_hold MAX amount of time before driving SDA. The MIN and MAX specs target slightly different aspects of the communication. T_hold MAX is really the data valid time.
In the timing you provided, for WRITE, the setup time and hold time are both 1.6uS which is fine. For read, Tpd is guaranteed to be 900nS or less by the spec as shown here. And you have 3.2uS CLOCK low before the read edge, which is also plenty.
Regarding why the data sheet was updated, here is the reason. We noticed some customers were not paying much attention to timing, the clock and data edges were often placed together, with 0 setup/hold time. As a result, we had quite a few inquiries/requests on this matter. To help future customers, we updated the data sheet, emphasizing the importance of satisfying the timing requirements.
Please do let me know if the customer has further questions/concerns, we’ll do our best to assist.
Regard, Guang
Apps-Sensing Products
