[BUG][alsabat] glitches happened on MTL-HDA when testing headset recording

[keqiaozhang]

Describe the bug
I'm enabling alsabat test for headset recording and found some glitches on MTL-HDA. The reproduction rate is about 50% with all three test frequencies(599,821 and 997).

To Reproduce
~/sof-test/test-case/check-alsabat.sh -c hw:sofhdadsp,0 -p hw:CODEC,0 -C 2 -F 599
Need PR thesofproject/sof-test#1012

Reproduction Rate
50%

Environment

1. Branch name and commit hash of the 2 repositories: sof (firmware/topology) and linux (kernel driver).
   • Kernel: tip sof-dev
   • SOF: tip main
2. Name of the topology file
   • Topology: sof-ace-tplg/sof-hda-generic-4ch.tplg
3. Name of the platform(s) on which the bug is observed.
   • Platform: MTLP_RVP_HDA

[lgirdwood]

@keqiaozhang periodic issue with stream or just happens once in stream ? (if once when ?)

[keqiaozhang]

@keqiaozhang periodic issue with stream or just happens once in stream ? (if once when ?)

From my observation, this is not a periodic issue, it only occurs at the beginning.

[lgirdwood]

@keqiaozhang periodic issue with stream or just happens once in stream ? (if once when ?)

From my observation, this is not a periodic issue, it only occurs at the beginning.

@keqiaozhang Is the timing of the glitch from beginning always consistent ? @btian1 is copier clearing buffer at start ?

[keqiaozhang]

Is the timing of the glitch from beginning always consistent ?

Yes, it's consistent and occurs within the first 1-2 milliseconds.

[juimonen]

@keqiaozhang could you attach the mtrace also?

[lgirdwood]

@singalsu fyi - your glitch dtetector should be able to emit output if glitch is in initial periods (i.e. trigger start/release) errors.

[singalsu]

@singalsu fyi - your glitch dtetector should be able to emit output if glitch is in initial periods (i.e. trigger start/release) errors.

Currently it starts detect after stable signal level (waiting for ramp to complete) and analyze the FFTs after that. I would need to add a time domain glitch detector for the beginning part during ramp. The time domain is not as sensitive as FFT but it should work for glitches like this. I will try in next version for thesofproject/sof-test#1016

[lgirdwood]

@singalsu fyi - your glitch dtetector should be able to emit output if glitch is in initial periods (i.e. trigger start/release) errors.

Currently it starts detect after stable signal level (waiting for ramp to complete) and analyze the FFTs after that. I would need to add a time domain glitch detector for the beginning part during ramp. The time domain is not as sensitive as FFT but it should work for glitches like this. I will try in next version for thesofproject/sof-test#1016

Thanks - this will save some triage time as these are two quite different bugs with different solutions when they occur.