Those gaps are way too big for it to be microschedule or xhci buffer-management related (the controller doing the buffer management, not the host). Assuming a nominal ADC wordsize, the biggest one translates to roughly 20-30ms of idle. That does strongly suggest the driver or host application starved the xhci controller for whatever reason.
Can you detect drops in real-time, and trigger an event capture based on that? (Run with event tracing enabled, then disable it when the first data byte gets dropped.)
Why doesn't Pi 4 suffer from the same?
What is the output of uname -a for both systems?
Can you detect drops in real-time, and trigger an event capture based on that? (Run with event tracing enabled, then disable it when the first data byte gets dropped.)
Why doesn't Pi 4 suffer from the same?
What is the output of uname -a for both systems?
Statistics: Posted by jdb — Thu Aug 08, 2024 5:46 pm