Noise floor and digital black

[Soundmatters]

Factor out any digital black in the audio data from the calculation of noise floor level.

[dericed]

so any audio frame that is filled with 0 values should be omitted? or every 0 sample? This may add some processing time as I'd have to run another astats instance but filter out the 0 samples.

[Soundmatters]

Not sure about this. What would cause the Noise_floor_dB in the csv to report a value of "=-inf" I thought that "=-inf" implied 0 sample value, but after checking a few examples (w/ Audition's Amplitude Statistics) that does not seem to be the case.

[Soundmatters]

In case you've not seen it, here is an example of that reporting:

[richardpl]

Try latest ffmpeg version from git, it should have fixed noise floor calculation, no longer uses histogram approach, and was limited to 8192 bins - which caused that very small noise floors would be reported as -inf while in fact they where not 0.

[retokromer]

@richardpl Do you know if a new version of FFmpeg will be released soonish?

[richardpl]

Not really, there was plan to release 6.1 in July but overall situation and morale is pretty low since "introduction to mailing list/pushing for committing" of avradio/(Software Defined Radio) directly into FFmpeg.

[retokromer]

Thank you for the information! This confirms my feelings.