LyraT 4.3 records peak noise with rare chance in repeating recording
Posted: Wed May 29, 2019 1:13 pm
Hi,
I am testing with a LyraT 4.3 board, and running example pipeline_wav_sdcard. Firstly the origin example was running great and record one .wav file each time I ran it.
Then I wrapped it to record repeatedly with an increasing number as .wav filename. Something strange was discovered that with a small chance, the L or R channel data may be corrupted with pike noise. Say I ran for 1000 recording files, I might get 20 to 25 files with corrupted data, either L or R. I never see L and R corrupted at the same time.
Some investigation has been done. I am pretty sure the whole modification is just for wrapping recording into a separate task, and not relative to pipeline itself. Looking into the .wav data, it seems the corrupted voice data caused by slipping 1 bit, and then LSB of one is regarded as MSB of the next.
Did someone ever see this phenomenon, or even have an answer/solution in place? Thanks a lot.
I am testing with a LyraT 4.3 board, and running example pipeline_wav_sdcard. Firstly the origin example was running great and record one .wav file each time I ran it.
Then I wrapped it to record repeatedly with an increasing number as .wav filename. Something strange was discovered that with a small chance, the L or R channel data may be corrupted with pike noise. Say I ran for 1000 recording files, I might get 20 to 25 files with corrupted data, either L or R. I never see L and R corrupted at the same time.
Some investigation has been done. I am pretty sure the whole modification is just for wrapping recording into a separate task, and not relative to pipeline itself. Looking into the .wav data, it seems the corrupted voice data caused by slipping 1 bit, and then LSB of one is regarded as MSB of the next.
Did someone ever see this phenomenon, or even have an answer/solution in place? Thanks a lot.
