[Freeswitch-users] garbled audio with G726-32, other codecs are fine

Ivan ivan at c3i.bg
Tue Aug 13 10:57:50 MSD 2013


Ah, I didn't know there was always bit packing - I guess I should read 
how G726 works, thanks for pointing that.

The thing is, both linksys PAPs and linphone have the same problem, and 
they are not "exotic" endpoints so that makes me think the cause is my 
freeswitch setup. I'll try to test with Xlite when I have a chance, and 
also see if wireshark shows enough codec detail to find out if the 
endpoints get it wrong. I'm of course interested if you have other ideas 
on how to debug that.

ivan


On 08/12/2013 05:14 PM, Michael Jerris wrote:
> There is always bit packing, but there are 2 different ways to do the bit packing.  A lot of devices get it wrong so its worth looking at that.
>
> Mike
>
> On Aug 4, 2013, at 1:53 AM, Ivan Mitev <imitev at c3i.bg> wrote:
>
>> Thanks for the suggestion but I'm testing with G726-32, not AAL2-G726-32
>> ; so bitpacking shouldn't be used. By the way when I tested with AAL2 to
>> the linphone client I only got cracks and whitenoise, I've forgot to
>> mention that in my post.
>>
>> That said I've tried to uncomment and set <param name="bitpacking"
>> value="none"/> in internal.xml ("none" is a wild guess - I couldn't find
>> any doc on values accepted by this parameter), but that doesn't help.
>>
>> Speex: the ATAs don't support it. And being stubborn I'd like to
>> understand what's wrong with G726 :)
>>
>>
>> On 08/04/2013 05:22 AM, Jeff Leung wrote:
>>>
>>> You can turn off G726 AAC bit-packing in spandsp.conf.xml.
>>>
>>> By the way, there are other codecs out there you can try. SPEEX comes
>>> to mind if all your endpoints don’t deal with the PSTN.
>>>
>>> *From:*freeswitch-users-bounces at lists.freeswitch.org
>>> [mailto:freeswitch-users-bounces at lists.freeswitch.org] *On Behalf Of
>>> *Brian Foster
>>> *Sent:* Saturday, August 3, 2013 2:37 PM
>>> *To:* FreeSWITCH Users Help
>>> *Subject:* Re: [Freeswitch-users] garbled audio with G726-32, other
>>> codecs are fine
>>>
>>> AAC bitpacking by any chance? I thought I had a similar issue,
>>> happened so long ago I cant remember what I did.
>>>
>>> Thank you,
>>>
>>> Brian Foster
>>> Project Manager/Owner's Rep.
>>> Davri Investments, Inc.
>>> O: 317-787-2686 x2102
>>> M: 317-600-9753
>>> E: bdfoster at davri.com <mailto:bdfoster at davri.com>
>>> Indianapolis, Indiana
>>>
>>> Sent from a mobile device.
>>>
>>> On Aug 3, 2013 5:20 PM, "Ivan Mitev" <imitev at c3i.bg
>>> <mailto:imitev at c3i.bg>> wrote:
>>>
>>> Hello
>>>
>>> I'm migrating an office setup from asterisk to FS and in the process I
>>> was considering using G726-32 for some bandwidth starved remote
>>> endpoints. However I only get metallic/garbled audio with that codec
>>> even when simply playing moh to the endpoint, while other codecs work
>>> fine (G711U/A, G722, GSM). G732-16 is inaudible, G732-40 sounds
>>> marginally better but still garbled and really worse than G711.
>>>
>>> The setup is FS 1.2.12 from FS' yum repo on a centos6 64bit KVM guest
>>> (centos6 64bit host). But please don't shoot ! :) - I know about virtual
>>> environment limitations but for these tests the host is only lightly
>>> loaded, there aren't any calls to the FS instance except my tests, and
>>> the fact that it works with other codecs makes me think that
>>> virtualization is not the issue here. I may be wrong though.
>>>
>>> Is there any guide for debugging that kind of problem before reverting
>>> to a fresh install on bare-metal with the latest HEAD ? Until now I've
>>> tried:
>>>
>>> - improving timers ; but the default soft timer (which I guess uses
>>> timerd) works best. The time interval between sent packets on a tcpdump
>>> trace looks identical to the output of "timer_test", so that doesn't
>>> seem to be a network/jitter problem. And there's no problem with other
>>> codecs, but maybe G726-XX is specific. For info the guest's clocksource
>>> is kvm_clock, while the host uses tsc.
>>>
>>> - using different endpoints: the production ones are Linksys PAP2
>>> ("fixed" for 20ms psize, and G726-32 SDP type indentification), but the
>>> same thing happens with linphone on a fedora 19 laptop.
>>>
>>> A call with rtp media going through FS without transcoding - G726-32 to
>>> G726-32 - works perfectly (I can't hear the difference with G711). The
>>> problem is only when there's transcoding to G726 (from wav for moh, or
>>> from any other codec when bridging). I've looked at the wiki, posts,
>>> changelogs, jira, ..., but am a bit at a loss now.
>>>
>>> Any pointers ?
>>>
>>> Except that little problem, FS rocks, and I'm happy I can finally ditch
>>> asterisk. Kudos to the core devs and contributors.
>>>
>>> Ivan
>>>
>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> 
> 
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://wiki.freeswitch.org
> http://www.cluecon.com
>
> FreeSWITCH-users mailing list
> FreeSWITCH-users at lists.freeswitch.org
> http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
> UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
> http://www.freeswitch.org
>



Join us at ClueCon 2013 Aug 6-8, 2013
More information about the FreeSWITCH-users mailing list