I've got a simple I2C interface design in prototyping, and due to the slow speed of our I2C implementation (10 KHz max) and short runs, I opted to use the internal 10K pullups in this part. However, in prototyping, with OE pulled high and no loads on inputs or outputs, the pullups only produce an approximate 2.2 volt output, which is in error, as the pullups are supposed to go to the rail of the appropriate side. I can't find if this part has any errata on it, and the behavior I report has been observed with more than one part. The basic design is a 3.3V (VCCA) to 5.2V (VCCB) translator, with a TCA9535 as the only device on the 5.2 volt side. The power supplies are zener-clamped in the prototype phase, and have not been observed to overshoot anyway. OE is tied to 5.2V. Unfortunately, I did not include optional pullups due to space availability, and leaving the TCA9535 I2C inputs at 2.2 volts causes the device to pull some 30 ma from the 5.2 volt rail, which is not the case if the inputs are pulled fully up to 5.2V! Even with the TCA9535 removed (no device on either the 3.3 or 5.2 volt rail I2C bus), the pullups only pull to 2.2 volts on both sides! Logically, minus the pullup issues, the device does pass logic signals, just at the wrong voltage levels due to the pullup voltage. Our devices were samples, marked with "NFER", in the VSSOP package.
↧