Speex dev - Mar 2009 - echo cancellation input/output delay

Hi,

I've been reading the speex echo cancellation documentation
and it says :

".... It is important that, at any time, any echo that is
present in the input has already been sent to the echo
canceller as echo_frame. ... ";



All I'm doing in the echo cancellation process is
initializing it once with a call to :

SpeexEchoState* speex_echo_state_init(int frame_size, int
filter_length)

and calling  speex_echo_cancellation(m_echoState, rec, play,
out) to remove the far-end echo from the input signal.


So here is my question:

if the first sample in the rec buffer refers to a guiven
time 't' the first sample in the play buffer is supposed to
refer to a guiven time 't - delta' ???

or in other words:

speex_echo_cancellation(m_echoState, rec(t), play(t-delta),
out) is correct ?

If so... how many millisencons this 'delta' is supposed to
be ?

Thanks in advance,
Daniel

Quoting danflu at uninet.com.br:
> speex_echo_cancellation(m_echoState, rec(t), play(t-delta),
> out) is correct ?
Yes and no. If your audio buffering has no delay, then having a positive delta
means the AEC won't see play(t) when it needs to cancel at rec(t), which is
bad. On the other hand, most audio hardware introduces a lot of delay, so
including your delta manes that you reduce that delay, which is OK. As long as
delta is never bigger than the audio delay (OS/hardware-specific).

   Jean-Marc
> If so... how many millisencons this 'delta' is supposed to
> be ?
>
> Thanks in advance,
> Daniel
> _______________________________________________
> Speex-dev mailing list
> Speex-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/speex-dev
>
>