[*] Latency on VoIP

[John Lange]

For the sake of interest, here is mine.

         g723   gsm  ulaw  alaw  g726 adpcm  slin lpc10  g729 speex  ilbc
   g723     -     -     -     -     -     -     -     -     -     -     -
    gsm     -     -     2     2     2     2     1     5     8     -    13
   ulaw     -     3     -     1     2     2     1     5     8     -    13
   alaw     -     3     1     -     2     2     1     5     8     -    13
   g726     -     3     2     2     -     2     1     5     8     -    13
  adpcm     -     3     2     2     2     -     1     5     8     -    13
   slin     -     2     1     1     1     1     -     4     7     -    12
  lpc10     -     3     2     2     2     2     1     -     8     -    13
   g729     -     3     2     2     2     2     1     5     -     -    13
  speex     -     -     -     -     -     -     -     -     -     -     -
   ilbc     -     4     3     3     3     3     2     6     9     -     -

However, I have dropped all compression from our system. We use only
g711u now end-to-end unless the client configures their device to do
otherwise.

John

On Tue, 2005-04-12 at 08:13 -0500, LES.NET (1996) INC. wrote:
> There was a discussion on asterisk-biz about something, and some people
> posted info on the translation time from different codecs..
> 
> 
> 
> The discussion on *biz was partly about codec CPU horsepower and
> bandwidth.  Because it takes one big machine to do say.. 96 concurrent
> ulaw->g729 conversions, and not so much of a machine to keep it ulaw.
> 
> Of course, your bandwidth costs go up without g729, but you may be able to
> cram more simultaneous channels in a single machine (thus a single amount
> of physical space (colocation space costs reduced)) that it may actually
> make more economical sense to use ulaw than to use g729 in some cases.
> 
> You can see by show translation (below), that converting from ulaw -> g729
> takes 27 milliseconds and then 6 milliseconds back to ulaw.   My current
> server is a Dual P3 1.13ghz, faster systems should translate quicker..
> 
> So I would assume, that having an asterisk box on either end, similarly
> equiped, would add about 33 milliseconds to the overall latency
> 
> pstn -> ulaw -> g729(+27/27ms) -> internet(+50?/77ms) -> g729 ->
> ulaw(+6/86ms) -> pstn
> 
> Here is my codec translations listing:
> astpbx*CLI> show translation
>          Translation times between formats (in milliseconds)
>           Source Format (Rows) Destination Format(Columns)
> 
>          g723   gsm  ulaw  alaw  g726 adpcm  slin lpc10  g729 speex  ilbc
>    g723     -     -     -     -     -     -     -     -     -     -     -
>     gsm     -     -     2     2     5     2     1     5    27     -    23
>    ulaw     -     5     -     1     5     2     1     5    27     -    23
>    alaw     -     5     1     -     5     2     1     5    27     -    23
>    g726     -     7     4     4     -     4     3     7    29     -    25
>   adpcm     -     5     2     2     5     -     1     5    27     -    23
>    slin     -     4     1     1     4     1     -     4    26     -    22
>   lpc10     -     7     4     4     7     4     3     -    29     -    25
>    g729     -     9     6     6     9     6     5     9     -     -    27
>   speex     -     -     -     -     -     -     -     -     -     -     -
>    ilbc     -     8     5     5     8     5     4     8    30     -     -
> 
> Anyone else want to post their show translations?
> 
> Les.
> _______________________________________________
> Asterisk mailing list
> Asterisk at muug.mb.ca
> http://www.muug.mb.ca/mailman/listinfo/asterisk
>