[Asterisk-Users] Re: G729 vs. gsm

[Steve Underwood]

Tony Mountifield wrote:

>In article <200505281615.01031.akohlsmith-asterisk at benshaw.com>,
>Andrew Kohlsmith <akohlsmith-asterisk at benshaw.com> wrote:
>  
>
>>You can convert between ulaw and alaw 10000 times and it'll sound exactly the 
>>same as the original 8000Hz 16-bit sample because... well... it is.
>>    
>>
>
>Just to pick a nit, the above can't be right. G.711 is a logarithmic
>recoding of 16 bits into 8. So each 8-bit value can be reached from
>a whole lot of different 16-bit values. The information as to which
>16-bit value it came from is lost, so you can't get back to the
>original values, just to approximations of them.
>
>Also, I believe the 16-bit range is chopped up in slightly different
>places for aLaw than for uLaw, which could mean a slight loss of
>information going back and forth repeatedly between the two.
>
>Cheers
>Tony
>  
>
It is actually 12 or 13 bits compressed to 8. As you say, it is lossy. 
However, after the first round trip, subsequent round trips give the 
same answer. This was a design requirement. Tandem operation was not 
allowed to give endless degradation. Even G.726 follows this rule. You 
can chain a mixture of G.711 and G.726 links, without the degradation 
building up. It is one of the reasons G.726 is so very complex for an 
ADPCM codec.

Regards,
Steve