[FFmpeg-devel] Swresample strangeness

[FFmpeg-devel] Swresample strangeness

[Marton Balint]

Hi,

I am working on a patch to select a better default packet size for the wav 
demuxer, since it reads at most 4096 bytes which can cause insanely small 
packets for files with high channel count.

Unfortunately some swresample fate tests blew up, and I wonder if the 
result of the resampling should be dependant on size of the audio chunks 
which are being fed to it. My assumption is that if it causes 
audible difference in the output, then it is likely a bug in swresample, 
but I am not familiar with its code and its many tunable variables, so if 
someone could confirm, (or even better, fix it :)) that would be great.

Here is a simple command line which should produce silence, but there is 
an audible regular clicking instead:

ffplay -f lavfi -i \
"sine=440:r=8000:samples_per_frame=2048,aresample=44100:filter_size=1:phase_shift=0,aresample=8000,aformat=cl=mono[a];
  sine=440:r=8000:samples_per_frame=320, aresample=44100:filter_size=1:phase_shift=0,aresample=8000,aformat=cl=mono[b];
  [a][b]amerge,pan=mono|c0=c1-c0"

Thanks,
Marton
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Re: [FFmpeg-devel] Swresample strangeness

[Michael Niedermayer]

[-- Attachment #1.1: Type: text/plain, Size: 1369 bytes --]

On Sun, Feb 25, 2024 at 05:38:27PM +0100, Marton Balint wrote:
> Hi,
> 
> I am working on a patch to select a better default packet size for the wav
> demuxer, since it reads at most 4096 bytes which can cause insanely small
> packets for files with high channel count.
> 
> Unfortunately some swresample fate tests blew up, and I wonder if the result
> of the resampling should be dependant on size of the audio chunks which are
> being fed to it. My assumption is that if it causes audible difference in
> the output, then it is likely a bug in swresample, but I am not familiar
> with its code and its many tunable variables, so if someone could confirm,
> (or even better, fix it :)) that would be great.

From what you describe, the first thing that comes to mind are timestamps.
If you resample with timestamps that mismatch the exact values expected from
the sample rate. You will have different resuls when the packet size is changed.
Its the same data but at a different point in time and thats different then.

I have too many things to do though so i cant really look into this.

thx

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Take away the freedom of one citizen and you will be jailed, take away
the freedom of all citizens and you will be congratulated by your peers
in Parliament.

[-- Attachment #1.2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

[-- Attachment #2: Type: text/plain, Size: 251 bytes --]

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Re: [FFmpeg-devel] Swresample strangeness

[Marton Balint]

On Sun, 25 Feb 2024, Michael Niedermayer wrote:

> On Sun, Feb 25, 2024 at 05:38:27PM +0100, Marton Balint wrote:
>> Hi,
>>
>> I am working on a patch to select a better default packet size for the wav
>> demuxer, since it reads at most 4096 bytes which can cause insanely small
>> packets for files with high channel count.
>>
>> Unfortunately some swresample fate tests blew up, and I wonder if the result
>> of the resampling should be dependant on size of the audio chunks which are
>> being fed to it. My assumption is that if it causes audible difference in
>> the output, then it is likely a bug in swresample, but I am not familiar
>> with its code and its many tunable variables, so if someone could confirm,
>> (or even better, fix it :)) that would be great.
>
> From what you describe, the first thing that comes to mind are timestamps.
> If you resample with timestamps that mismatch the exact values expected from
> the sample rate. You will have different resuls when the packet size is changed.
> Its the same data but at a different point in time and thats different then.
>
> I have too many things to do though so i cant really look into this.

Ok, thanks anyway. For the async tests the timestamps were randomly 
generated, and that changed. For other tests however I am quite 
sure about the swresample bug, so I will send a patch soon to fix that.

Regards,
Marton
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".