alsa-utils uses incorrect format for 24 bit audio?

alsa-utils uses incorrect format for 24 bit audio?

[Ricard Wanderlof]

When attempting to use arecord to capture audio from a 24 bit source, and 
create a 24-bit .wav file I ran into the problem that the resulting file 
does not seem valid. Looking closer at the contents, it appears that 
arecord writes the 32-bit S24_LE samples verbatim to the file, and sets 
the number of bits per sample specifier to 24 in the .wav header, at the 
same time as it sets the 'block align' to 4 which I suppose is logical, 
although normally the block align corresponds to the frame size (in bytes) 
for the sample size specified in the bits per sample specifier. 

A number of utilities and applications I tried refused to play the file 
back properly, including Audacity, VLC, sox and kwave, so I'm guessing the 
format written by arecord is not valid in this case. 24 bit audio as 
exported from Audacity for instance in reality corresponds to the ALSA 
S24_3LE format, i.e. the data is sent with no padding (resulting in a .wav 
'block align' specifier of 3 for a mono stream).

Is this a known problem with arecord/aplay, or is it the other 
applications which have been less than accurate when implementing .wav 
file standard?

/Ricard
-- 
Ricard Wolf Wanderlöf                           ricardw(at)axis.com
Axis Communications AB, Lund, Sweden            www.axis.com
Phone +46 46 272 2016                           Fax +46 46 13 61 30

Re: alsa-utils uses incorrect format for 24 bit audio?

[Takashi Sakamoto]

Hi,

On Sep 6 2017 04:32, Ricard Wanderlof wrote:
> When attempting to use arecord to capture audio from a 24 bit source, and
> create a 24-bit .wav file I ran into the problem that the resulting file
> does not seem valid. Looking closer at the contents, it appears that
> arecord writes the 32-bit S24_LE samples verbatim to the file, and sets
> the number of bits per sample specifier to 24 in the .wav header, at the
> same time as it sets the 'block align' to 4 which I suppose is logical,
> although normally the block align corresponds to the frame size (in bytes)
> for the sample size specified in the bits per sample specifier.
> 
> A number of utilities and applications I tried refused to play the file
> back properly, including Audacity, VLC, sox and kwave, so I'm guessing the
> format written by arecord is not valid in this case. 24 bit audio as
> exported from Audacity for instance in reality corresponds to the ALSA
> S24_3LE format, i.e. the data is sent with no padding (resulting in a .wav
> 'block align' specifier of 3 for a mono stream).
> 
> Is this a known problem with arecord/aplay, or is it the other
> applications which have been less than accurate when implementing .wav
> file standard?

Would you please dump headers of RIFF/Wave format file with the captured 
samples? Like:

$ hexdump -C /tmp/test.wav | head -n5
00000000  52 49 46 46 70 31 00 00  57 41 56 45 66 6d 74 20 
|RIFFp1..WAVEfmt |
00000010  10 00 00 00 01 00 01 00  a0 0f 00 00 80 3e 00 00 
|.............>..|
00000020  04 00 18 00 64 61 74 61  28 31 00 00 00 00 00 00 
|....data(1......|
00000030  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00 
|................|
*

For confirmation, when capturing any samples, please add '-v' option to 
arecord command line to dump the type of handled samples. Like:

$ LANG=C arecord -v -D plughw:0,0 -t wav -f S24_LE /tmp/hoge.wav
Recording WAVE '/tmp/hoge.wav' : Signed 24 bit Little Endian, Rate 8000 
Hz, Mono
Plug PCM: Rate conversion PCM (48000, sformat=S24_LE)
Converter: libspeex (external)
Protocol version: 10002
Its setup is:
   stream       : CAPTURE
   access       : RW_INTERLEAVED
   format       : S24_LE
...


Regards

Takashi Sakamoto