My customer asks:
If the power supply is allowed to come up to ~250mV, then the circuit is turned on the THS7353 jams the I2C bus interface on the SDA line (pin 12). It's like it has come to the conclusion it should be driving its output. If clocks are provided it eventually lets go of the bus. Page 37 of the datasheet specifies having a single clock to avoid an issue, but adjusting our firmware to provide that single clock resulted in the issue being even more prevalent.
Leftover from earlier development, our circuit provides some extra I2C clocks at power up, I've attached scope shots of the problem in action when these extra clocks are issued (the first activity on the bus after power-up is this shot). the top trace is SDA, and the bottom is SCL. "bus behavior just after turn on.jpg" illustrates what's expected to happen, and "Bus latchup at turn on.jpg" illustrates the issue. If the THS7353 is lifted, the problem goes away so I am certain the SDA line is being driven by the THS7353 at this time. Also, the output waveform is not consistent, but the shot of the latchup is the most typical."
"bus behavior just after turn on.jpg" illustrates what's expected to happen:
"Bus latchup at turn on.jpg" illustrates the issue