Originally posted by timcupery 1.  The tinny ring in the background may be related to the exaggerated brightness that a lot of people have noticed in ogg-encoded files.

Originally posted by JohnV
I wouldn't make this kind of conclusions about purely tonal sine wave regarding music encoding in general...

Originally posted by Ivan Dimkovic


Well, it is a bug - and it should be fixed, because other codecs do not produce this bug.

Originally posted by timcupery

Ogg encodes the tone itself seemingly fine, but adds this tinny background ringing buzz which is incredibly annoying.  Not a good thing.  This potentially points to a couple problems:

1.  The tinny ring in the background may be related to the exaggerated brightness that a lot of people have noticed in ogg-encoded files.  This isn't very noticeable or annoying within songs with a lot of different sounds going on, but a clip where it can be isolated - like the one in question here - points out a serious problem with the encoder algorithm.

2.  The algorithm which decides the output bitrate when using the quality settings may need to be "smartened" so that it puts out a higher average bitrate for clips which it can't handle, such as this one.  Even if this is fixed, however, vorbis should be able to handle such a sound clip at lower bitrates.

Originally posted by Ivan Dimkovic
hmm.. I am sorry if this has been discussed before, but since other codecs encode this file properly I thought it might be a bug of some sort in Oggenc..

anyway I >think< it should be encoded properly.

Hi all,

I have implemented ReplayGain support for Vorbis.

If you are not familiar with it, it is basically
a method of making sure all your files have equal loudness,
remove the need for normalization and prevent clipping
during playback. The process is totally lossless,
and supporting it requires minimal work.

More info about the exact workings can
be found on www.replaygain.org (recommended reading) 

The support consists of three levels:

a) An agreed tag format for storing the ReplayGain
data. I hereby propose:

RG_RADIO=+0.0 dB
RG_AUDIOPHILE=+0.0 dB
RG_PEAK=1.000

The exact meaning of these values is described
on www.replaygain.org. This format should be
self-explanatory otherwhise. Note that the Vorbis
version has a human-readable form, in accordance
with the fact the they are in the 'comment' field
of a Vorbis file.

If possible, and if this format is agreeable
for everyone, I would ask that they be included
in the Vorbis comment/tag format documentation.


b) An analysis program, that reads .ogg files,
calculates the correct gain values, and stores
them back into the oggs as tags.

This program is already written and available at
http://sjeng.org/ftp/vorbis/ReplayGain-0.3.tar.gz.
It should work on both Unix and Windows, although I
have not tried it on the latter yet.

Usage is as simple as:

replaygain <file1> <file2> <...>

It also supports

replaygain -a <track1> <track2> <...>

to perform 'Audiophile Gain' analysis. This stores
additional ReplayGain values for entire albums.


c) The decoder software

ReplayGain must be supported by the decoder/player to
have any effect. If the decoder/player does not support
ReplayGain, the files will still play as usual.
In fact, nothing will change.

Basic support can be done in a minimal amount of
code:

1) Read out the RG_RADIO tag
2) Calculate scale = 10^(RG_RADIO/20)
3) Multiply all PCM samples with 'scale'

That's it!

If the decoder author wishes to do so, he can also
add support for more advanced features (in order
of complexity):

4) Offer a selection between 'Radio' and 'Audiophile'
ReplayGain and use RG_RADIO or RG_AUDIOPHILE depending
on the selection. (trivial)

5) Use the RG_PEAK value to prevent clipping and thus
improve audio quality. (also very easy, see example code)

6) Implement a preamp and a hard limiter to allow the
user to increase the loudness without causing the audio
to clip. (more tricky to do fast)

Example code, including all of the 'advanced' features is provided
in the form of an enhanced XMMS decoder with ReplayGain support
at:

http://sjeng.org/ftp/vorbis/ReplayGain-XMMS-0.2.tar.gz

Originally posted by xiphmont
BTW, Peter has implemented replaygain in the most recent WinAmp plugin

Originally posted by timcupery
I normalized Ivan's 1k.wav file to 98% using EAC, and encoded the result at -q 4.9.  Could still hear the tinny background ring, clipping evidence.
Normalized the 1k file to 95%, same encode, this time it sounded perfect.  Hope that this is helpful in some way or another.  And may [evil] monty please forgive my wild speculation (i.e., hypothesizing) above.

Running in Radio/Single Track mode

Analyzing...

   Gain   |  Peak  | Scale | New Peak | Track

---------------------------------------------

-14.16 dB |  33467 |  0.20 |     6556 | 1k.ogg

Done.