Podcast sample rate

[RS]

>From: Jim web
>Sent: Wednesday, July 19, 2017 14:33

>I've just been told unofficially that this is a 'legacy' issue. i.e. to
>ensure that as many types of device as possible can play them, even ancient
>kit.

Thanks, Jim, for making the enquiry.  Noone can quarrel with providing for 
backwards compatibility of the MP3 podcasts.

I wish the BBC would think more about backwards compatibility of AAC.  There 
is no problem with software players.  HTML5 and recent versions of VLC fully 
support HE-AAC v2 with LC (Low Complexity), SBR (Spectral Band Replication) 
and PS (Parametric Stereo).  The problem is with hardware players.

I am not alone in this list in having had difficulties playing AAC-LC at 
all.  When trying to find hardware players which support SBR and PS the 
problem gets worse.  Until Vangelis pointed out what it meant, I had been 
comforted by the glib statement in the AAC specification that if a player 
does not understand SBR or PS it should ignore them.  SBR and PS are 
presented as add-ons, but you don't get something for nothing.  Since the 
aim is to improve compression at low bit rates, the add-ons mean something 
else has to be taken away.

The BBC does not use PS but if a recording does include PS and the player 
does not understand PS, there is NO stereo information.  The result is mono 
playback.

SBR encodes high frequencies more efficiently.  Only half of the sampling 
rate is used for the LC component and half is used for SBR.  If the player 
does not understand SBR, instead of a roll-off at 14kHz, or 17kHz if you 
change the encoder settings, you get a roll-off at 7kHz or 8.5kHz.

That is what should theoretically happen.  Real life players make things 
even worse.  I have run the Fraunhofer tests
https://www2.iis.fraunhofer.de/AAC/stereo.html
on my satellite receiver and my SanDisk Clip Jam and the results were not as 
expected.

For the moment we have HLShd with 96kbit/s sound without SBR but that is 
described as a legacy provision.  At present we only have to resort to HVF 
with SBR for red button coverage of sporting events, where sound quality of 
the commentary is unimportant, and pre-HD repeats.  According to the BBC 
blog, Video Factory is the future.

I have not yet tried DVF.  As I understand it the video and audio are 
downloaded separately and then have to be stitched together with ffmpeg.  It 
may that in that process the audio could be transcoded back to AAC-LC.  Does 
anyone know?