[asterisk-dev] iLBC packet loss concealment (was: code-cleanup concerns)

[Matthew Fredrickson]

On Apr 17, 2006, at 6:04 AM, Steve Underwood wrote:

> Denis Smirnov wrote:
>
>> On Mon, Apr 17, 2006 at 09:30:30AM +0800, Steve Underwood wrote:
>>
>> SU> Internally Asterisk uses 16 bit samples of signed linear audio at 
>> 8k SU> samples per second. This is 128kbps, not 64kbps. Right now any 
>> codec is SU> converted to this during transcoding. This is a bad 
>> thing. It ties SU> Asterisk to 4kHz bandwidth. It is something that 
>> needs fixing.
>>
>> transcode_via_sln, that used by default, is a VeryBadThing (tm)
>>
> I think very bad is overstating it. However, its not right. The ITU 
> specs define specific ways to go ulaw->alaw->ulaw and 
> ulaw->G.726->alaw->G.726->ulaw, and those are not followed. If they 
> were, tandem operation would be better. I'm not sure that is a huge 
> issue, though. Most of the transcoding is for things like G.729 or 
> Speex, where the only way is via 16 bit linear audio.
>

That's not QUITE correct.  The capability within the transcoding 
architecture to to do so, and direct alaw to ulaw is even supported.  
However, most codecs (g.726 in this case) don't currently provide a 
direct translation path to something other than signed linear.

Matthew Fredrickson