I have a number of full-speed USB-powered devices I'm working with. All use STMicro microcontrollers (a few different models, including STM32F205, L432, etc.). All of the different devices use the same USB circuitry (including the 1.5k internal pullup on the D+ line), but for some reason on the latest build (700 pieces) of a particular device we're seeing intermittent USB enumeration failures. If I continually cycle USB power, I can get SOME devices to fail anywhere from 10-75% of the time, while some seemingly never fail. Using a USB protocol analyzer, I captured data both when one device did not enumerate and when it does.
Here's an instance of not enumerating:
As you can see, there (sometimes) seems to be an issue with the appropriate voltage during the K/J chirp sequence. However, I’m confused why on index #2 of the “bad enumeration�? capture the speed seems to be set to high speed without any J/K chirps having taken place yet. Further, in reading the USB Background article posted by the protocol analyzer manufacturer, it seems the Tiny K/Js I am seeing should be considered spurious and may foul communications. The say the cause is generally the pullup on the D+ line being released too slowly, which creates a voltage divider with the host termination resistors and causes a voltage mismatch. But I seem to see these both during successful and failed enumerations, so I am not sure if they are just a red herring. Lastly, if the micro being slow to disconnect its internal pullup is causing these spurious chirps, do you have any suggestions for how to improve robustness to avoid this?
Here's the USB circuit:
