Latency increase in Jetson Nano V4l2 timestamps over time

The timestamps reported by V4L2 seem to drift over time from the monotonic clock for a camera source I am using, about 500ms of drift after eight hours. Is this something to be expected? I don’t have too many capture buffers so I’m dequeueing them fast enough, but somewhere the system is accruing a few hundred ms of latency over time.

This seems to be expected. It may be due to accuracy of the oscillator. Do you hit certain issue from the offset ?

Yes, we’re using the timestamp to generate PTS for the video to be sent over the network. As a result we see the latency increase over long streams and I’m worried this will cause a buffer overrun at some point in addition to desyncing the audio and video. Is there a way to more reliably timestamp the buffers? I noticed the monotonic clock that ALSA uses doesn’t seem to drift.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.