[vorbis] Frequency response issues with native ogg vrs re-encoded ogg.

[k_myers at irito.com]

Thanks for your comments back...

Yes, I use oddcast to do the down step to 96kbps on the icecast2 stream and
what is strange it that it sound better. (To me at least) This is with
respect to bass, and over all fidelity, but after listening to a few
different sets, it does it with state of the union more then any other set
that I have encoded from the original wave files to ogg. I did the encoding
to ogg with oggdrop (not exactly sure if that had the RC2 encoder in it). On
the mp3 side I used the Radium "enhanced" Fhg Professional codec and WaveLab
3.0 at VBR 128kbps and VBR 256kbps. When I stream mp3 I am using oddcast dsp
beta 25 and of course that uses LAME with the settings using VBR "ABR" mode
with no hi or low pass cut off.

My problems seem to be the opposite of what people understand with mp3 vs
Vorbis. I am  getting the bass response I want to hear out of mp3 and I am
not talking "just being louder". It seems that vorbis does not have the same
kind of "punch through" at frequencies ranging from 70hz (and probably
lower) to 240hz. I am going to run some tests tomorrow when I get into the
studio using the original wave file for State of the Union (mastered with
the effects) and encode to Vorbis 350kbps and MP3 (at 360kbps if I can get
it that high.. not sure though as I am using Vegas 2.0 to do this) and then
take the files back into wave format and put them through a set of freqency
analyzers running at the same time in sync and see what I get. I may also
just out of curiosity try AC3. It would be interesting to hear the
differences in each side by side.

I know what you mean with Sony head phones in general however the MDR-7506
are not what I call "DJ headphones" since they are more often found wrapped
around the necks or the heads of directors on set (which is why I have a
pair as I do some ADR from time to time) and are the defacto standard in
just about any dialog mixing studio I have been in. The MDJ-700's are
definitely over the top with bass, and I rarely use them for mixing final
masters. They are just too over powering at times.

Once I get a few tests completed, I will post the resulting files on my ftp
site or have them for DL somewhere. I am also going to encode some test
pattern type files taken from the Berhinger test disc that ships with their
mixer consoles (we recently picked up one of their small Euroracks which
sounds fantastic for under 300 bucks Canadian) into both mp3 and Ogg. They
have some great tests for setting up gear and I think it would be good for
everyone to have a locked set of these files kicking around to use in these
situations instead of dealing with variables all the time. Variables are
good however, because no matter what... that is the way the world works.. :)
thanks again for your help.

Lithium

----- Original Message -----
From: "Gian-Carlo Pascutto" <gcp at sjeng.org>
To: <vorbis at xiph.org>
Sent: Monday, September 10, 2001 2:52 AM
Subject: Re: [vorbis] I hate myself for asking this, but...

> ----- Original Message -----
> From: <k_myers at irito.com>
> To: <vorbis at xiph.org>
> Sent: Monday, September 10, 2001 2:15 AM
> Subject: Re: [vorbis] I hate myself for asking this, but...
>
>
> > http://www.djlithium.com:8064/djlithium.ogg (vorbis stream at 96kbps
using
> > 350kbps oggdrop encoded files run through Oddcast Beta 25)
>
> If the source files are 350kbps how do you get the 96kbps stream?
>
> There are no peelers yet, so you must be reencoding. Vorbis currently
> does terrible with reencoded stuff, and problems due to reencoding
> aren't real quality problems anyway far as I am concerned. One should
> never ever reencode, period. Thats why offering peeling is such a big
> 'plus' for Vorbis. (Well, once we actually have it at least ;)
>
> > I was listening to the native Ogg 350kbps version of State of
> > the Union on-demand streaming from KasterBlaster yesterday with another
> copy
> > in mp3 running on a seperate box coming from my other server at 256kps
> > on-demand
>
> Ah, so I gather those aren't reencoded streams. Are they rc2? Anything
> before rc2 _did_ have bass problems.
>
> I notice the first part of that stream has a bass sample in the
> background that I know causes a lot of trouble with Vorbis, but that
> wouldn't explain why you percieve an overall lack of bass. (It would
> certainly explain the preecho though)
>
> > (check my site for the on-demand links in mp3 format at
> > www.djlithium.com under the streaming section), and I can definately
hear
> > the difference.
>
> It could also be of course that MP3 is terribly messing up on the
> bass causing it to come out louder.
>
> (Okay okay, I know it's farfetched, but without the originals, there's
> really no way I can tell anything here)
>
> > I have tried both sets of head phones (Sony MDR-7506 and
> > Sony MDJ-700's)
>
> Sony's tend to overemphasize the bass (I guess that is why
> they are popular with DJ's ?) so that would explain why you
> percieve the problems so good exactly in the bass section.
>
> That doesn't explain why you percieve the MP3's to be better
> though...even if the artifacts are exagerrated Vorbis shouldn't
> have more than MP3 in the first place ;)
>
> > Does bit rate affect this in anyway? I am listening to the 96kbps ogg
> stream
> > now, an its actually got more bass in it then the 350kbps files, and it
> also
> > seem there is more "pre-echo" in the 350kbps versions.
>
> That is certainly very weird. Perhaps the 'more bass' _is_
> actually an artifact after all?
>
> > I use some spacial expansion effects on the mastering of the waves
before
> > I encode them, but these wave files were used to generate both the MP3
and
> > Ogg versions with effects on for each, so I am guessing the outcome
would
> > be completely independent of that,
>
> That's correct. If you did this _before_ encoding it should be ok.
>
> I could imagine it's possible that some very weird stereo effects
> play foul with the advanced stereo modes, but the 350kbps mode
> doesn't use those.
>
> > but I am wondering if Ogg (or Vorbis really)
> > has problems with dealing with signals that are very wide in stereo
image.
> Does
> > this make sense??
>
> No. Certainly not for the 350kbps stream, which would be losslessy
coupled,
> and very likely not for the 96kbps stream with the advanced stereo modes
> even.
>
> If you could snap a part where you percieve the problems to be
> most obvious out and put up the original of that somewhere, that
> would be very helpfull. (I'm also interested in the MP3 encoder
> and settings you used)
>
> --
> GCP
>
>
> --- >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.