Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: "Up-front" high frequencies (Read 5519 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

"Up-front" high frequencies

Hello everyone.  I've been lurking on this board since it first began and decided to finally register and post something.  I still lurk on the r3mix board... but have no desire to post there

Anyway, on to the subject.  I've been playing with the latest BQOGGDROP at quality settings 5 and 6.  In general, the files sound very, very good for the bitrate.  However, I have noticed that high frequency sounds seem to be more "up-front" in the ogg-encoded files than they are in the original wavs. 
One example I found was when I encoded "Ridiculous Thoughts" by the Cranberries.  Around the one minute mark, a tambourine starts and seems very "up-front" in the ogg file.

Am I just hearing things, or is anyone else experiencing this?

"Up-front" high frequencies

Reply #1 –
When you say up front what exactly do you mena by that.

"Up-front" high frequencies

Reply #2 –
They seem to be exagerated somewhat.

"Up-front" high frequencies

Reply #3 –
I agree that sometimes ogg encoded files can sound a little 'bright' compared to the original. Personally, I prefer that to the way some other codecs make things sound a little 'dead' by comparison. I think it's a matter of personal choice. However you mess with the original is surely bound to 'colour' the result in some way or another.

john33

"Up-front" high frequencies

Reply #4 –
Do a double blind test with the parts you think sound bright and you will know if it's real or that you're just imagining things. You can get a program - ABX Comparator - to do these tests on www.pcabx.com

"Up-front" high frequencies

Reply #5 –
I was just checking through the ogg directory of the forum and found this thread.  Glad to know that someone else has picked up on the higher-frequency sounds being exaggerated ("up-front" or "bright") with ogg.  On certain songs, I can easily distinguish between ogg files and the original wav files, and can do so consistently on double-blind tests.  The problem is worst at low bitrates, and RC3 has improved on RC2 in this regard, but I can still differentiate between wav and ogg up until -q 6.  I'm not picking up on any annoying artifacts here, but the increased brightness is still a negative.  Seemingly it would be possible to to not "colour" the sound much, erring neither on the side of deadness nor brightness.  The files on which I have found this increase in brightness to be most noticeable are the opening measures of Nirvana - In Bloom (1991), and the opening instrumental measures of Stryper - In God We Trust (1987).  Also useful here is the opening measures of 77's - Look (1991), where the snare drum becomes noticeably brighter with ogg.
God kills a kitten every time you encode with CBR 320

"Up-front" high frequencies

Reply #6 –
Well, I may be committing a sin by revealing comments on the 128 listening tests before it's closed, but Monty had this to say about one of the samples encoded with the pre-RC3 ogg:

Quote
Uniformly bright, background noise mostly stays put... most of the time


He ABX'd it 16/16.  Sounds like it's something which needs work on.

ff123

"Up-front" high frequencies

Reply #7 –
Quote
Originally posted by ff123

He ABX'd it 16/16.  Sounds like it's something which needs work on.

ff123


I correected that problem about a month ago.  The masking bug I mentioned before that Dibrom found a few weeks before the official rc3 is the same problem, and I fixed it then; when I randomize an official rc3 encoding into the ABX mix, it beats the other codecs and has no wandering background noise.

The version of Vorbis being used in the second round of tests you're collecting is between rc2 and rc3 and is missing this fix and several others.

BTW, when I said 'uniformly bright' I meant 'didn't sound like it was lowpassing to save its skin on every synth hit like the others'.  I said the same about MPC :-)

Monty (who was able to ABX *all* the wayitis samples at 16/16 IIRC)

"Up-front" high frequencies

Reply #8 –
It's not annoyingly bright on most of my music.  Ogg noticeably exagerates the high frequencies on only a few tracks.

On a different note, I've been playing around with different quality settings.  I was totally shocked when I listened to a track encoded at q0.  I expected it to be very bad, but I found that the artifacts weren't nearly as terrible as I expected.  If you are reading this and haven't tested q0, give it a try and see if it startles you like it did me!

"Up-front" high frequencies

Reply #9 –
Quote
Originally posted by Bones
It's not annoyingly bright on most of my music.  Ogg noticeably exagerates the high frequencies on only a few tracks.

On a different note, I've been playing around with different quality settings.  I was totally shocked when I listened to a track encoded at q0.  I expected it to be very bad, but I found that the artifacts weren't nearly as terrible as I expected.  If you are reading this and haven't tested q0, give it a try and see if it startles you like it did me!


To be honest, I don't like the quality at -q0 right now.  I've been working on it and decided to trade 'fidelity for usability' at the lowest quality level in rc4...  I'm going with a somewhat artificial brightness quality rather than having the high end descend into murk and swirl like it often does now (when I say 'artificial', a little like mp3pro, but not as pronounced.  What it does do is make the sample noticably hissier).

It's a trade off and definately a compromise, but I consider it better than what's there now.  For a demo of the tuning-in-progress, see:

http://www.mit.edu/afs/athena/user/x/i/xip...blic/ogg-64.ogg

a decoded comparable WMA sample (in .WAV) at 64 is at:

http://www.mit.edu/afs/athena/user/x/i/xip...blic/wma-64.wav


I'll upload the original comp.wav these were generated from into the same dir now.

"Up-front" high frequencies

Reply #10 –
I've decided to close the test on Saturday (see my other post).  So listener comments will be available for people to read then.

ff123

"Up-front" high frequencies

Reply #11 –
Monty: I dont like to ask this but could you give me an idea what i would need to change to enable lossy coupling on all q settings.
I would like to experiment with this setting, as would a couple of others on this forum.

Thanks alot
Alex

"Up-front" high frequencies

Reply #12 –
Quote
It's a trade off and definately a compromise, but I consider it better than what's there now. For a demo of the tuning-in-progress, see: 

http://www.mit.edu/afs/athena/user/x/i/xip...blic/ogg-64.ogg 

a decoded comparable WMA sample (in .WAV) at 64 is at: 

http://www.mit.edu/afs/athena/user/x/i/xip...blic/wma-64.wav


Ogg is better than wma on most, if not all of those clips.  The most annoying thing to me about the ogg is the exaggerated sibilance, especially noticeable on 41_30sec.wav

ff123

"Up-front" high frequencies

Reply #13 –
Quote
Originally posted by Alex
Monty: I dont like to ask this but could you give me an idea what i would need to change to enable lossy coupling on all q settings.
I would like to experiment with this setting, as would a couple of others on this forum.

Thanks alot
Alex


It's already active on my CVS branch (but that branch is not currently usable).  BTW, changing the stereo coupling assumes that all the codebooks needed to represent the different mode are populated.  They are actaully in rc3, but only up to -q6.

"Up-front" high frequencies

Reply #14 –
This demo ogg64 beats mp3pro at 64 as well.  I downloaded the wav and encoded with Nero.

ff123

"Up-front" high frequencies

Reply #15 –
Quote
Originally posted by xiphmont


It's already active on my CVS branch (but that branch is not currently usable).  BTW, changing the stereo coupling assumes that all the codebooks needed to represent the different mode are populated.  They are actaully in rc3, but only up to -q6.


Oh, sorry... for playing around purposes, you can change the _psy_stereo_point_dB_44 and _psy_stereo_point_kHz_44
numbers in 'lib/modes/psych_44.h'

the 'dB' is a mode setting from 0 (lossless) to 3 (18dB) integer, and the point_kHz is where it changes to point stereo in kHz for short/long blocks.

"Up-front" high frequencies

Reply #16 –
Thanks Monty, Ive been able to compile a version of oggenc with lossy coupling upto 6.99x.
I will do some testing with this, both listening and EAQUAL, though how usefull the results of that are is questionable.

Im very much looking forward too RC4 and 1.0, I wish there was more I could do too help somehow.

Thanks very much
Alex

"Up-front" high frequencies

Reply #17 –
His versions are already at my page.

Regards;

Roberto.

"Up-front" high frequencies

Reply #18 –
Quote
Originally posted by rjamorim
His versions are already at my page.


I cannot find it. Also I noticed that "Custom compiles of OggEnc" not available now. Where's the new compiled version?

Anyway, thanks for your kindness, rjamorim.

"Up-front" high frequencies

Reply #19 –
You should find that the link for Custom compiles of OggEnc at Roberto`s site is the link to the new files.

The Quality Optomized encoder is no longer there, it just wasnt of any use, there were no case in which it produced files audibly better and there were some results showing it produced worse sounding files.

Theres now a standard version of Oggenc compile with Intel c++ 5 compiler, that produces files identical to oggenc from www.vorbis.com.

And there is also an experimental version of oggenc with lossy channel coupling enabled upto -q 6.99.

Thanks to Roberto for hosting the files.