[vorbis] Low bitrate encoding [non-mockery]

[Erik J. Kruus]

Dan, it would have been nice if you had looked up
the sox home page yourself:

       http://home.sprynet.com/~cbagwell/sox.html

and answered all your own questions.  You did not answer
my one question to you.  So I'll ask you to answer your own
questions by telling you where you can find the answers.

Recently [Sept. 4, 2000] sox is also accessible via

   http://sourceforge.net/projects/sox/

[Note: Since 1996, sox has a new maintainer.  Be sure, that
you did not go to the old, out-of-date 1995 page of the
original sox author, whose last official sox release was in
1995, when installing your "official" sox version.  It actually
popped up before the new sox home page when I did my
search today!]

I'm not a sox advocate.  I'm telling you how I avoided problems
that you had with sox.  That's it.  You can use any resampling
you wish, including the possibly better code from the second
URL I gave.  You could even take the average of every pair
of consecutive samples and live with monty's mockery, or
use your own proprietary resampling code. I don't care.
Don't worry. Be happy.

Dan Hollis wrote:

> On Thu, 11 Jan 2001, Erik J. Kruus wrote:
> > Dan Hollis wrote:
> > > On Wed, 10 Jan 2001, Erik J. Kruus wrote:
> > > > Now I ask you to read about how sox does it better at two URLS.
> > > > Sox has three ways, compared with graphs and interpretation
> > > > at
> > > >           http://leute.server.de/wilde/resample.html
> > > The polyphase resampler is best (to my ears) when the source and destination
> > > frequencies are not perfect multiples of each other, but the sox code is
> > > really buggy (crashes 90% of the time).
> > Hmmm. Do you have a recent sox? Personally I've only bothered with the
> > best algorithm, which around sox 12.17 was resynced with the excellent
> > source code obtained from the second URL I mentioned.
>
> Does the production release sox have polyphase resampler included?
> Last one I checked was simple interpolation filter, polyphase was still
> external unsupported patches.
>

Well, I when I said "Sox has three ways", I meant "Sox has three ways".
Check it out for yourself. Go to the web page.  Read under "Technical
Information".  If you still have doubts check out the source code or CVS
trunk for sox-12.17.1 and see if it's up to the task for which you need it.

> I don't care about specific distribution hacked sox, I'm talking about the
> official source code releases.

I install sox as debian or mandrake version, and I ALSO install from either
the "official source code releases" [or from sox CVS trunk if there's a recent
addition I need] and call that binary soxx. I can use either one at will.
I tend to use the most recent release myself.  Today, one may download
this release [12.17.1] as source.tar.gz, Redhat rpm for {i386,ALPHA,Sparc,source}
Win32 binary and OS/2 binary.  Only Debian is an older version.

However, if you feel that some "distribution hacked sox" is at the
root of your 90% crash experience, then perhaps you should test
my 0% crash experience by following my advice to get a recent version.

Of course, nobody can know what your exact problem is if you give
no information.  Consider, Dan, that if you had told people -which-
version and -which- "distribution hacked sox" is the one to avoid,
this would be a generally helpful contribution.

But please don't stop at just looking at the sox code.  Possibly the
best resampling code can be obtained from the second URL I gave.
If this code also fails your criteria for excellence, then I'm sure
many would be interested to know precisely why it fails to
satisfy you.
 [ http://www-ccrma.stanford.edu/~jos/resample/resample.html
    "Digital Audio Resampling Home Page" for those who missed it
   comes with efficient 'C' implementation of the algorithm.]

Then, if after having done some research yourself, you still feel
that I'm giving misleading advice to people who want to resample
with a good algorithm you may badmouth and pooh-pooh to your
heart's content and I'll publicly apologize for having misled folks.

But don't just say "it doesn't work, didn't the last time I looked"
without supplying information relevant to solving your problem
and without listening to folks advising you to please look again
and telling you where to look. OK?  Deal?

But the general vorbis folks are definitely invited to check out
the sox home page URL and code themselves, and believe
neither Dan nor me. Open Source is a great and powerful thing.

If you still have questions about resampling, take a look at some
of the dsp mailing lists.  They used to discuss resampling algorithms
fairly regularly, and by now may even have a good section about it
in their FAQ.

Let's kill this resampling thread.  It's known, standard technology,
and freely available to all, and quite a bit better than just averaging
every pair of samples.

Erik.

>
> -Dan
>
> --- >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

--- >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.