[mpeg2] Re: kfir latency revisited

["Rob Morris" <rmorris@cdlsystems.com>]

Can't believe I said two I frames.

P's are predicted from I, then B's are calculated from P and I (or 2 P's).
Consequently, we must have the 4rth frame before we can decode the 2nd in
the sequence you specified. Hence the latency of at least 3 frames.
----- Original Message -----
From: "Ronan KERYELL" <Ronan.Keryell@enst-bretagne.fr>
To: "Bram Stolk" <bram@sara.nl>
Cc: <mpeg2@linuxtv.org>
Sent: Thursday, August 08, 2002 11:46 AM
Subject: [mpeg2] Re: kfir latency revisited


>>>>> On Thu, 8 Aug 2002 14:01:51 +0200, Bram Stolk <bram@sara.nl> said:

    Bram> Hello, I am trying to determine the latency of the BMK kfir
    Bram> card.  Rob Morris mentioned previously on this list, that the
    Bram> latency is at least as long as the time between two I frames.

    Bram> With GOPChop, I see that by default, the card generates GOPs of
    Bram> 15 frames, like this:

    Bram> IBBPBBPBBPBBPBB

    Bram> So I expect a 15 frame delay.

    Bram> However, experiments suggest a shorter delay.

I hope that the P are derived from I and previous P frames so that in this
case the theoretical delay should be down to 3.

Of course, what is the real workflow in the card?...
--
    Ronan KERYELL              |\/
    Labo Informatique Télécom  |/)  Tel:    (+33|0) 2.29.00.14.15
    ENST Bretagne, BP832       K    Fax:    (+33|0) 2.29.00.12.82
    29285 PLOUZANE CEDEX       |\   E-mail: Ronan.Keryell@enst-bretagne.fr
    FRANCE                     | \
http://www-info.enst-bretagne.fr/~keryell