> As I can't compile for dos/windows I was relying on the 1.1.6 binaries
from
> the download area.
If you don't feel like recompiling, just discard the last byte and don't
rely on the terminator.
> Just to let you know, I've been working on a port for the eCOG
> microcontroller (www.cyantechnology.com).
Cool.
> I managed to get a decoder-only build by adding a large number of ifdefs to
> the code - it's certainly not a neat solution, but it's got me
started.
> Things look promising for our next variant of the chip. I'll let you
know
> how we get on in a couple of months.
Please keep us posted.
Jean-Marc
> Andy
>
>
> >From: Jean-Marc Valin <Jean-Marc.Valin@USherbrooke.ca>
> >To: Andrew White <andrewmarkwhite@hotmail.com>
> >CC: speex-dev@xiph.org
> >Subject: Re: [Speex-dev] 8.4kbps data rate in 8kbps nb mode
> >Date: Fri, 18 Nov 2005 20:30:45 +1100
> >
> >Indeed there was a bug in the speex_bits_insert_terminator() function.
> >It's now fixed in svn. The only people this bug affected are the
ones
> >relying on detection of the end of packed Speex frames and some wasted
> >bytes when using 8 kbps mode.
> >
> > Jean-Marc
> >
> >
> >Le jeudi 17 novembre 2005 ? 23:18 +0000, Andrew White a ?crit :
> > > Hi Jean-Marc,
> > >
> > > Thanks for your response.
> > >
> > > I wrote a script to remove the Speex bits from an Ogg file - As
I'm
> >using
> > > CBR
> > > mode I already know the number of bits in each frame and can do
without
> >the
> > > Ogg
> > > overhead.
> > >
> > > In the Ogg Format you have a segment table that shows the length
of each
> > > segment. When --nframes is set to the default of 1, one Speex
frame
> >equals
> > > one
> > > segment/packet.
> > >
> > > If you encode at 8kbps each segment is 0x15(21) bytes long, if
you
> >encode at
> > > 15kbps each segment is 0x26(38) bytes long. You can easily see
this if
> >you
> > > view
> > > the Speex file with a Hex text editor as the segment table is the
> >segment
> > > size
> > > repeated several times.
> > >
> > > In 15kbps mode you get 300 Speex bits in one frame. This fits in
the 38
> >byte
> > > segment with 4 bits to spare. I expected 160 Speex bits in 8kbps
mode to
> >fit
> > > into a segment 20 bytes long, but speexenc gives a file with 21
byte
> > > segments.
> > >
> > > If I drop the 21st byte and decode the stream I don't seem to
get any
> >loss
> > > of
> > > quality, so I was wondering if it was extraneous data, or if the
Ogg
> >segment
> > > needs to be slightly larger than the Speex data for another
reason. I
> >looked
> > > through the code, but couldn't see why the segment size would
need to be
> > > larger
> > > than the Speex data itself.
> > >
> > > Andy
> > >
> > >
> > >
> > >
> > > >From: Jean-Marc Valin <Jean-Marc.Valin@USherbrooke.ca>
> > > >To: Andy White <andrewmarkwhite@hotmail.com>
> > > >CC: speex-dev@xiph.org
> > > >Subject: Re: [Speex-dev] 8.4kbps data rate in 8kbps nb mode
> > > >Date: Fri, 18 Nov 2005 09:00:08 +1100
> > > >
> > > >I'm not aware of this. How did you measure this value?
Note that in
> > > >practice, there is a small overhead to Ogg packets.
> > > >
> > > > Jean-Marc
> > > >
> > > >Le mercredi 16 novembre 2005 ? 12:18 +0000, Andy White a
?crit :
> > > > > I know that most of the narrow band submodes don't
have an exact
> > > >multiple of
> > > > > eight bits in each frame
> > > > >
> > > > > 79 bits for 3.95kbps
> > > > > 119 bits for 5.95kbps ...etc
> > > > >
> > > > > This means that you have a few unused bits when
transporting Speex
> >in
> > > >the Ogg
> > > > > format (you can achieve a lower number of unused bits
if you change
> >the
> > > >number
> > > > > of frames per Ogg packet to suit the mode).
> > > > >
> > > > > However, in 8kbps mode, the number of bits per frame is
160. I would
> > > >have
> > > > > expected this to fit into an Ogg segment 20 bytes long.
However,
> >when I
> > > >use
> > > > > (speexenc -n --quality 3) to encode at 8kbps the Ogg
segments are
> >each
> > > >21 bytes
> > > > > long.
> > > > >
> > > > > Is this an error, or is there extra information put in
the 21st
> >byte?
> > > > >
> > > > > I'll have a hunt through the code, but thought
someone might have
> >come
> > > >across
> > > > > this before.
> > > > >
> > > > > Regards,
> > > > >
> > > > > Andy
> > > > >
> > > > > _______________________________________________
> > > > > Speex-dev mailing list
> > > > > Speex-dev@xiph.org
> > > > > http://lists.xiph.org/mailman/listinfo/speex-dev
> > > > >
> > >
> > >
> > >
>
>
>