New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 3929 link

Starred by 4 users

Issue metadata

Status: WontFix
Owner:
Closed: Jan 2017
Cc:
Components:
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

AEC on Mac can be put in non-causal state

Project Member Reported by bjornv@webrtc.org, Oct 17 2014

Issue description

Run a Hangout on a MacBookPro without headset and the other end will get echo bursts.

There are multiple things causing this and they are gathered in https://code.google.com/p/webrtc/issues/detail?id=3800

When investigating the AEC delay in a recording it is seen that occasionally the delay becomes negative, hence the AEC is put in a non-causal state. This has as an effect that the AEC completely fails, since it can't estimate the echo path.

I've attached a plot of this to show it. It is a 3 second audio snippet of a recording. After 1 second the delay becomes negative for the duration of roughly 200 ms.
 
non-causal-delay-in-aec.png
165 KB View Download
Project Member

Comment 1 by tina.legrand@webrtc.org, Oct 30 2014

Labels: EngTriaged IceBox
Project Member

Comment 2 by bjornv@webrtc.org, May 26 2015

Cc: bjornv@webrtc.org
Project Member

Comment 3 by bjornv@webrtc.org, Jun 9 2015

Cc: henrik.lundin@webrtc.org
Owner: ----
Status: Available
Project Member

Comment 4 by bjornv@webrtc.org, Jun 30 2015

Cc: -bjornv@webrtc.org
Project Member

Comment 5 by tina.legrand@webrtc.org, Nov 24 2015

Owner: p...@webrtc.org
Per, this is an old issue, but interesting findings I think you should be aware of.
Project Member

Comment 6 by henrik.lundin@webrtc.org, Oct 5 2016

Components: Audio
Project Member

Comment 7 by henrik.lundin@webrtc.org, Oct 5 2016

Components: -SignalProcessing
Project Member

Comment 8 by tina.legrand@webrtc.org, Nov 16 2016

Status: Assigned (was: Available)
Project Member

Comment 9 by p...@webrtc.org, Jan 23 2017

Labels: -Pri-2 Pri-3
I've never seen this occurring and I know of nothing in the code that can cause this apart from the AEC code that adjusts the delay in the AEC render buffer. That code is not at all straightforward to follow and is in need of refactoring and proper testing.

However, since the current plan is to roll out AEC3, which includes a fully rewrite of the render buffer delay adjustment, the plan is to be able to close this issue once AEC3 is fully launched.
Project Member

Comment 10 by henrik.lundin@webrtc.org, Jan 23 2017

Status: WontFix (was: Assigned)

Sign in to add a comment