[vorbis] Vorbis codec delay

Michael Smith msmith at labyrinth.net.au
Thu Aug 8 23:47:13 PDT 2002



At 03:30 AM 8/8/02 -0700, you wrote:
>I'm evaluating audio codecs for use in a video
>conferencing application, and was interested in vorbis
>due to its high quality and patent-free algorithm.
>Encoding/decoding delay is one parameter vital for
>conferencing usage which I have not been able to find
>in any FAQ or documentation. Before I start to dig
>into the code to "reverse engineer" the delay I would
>like to ask if anyone on the list has sucessfully
>determined the delay of the codec. (Maybe I have just
>missed some document?)

I'm not sure if anyone has done an exact determination of the
delay, but here's a rough outline:

Ogg (the framing layer on top of vorbis) has a very high delay,
because the encoder uses 4 kB 'pages' by default. At low bitrates,
this can be a couple of seconds. However, using ogg is optional,
though it's substantially more work to do it in other ways, and
if you DO stick with ogg, you can (at a cost of slightly more
overhead) reduce the page size, or even ensure only one vorbis
packet is used per ogg page. So this isn't a major problem.

Decoder: the decoder obviously has some delay, but this is
dependent on the block sizes used for a particular stream,
so it's hard to give a single answer. This is not particularly
large, however.

Encoder: the encoder has a high delay of several long block
lengths at a minimum (and much more if you're using managed modes,
so you would use VBR for lower delay). Of course, this is not
a fundamental part of the format, just a design decision for the
current encoder. Other encoders (if they existed) could do differently.

As a rough estimate, I'd expect the vorbis delay (i.e. encode+decode
delay, but NOT including ogg, since you can reduce that to 0 additional
delay if you want) to be around 4k samples with the current software.
I may be way off there, though, I wouldn't take any numbers I've given
as being accurate.

Michael

<p><p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.



More information about the Vorbis mailing list