Per advice in this post I am using a 74AHCT2G125DC (datasheet) to drive a Neopixel/WS2812b strip's data pin from a 3.3v MCU. The MCU is driving A1, and the addressable LEDs are chained from Y1. VCC on the high side is 5V and the ground is shared.
I'm seeing numerous communication errors and flickers on the LEDs although there is no voltage sag on the 5V rail. Oscilloscope probing the input side of the 74AHCT2G125DC buffer shows a clean signal between 0-3.3v. Probing the output side shows a much noisier signal ranging from 1.2-5.0V. It never reaches the ground which I suspect is causing some of the issues.
What is the best explanation and how do I fix this? My current theory is that somehow driving down the LED chip's pin to GND must sink more than the current limit of the driver (8mA, if I read it correctly.) Or am I misunderstanding that a driver of this sort can feed a digital signal out to a device like a WS2812b? Is there any explanation for why on a low signal, the output doesn't fall to zero?
Note: The reason for using a tri-state driver is so I can shift the digital out to a high-impedance state during circuit shutdown preventing an observed current backflow to the MCU when the strip is unpowered.