Error resolution or example for webrtcdsp with webrtcechoprobe

[Nicolas Dufresne]

Le mar. 26 mars 2019 08 h 55, ankit <ankit.patel1 at einfochips.com> a écrit :

> Hello Nicolas,
>
> > Maybe your ALSA driver isn't exactly of the quality you need ? Maybe
> > you didn't really configure 80ms the right way ? Please share a little
> > more information and we may be able to help.
>
> I am not getting this about quality. Current implementation works good
> without dsp implementation.
> Let me know in terms of what you are asking.
>

As for every embedded system, drivers aren't made for multiple purpose.
They are generally written in a rush for a specific use case. Whenever you
deviate from the original use case, it often fail or miss-behave. In this
case, with the echo canceller you now have a hard dependency on correct
latency reporting, and a round-trip latency being less the. 200ms.

Also, it may not be latency, maybe it picks a different a different rate or
format, and your driver might not be handling it properly. A large amount
of android audio drivers ignores the rate and always assume 48kHz as an
example.

You could also simply have a fpu that is too slow for this library.
Remember that this is a PC solution. So you need PC like fpu to make it
work.


> > I can suggest trying the experimental delay-agnostic feature, but
> > without context, I'm proposing you some plaster which may lead to sub-
> > optimal performance. To validate your latency, I strongly recommend
> > using audiolatency element.
> >
> > It works better when you wire the signal back in a close loop, but
> > should also work with a mic facing speaker. You simply run this, and it
> > will measure the roundtrip latency for you (with the appropriate setup
> > added of course).
> >
> >   alsasrc ! audiolatency ! alsasink buffer-time=80000
>
> As I am working on openembedded(yocto) platform which don't provide
> audiolatency plugin.
> I also tried to enable delay-agnostic in webrtcdsp but output result is the
> same with echo.
>
> Let me know if more information required.
>

It is really easy to bump your GST version in Yocto, so that should not be
an argument to not use the latest debugging tools available. audiolatency
have not external depency, you could even build it on the side using a
simple gcc call in its sub-directory.


> Regards,
> Ankit
>
>
>
> --
> Sent from: http://gstreamer-devel.966125.n4.nabble.com/
> _______________________________________________
> gstreamer-devel mailing list
> gstreamer-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/gstreamer-devel
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/gstreamer-devel/attachments/20190326/e6e71f8e/attachment.html>