Part Number: TXS0104E
I have implemented this four-channel level translator on numerous designs since 2017. It always worked until recently, and even now it only fails on one variation of the application it is used on - in some cases. The circuit simply translates the four SPI lines from a microcontroller from 3.3.V to 1.8V to match the levels of an imaging sensor SPI port. There is a FET that is driven from the Chip Select signal that inverts that signal and drives the Enable OE pin of the TXS0104. Chip Sel LOW = OE High, so that OE always follows CS and maintains a high impedance buffer on the shared SPI bus signals.
I am now seeing large voltage spikes (~1.2V) on the B-side inputs. The frequency is ~700KHz. They are only present when the input signals are low. When the spikes show up on the CS line they are high enough to drive the inverting FET, which in turn disables the OE pin momentarily. Then the cycle repeats. Sometimes the first cycle is clean, sometimes all four are noisey. CS = Yellow OE = Green
Zoomed in the signal looks like this:
If I hold the OE pin at ground, disabling the device, there are no spikes on any input signals.
If I hold the A4 (CS Out) pin at ground there are no spikes on the input signals.
My hunch is that the A4 pin is turning on and pulling the output high through the internal pull-up before the device has a chance to pull it low to match to input level. This in turn sets off a repeating process.
I replaced one device on one board with a device that was about 4 years old. It worked fine. I had six other boards sent back and the TXS0104 replaced. All six still do this.
The question is: WHY?!
