[Freeswitch-users] FS6947 - Tuning Opus bandwidth

Giacomo Vacca giacomo.vacca at gmail.com
Wed Oct 26 13:01:47 MSD 2016 

> As far as an Opus manual, I also love the idea and can't wait to read it.

This is the document trying to address the need for an "Opus manual":
https://freeswitch.org/confluence/display/FREESWITCH/FreeSWITCH+And+The+Opus+Audio+Codec

As mentioned there, questions and other feedback are welcome.

It can be easily exported from Confluence as PDF, should it be desired.

Giacomo



On 14 October 2016 at 18:48, Emrah <lists at kavun.ch> wrote:

> Hey there,
> Thanks for this extensive response. I am not sure I got everything, but
> I'll try to answer what I can.
>
> Why would you say opus at 8000@20i does not sound very good ? What do you
> mean exactly ? Your call quality is bad or you mean you don't like the
> fact that the sampling rate is low ?
>
> Opus at 8khz doesn't give you the same fidelity as PCM at 8khz. You
> definitely hear the compression. However, Opus at 48khz transcoded into
> PCMU / PCMA is hardly noticeable. Therefore, I wanted to find a decent
> compromise, one that would save bandwidth and CPU for calls that don't
> require a full band audio.
>
> You can play with the sampling rates by setting maxplaybackrate and
> sprop_maxcapurerate and enabling asymmetric_samplerates (which is an
> experimental feature so far) in your opus.conf.xml  .
>
> That sounds very interesting, however I do not want to impose global
> limitations as many of my Opus calls are full band calls. Would these
> settings help me work on a per call basis?
>
> FS does not have 12 khz or 24 khz because they are not much used for
> Voip. As for opus @ 16 khz perhaps it will be added.
>
> I guess this reconnects with my observation above. The idea would be to
> use something slightly higher than 8khz, yet not as intensive as 48khz,
> just to get a clearer call for "PSTN like" connections.
>
> But if you use opus @ 48 khz and you just change maxaveragebitrate and
> maxplaybackrate accordingly you should get only WIDEBAND from the
> encoder anyway .
>
> Not sure I got that right. Could you elaborate more? Can this work on a
> per call basis?
>
> As for the issue why codec settings cannot be set from FS's dialplan , I
> think its a missing feature that affects other audio codecs too and I
> think it would be very useful.
>
> Yes, absolutely. This would definitely help a lot. Not only on the FS
> side, but also on how to involve the Opus stack on the client side.
>
> As far as an Opus manual, I also love the idea and can't wait to read it.
>
> Thanks again for this response,
> Emrah
>
> On Oct 14, 2016, at 2:27 PM, Dragos Oancea <dragos.oancea at athonet.com>
> wrote:
>
> Hi
>
> Why would you say opus at 8000@20i does not sound very good ? What do you
> mean exactly ? Your call quality is bad or you mean you don't like the
> fact that the sampling rate is low ?
>
> You can play with the sampling rates by setting maxplaybackrate and
> sprop_maxcapurerate and enabling asymmetric_samplerates (which is an
> experimental feature so far) in your opus.conf.xml  .
>
> FS does not have 12 khz or 24 khz because they are not much used for
> Voip. As for opus @ 16 khz perhaps it will be added.
>
> But if you use opus @ 48 khz and you just change maxaveragebitrate and
> maxplaybackrate accordingly you should get only WIDEBAND from the
> encoder anyway .
> The decoder should decode at any sample rate.
> We're working on a document (sort of manual) for the Opus module and
> hopefully it will be released soon.
>
> As for the issue why codec settings cannot be set from FS's dialplan , I
> think its a missing feature that affects other audio codecs too and I
> think it would be very useful.
>
> We needed opus at 8000hz for transcoding and I tried to explain some things
> here (see my comment at the bottom of the page ):
> https://freeswitch.org/confluence/display/FREESWITCH/mod_opus
>
>
> Basically if you do heavy transcoding to PCMA / PCMU which is 8000 hz
> you'll want to avoid resampling 48 khz <-> 8 khz  - we did tests and by
> avoiding resampling we were saving 20-30 % CPU .
>
> Regards,
> Dragos
>
>
>
> On 14/10/2016 14:02, Emrah wrote:
>
> Hi there,
> Revisiting this issue. I see that I can set my Opus codec with
> Opus at 8000@20i, but 8khz seems to be the only alternative profile I can
> use. I see only 2 extreme options when the module is loaded or unloaded
> that it's either 48khz, mono or stereo and packet size, or 8khz, mono or
> stereo and packet size. Can someone clarify why there is nothing in
> between? And what exactly this setting does? Opus at 8000h@20i definitely
> doesn't sound very good. I'd rather have a compromise for tough network
> conditions.
> Since these are parameters I can dynamically set on my dialplan, the
> question then becomes why can't I fully manipulate my Opus stack from
> the dialplan?
>
> Thanks!
>
> On Jun 2, 2015, at 6:23 AM, Emrah <lists at kavun.ch
> <mailto:lists at kavun.ch <lists at kavun.ch>>> wrote:
>
> Hi there,
> @Mike: yes, but in a commonsensical approach the Opus library on the
> client's side would resample and therefore optimize the codec and the
> bandwidth accordingly up to FS.
> @Julien, I saw the setting for Opus globally, but it defeats the
> purpose. I don't want to limit the bandwidth of Opus for all
> instances. I'd like to optimize Opus on a per call basis.
>
> Thanks for the replies
>
> On May 31, 2015, at 2:13 PM, Michael Jerris <mike at jerris.com
> <mailto:mike at jerris.com <mike at jerris.com>>> wrote:
>
> Side note, opening at the different rate I believe just makes the
> opus library do the re sampling instead of FreeSWITCH.
>
> On Sunday, May 31, 2015, Julien Chavanton <jchavanton at gmail.com
> <mailto:jchavanton at gmail.com <jchavanton at gmail.com>>> wrote:
>
>    Hi Emrah,
>
>    The settings exist but they are not available from the dialplan,
>    right now they can only be set globally .
>    https://freeswitch.org/confluence/display/FREESWITCH/mod_opus
>
>    You can control the bandwidth using maxplaybackrate and
>    maxplaybackrate this will control the local encoder and also adds
>    the corresponding FMTP parameters to the SDP to be used by the
>    remote encoder (if it does implement the following draft, the
>    draft is evolving but I think it as not changed)
>
>    https://tools
>    <https://tools/>.ietf.org/html/draft-ietf-payload-rtp-opus-11
>
>    Maybe something like :
>
>    maxaveragebitrate 24000
>    maxplaybackrate 8000
>
>    The discussion was getting slightly more complicated when we
>    where discussing about unnecessary resampling this was not a
>    problem but it was just adding extra load on the server.
>
>    On Sun, May 31, 2015 at 6:09 AM, Emrah <lists at kavun.ch
>    <javascript:_e(%7B%7D,'cvml','lists at kavun.ch');>> wrote:
>
>        Hi list,
>
>        I re-read FS6947 and don't understand how this problematic
>        was addressed and the issue fixed.
>        The scope is simple. There should be a setting in the
>        dialplan that allows downsampling of Opus for applications
>        that do not require the 48khz / 2 channels framework. I.e.:
>        terminating to the PSTN with Opus to take advantage of low
>        bandwidth and great PLC.
>        There seems to be a lot of confusion around bandwidth in
>        general there. It doesn't matter if the internal clock of the
>        device is always sampling at 48khz / 2ch. There are settings
>        that can facilitate a lower bandwidth consumption for
>        particular use cases, and it seems the reason it is not being
>        implemented in FS is just a matter of being confused about
>        the intent of the 48khz 2ch base.
>        Please revisit this issue. FS should allow tuning of Opus
>        audio / network bandwidth in the dialplan. It would optimize
>        greatly lots of use cases.
>        If I'm calling the PSTN, I'd rather have my client downsample
>        and stream at a lower bandwidth, even if my audio capture
>        would still be at 48khz / 2ch as per the RFC, and save on
>        bandwidth, than transcode the full 48khz spectrum into PCM on
>        my FS and minimize processing power on the client's side.
>
>        Jira here: https://freeswitch.org/jira/browse/FS-6947
>
>        Emrah
>        ____________________________________________________________
> _____________
>        Professional FreeSWITCH Consulting Services:
>        consulting at freeswitch.org
>        <javascript:_e(%7B%7D,'cvml','consulting at freeswitch.org');>
>        http://www.freeswitchsolutions.com
>        <http://www.freeswitchsolutions.com/>
>
>        Official FreeSWITCH Sites
>        http://www.freeswitch.org <http://www.freeswitch.org/>
>        http://confluence.freeswitch.org
>        <http://confluence.freeswitch.org/>
>        http://www.cluecon.com <http://www.cluecon.com/>
>
>        FreeSWITCH-users mailing list
>        FreeSWITCH-users at lists.freeswitch.org
>        <javascript:_e(%7B%7D,'cvml','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 <http://www.freeswitch.org/>
>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org <mailto:consulting at freeswitch.org
> <consulting at freeswitch.org>>
> http://www.freeswitchsolutions.com
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://confluence.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
>
>
>
>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://confluence.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
>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://confluence.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
>
>
>
> _________________________________________________________________________
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> http://www.freeswitchsolutions.com
>
> Official FreeSWITCH Sites
> http://www.freeswitch.org
> http://confluence.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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20161026/6cff58e9/attachment-0001.html

Join us at ClueCon 2016 Aug 8-12, 2016
More information about the FreeSWITCH-users mailing list