Mask short audio dropouts

[manolito]

ProjectX has a nice audio feature called "Mask short audio dropouts by repeatedly writing the previous audio frame". But of course ProjectX can only handle TS captures in the MPEG2 format... :(

A few months ago DVB-T was replaced by DVB-T2 here in Germany (HEVC video and E-AC3 or AAC-LATM audio). While the video quality is much better and less sensitive to transmission glitches than the old DVB-T format, there is a problem with short audible audio dropouts. The broadcasts mostly have 2 audio tracks, one is 6-ch, the other one 2-ch, and the dropouts are far more frequent in the 6-ch track.

This seems to be a general problem and has nothing to do with the receiver (a lot of forum posts indicate this).

Now my question:
Does anyone know an audio repair software or plugin which can automatically detect and fix such short dropouts?

Of course I know how to do this manually in a WAV editor, but this would be very time consuming.


Cheers
manolito

[tebasuna51]

These audio dropouts are produced by corrupt audio frames.
When there are a corrupt audio frame always exist the possibility of a audible glitch, but can be repaired or at least minimized.
There are 3 levels of repair:

1) Repair audio inside a container.

With the help of timecodes and indexes in container, a detected corrupt audio frame can be replaced with the previous one or silenced.
Seems is like ProjectX work .
Also eac3to can do this when detect gaps/overlaps, and work fine with some audio formats, but, despite eac3to was created to manage EAC3 format in EVO files (old HD-DVD), don't work with new EAC3 format used by DVB now.

I don't know a tool than work at this level with DVB-T2 captures (TS container also?)

2) Repair audio extracted from container.

a) If the audio track is extracted rejecting the corrupt audio frames we can finish with a correct track but with sync problems, and now without timecodes can be reapaired. If there are only a few corrupt frames maybe is unnoticeable.

b) If the audio track is extracted including the corrupt audio frames we finish with a corrupt audio track.
At this level there are some tools than can repair the track.
DelayCut can work with some formats but not with the EAC3-DVB (madshi mod support EAC3-EVO format).

The EAC3 format have CRC fields than can detect corrupt frames but I don't know a tool to repair them.

3) Repair audio decoded to PCM

a) If the audio decoder reject the corrupt frames we can finish with with a short audio like precedent 2-a)

b) If the audio decoder try decode a corrupt frame (by design or because can't detect the corruption) we finish with the glitches than you detect.
At this level is when a audio editor can repair it.
There are some tools inside editors than can help (Click Removal in Audacity, etc.), but not always 100% accurate.

[manolito]

Thanks very much for the detailed explanation... :thanks:



With the help of timecodes and indexes in container, a detected corrupt audio frame can be replaced with the previous one or silenced.
Seems is like ProjectX work .
I don't know a tool than work at this level with DVB-T2 captures (TS container also?)

My XORO receiver saves the clips in an MTS container. But I usually process it first with TSDoctor which converts it into TS.

The weird thing is that these audio dropouts are present in the recorded MTS file, but watching the TV content directly then I hear no dropouts. I use the analog TV input so the receiver has to decode audio to analog on the fly. Looks like the receiver tries to repair the dropouts in this case, but not when playing the recorded file.


At this level is when a audio editor can repair it.
There are some tools inside editors than can help (Click Removal in Audacity, etc.), but not always 100% accurate.
Since I always convert audio to WAV first (for cutting in StaxRip via VDubMod) this would probably be my workflow. I do have a couple of WAV editors (WaveLab, Audacity, Acoustica, Adobe Audition) installed plus some older VST plugin packs (mainly the Waves plugins), but so far the declickers have not worked for these dropouts. The closest I got so far was automatic silence detection, then manually fix the glitches. But this does take a lot of time...


Cheers
manolito