[asterisk-users] Two phones fail to agree on codec, asterisk at fault?

Brent Davidson brent at texascountrytitle.com
Fri Mar 28 15:49:12 CDT 2008


With canreinvite=no you are forcing asterisk to remain in the call 
path.  As long as Asterisk is in the call path, it is supposed to be 
transcoding the calls, so it doesn't care what the compatible codecs are 
between then endpoints.  Each leg of the call is phone<->asterisk so 
asterisk negotiates a compatible codec set with each phone.  If there is 
a codec difference between two legs of a call, it should be transcoding 
between them, unless you have that disabled somehow.  (A quick google 
and I don't see how to disable transcoding apart from limiting codecs.)

Now the other issue here is why Asterisk is offering GSM to the 
softphone and g726 to the C450IP.  Try setting the allow and disallow 
settings for each channel rather than in Global.  I tend to set things 
like codecs on a per-device basis rather than in global.  Global 
settings have a bad habit of being overridden.

Good luck,
Brent

martin f krafft wrote:
> Hi list,
>
> I am faced by a situation where I am trying to make a softphone and
> a Siemens C450IP talk to each other. Both are hooked up directly to
> the same asterisk, in the same IP net. 
>
>   - a softphone runs on 192.168.14.3
>   - the C450IP is at 192.168.14.30
>   - asterisk runs on the machine known as 192.168.14.1
>
> I am running Asterisk 1.4.11, backported to Debian Etch by Xorcom.
>
> If I set canreinvite=yes for both, everything works. However, I have
> reason to use canreinvite=no for both. But if I do, then the two
> phones fail to agree on a codec.
>
> So calls are going via an asterisk bridge and the symptoms of my
> problem are:
>
>   1 if C450IP calls softphone, they can talk fine
>   2 if softphone calls C450IP, voice only goes from C450IP to
>     softphone, not the other way around.
>
> I traced this down to the session description protocol, where there
> is funky stuff going on with the supported codecs each peer
> announces. Remember, asterisk is between them, and I set
> disallow=all,allow=ulaw,allow=alaw in [global].
>
> So in situation 1, when the C450IP calls the softphone, these codecs
> are announced. 0 is ulaw, 8 is alaw, 111 is g726-32, 3 is gsm.
>
>   C450IP to asterisk: 8, 0
>   asterisk to softph: 8, 3, 0
>   softph to asterisk: 8
>   asterisk to C450IP: 8, 0
>
> They both agree on 8 (alaw) and stuff is working, but it's already
> curious how asterisk adds the 3 (GSM) in the second line and the
> 0 (ulaw) in the last.
>
> In situation 2, no voice travels from the softphone to the C450IP,
> and this is the dialog:
>
>   softph to asterisk: 8, 0, 3
>   asterisk to C450IP: 0, 8, 111
>   C450IP to asterisk: 0
>   asterisk to softph: 3, 0, 8
>
> Again, notice how asterisk basically ignores what it was asked to
> relay. In the end, the softphone settles for 3 (GSM) but the C450IP
> chooses 0 (ulaw). Since the softphone has no problem decoding ulaw,
> it can hear whatever the C450IP transmits, but it returns GSM
> packets, which the C450IP can't decode, and therefore nothing comes
> out of that phone.
>
> What's going on here? From all I can tell, the clients do the right
> thing, each selecting the first codec offered by asterisk (which
> they support), but asterisk is going a bit lala here, isn't it?
>
> First of all, why does it even bother with 3 and 111, given how
> I disallowed them? And second, why does it *dare* to announce more
> than what is available to the peer to which it relays?
>
>   
> ------------------------------------------------------------------------
>
> _______________________________________________
> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>
> asterisk-users mailing list
> To UNSUBSCRIBE or update options visit:
>    http://lists.digium.com/mailman/listinfo/asterisk-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.digium.com/pipermail/asterisk-users/attachments/20080328/0748ccff/attachment-0001.htm 


More information about the asterisk-users mailing list