I'm working on Linux (Debian 11) audio recording application, in which precise timestamping of audio chunks is critically important.
Audio recording is done with ALSA. I use ALSA API snd_pcm_htimestamp() to obtain last position update hi-res timestamp (see https://www.alsa-project.org/alsa-doc/alsa-lib/group___p_c_m.html#ga3946abd34178b3de60fd5329b71c189b ). I read audio chunks of fixed duration and expect that difference between consequent timestamps will be close to audio chunk duration. But I see variations from chunk duration +- 0.0003 secs. This is not good enough for me.
Application runs on real-time priority.
Can be done something to reduce mentioned jitter ? Or this is inherent ALSA/kernel/Linux limitation ? Your help will be very appreciated.
Thanks