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: problematic track: crystal method - cherry twist (Read 7441 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

problematic track: crystal method - cherry twist

hi,

i'm not very good at discribing artifacts by language/knowledge issues, but i did some listening test with track "cherry twist" by the crystal method (i will not upload the track, as i don't have the cd, only mpc insane and don't have any webspace to hand) with ogg 1.0 @ -q 4 and compared to lame --alt-preset 128 (try listening to ~0:55 - 1:00). while ogg sounds *very* distored, lame does fine.

regards; ilikedirt

problematic track: crystal method - cherry twist

Reply #1 –
Quote
Originally posted by ilikedirtthe2nd
i'm not very good at discribing artifacts by language/knowledge issues, but i did some listening test with track "cherry twist" by the crystal method (i will not upload the track, as i don't have the cd, only mpc insane and don't have any webspace to hand) with ogg 1.0 @ -q 4 and compared to lame --alt-preset 128 (try listening to ~0:55 - 1:00). while ogg sounds *very* distored, lame does fine.


Is this the song "Cherry Twist" off of the Vegas CD?  I want to avoid doing that whole Lovelife thing over agian.    I have the CD at home.  I'll try to flac 0:50-1:05 tonight and post a link back to this thread to the file.

problematic track: crystal method - cherry twist

Reply #2 –
Quote
Originally posted by Ardax
Is this the song "Cherry Twist" off of the Vegas CD? 


yes, it is

problematic track: crystal method - cherry twist

Reply #3 –
ilikethedirtthe2nd:
Ok, so you tried to re-encode a file that has already been encoded once (MPC --insane, wasn't it?) and discovered artifacts? Not the best way of testing codecs in my opinion...

EDIT: The poor results could be due to a transcoding done between a file encoded with a subband codec (MPC) to a file encoded with a transform codec (Vorbis), or something like that...

Maybe you could find some info on this at this thread

problematic track: crystal method - cherry twist

Reply #4 –
Quote
Originally posted by kjempen

EDIT: The poor results could be due to a transcoding done between a file encoded with a subband codec (MPC) to a file encoded with a transform codec (Vorbis), or something like that...


Actually, this is the most favorable way of doing the evil transcoding task. But i agree, a test should only be performed with the original file, not a previously encoded one.

problematic track: crystal method - cherry twist

Reply #5 –
Okay, here we go:

http://ardax.dyndns.org/tmp/CherryTwist/

Has the flac and a few encodes.  Lame 3.92 --alt-preset 128, Ogg rc3 -q4.0, and Ogg 1.0 -q{4.0, 4.50, 4.99, 5.0}.

It sounds like the quality scale has slid a bit between rc3 and 1.0.  The ogg 1.0 -q4.0 encode is noticeably worse in quality than the rc3 -q 4.0 encode, but the file is much smaller.  Encoding with 1.0 -q4.50 yields a filesize within 1k of the rc3 -q4.0 encode, and they sound very similar.  Should still be fairly easy to abx against the original though.

Edit: looks -> sounds.

problematic track: crystal method - cherry twist

Reply #6 –
Does everybody think that the ogg 1.0 release got out too hasty to have a thorough listening test, thus make no room to tune more for these problematic sound clips?

Then should I trust the quality achieved by ogg 1.0?

problematic track: crystal method - cherry twist

Reply #7 –
To be honest, I don't think it was too early to have a proper listening test, for a few reasons:
1. Do you test the final or the beta..? It will never be perfect so do you beta it from now to the end? (not that I would have really minded either way but still...)
2. Just cos it is 1.0 doesn't rule out 1.1, 1.2, 1.2.1, 1.2.1.1.0.94 etc. Changes can still happen.
3. From what I have heard of monty's ears, I'd rather have him say 'yup, cool, ok' than a listening test involving 1000 clones of me.

Problem samples will keep on coming; they always seem to. To wait until there are no more samples with issues is to wait until people stop making new music (imho).

Go 1.0, go rc4, go whatever they want to call it. Go vorbis!

gnoshi, with his as usual unbiased and thoroughly objective viewpoint =)
happiness comes in brown paper bags.

problematic track: crystal method - cherry twist

Reply #8 –
Thanks for your quick reply, gnoshi. I really agree with your opinion.

So I cannot wait for the post-1.0 with wavelet implementation out for encoding my favorite classical music with this great music format.

problematic track: crystal method - cherry twist

Reply #9 –
lol! Sadly, I think you'll have to wait a while for that.
I believe that there a lot of other things on the agenda before wavelet implementation (though not being a developer, I could be way off track).

gnoshi
happiness comes in brown paper bags.

problematic track: crystal method - cherry twist

Reply #10 –
Quote
Originally posted by Enig123
Does everybody think that the ogg 1.0 release got out too hasty to have a thorough listening test, thus make no room to tune more for these problematic sound clips?

Then should I trust the quality achieved by ogg 1.0?


To me, it seems like they adjusted the quality scale a bit, not that 1.0 is worse than rc3.  On this clip, rc3 -q4.0 sounds much the same as (and is about the same filesize as) 1.0 at -q 4.5.

The quality should be fine, you'll just have to find a quality setting that satisfies you.

That said, I added an rc3 -q3.0 encode.  It's almost the same size (< 1k) as the 1.0 -q4.0 encode.  I think that the rc3 encode sounds better than the 1.0 encode.

Of course, this "1.0" is really the ultimate, final, 1.0.

[span style='font-size:9']Edit: What the hell was I thinking when I wrote that last line?  Please ignore it.  [/span]

problematic track: crystal method - cherry twist

Reply #11 –
Quote
Originally posted by Enig123
Thanks for your quick reply, gnoshi. I really agree with your opinion.

So I cannot wait for the post-1.0 with wavelet implementation out for encoding my favorite classical music with this great music format. 


Well, Vorbis I is never going to see wavelets -- Monty's said that much himself.  I'm sure that wavelets would require decoder changes to be really worthwhile (not necessarily to work, but to be useful).  Vorbis II (or 2.0, whatever it ends up getting called), on the other hand, should be rockin' with wavelets though.

problematic track: crystal method - cherry twist

Reply #12 –
Quote
Originally posted by Ardax


Well, Vorbis I is never going to see wavelets -- Monty's said that much himself.  I'm sure that wavelets would require decoder changes to be really worthwhile (not necessarily to work, but to be useful).  Vorbis II (or 2.0, whatever it ends up getting called), on the other hand, should be rockin' with wavelets though.

It's a repeated refrain about how wavelets are going to transform Ogg Vorbis into some unbeatable encoder, but I worry about the hype and the unreasonable expectations. If wavelets are the "silver bullet" why do people simply talk about the technology rather than develop it?

problematic track: crystal method - cherry twist

Reply #13 –
Quote
Originally posted by mithrandir
If wavelets are the "silver bullet" why do people simply talk about the technology rather than develop it?


And - more important - why hasn't it been implemented in any other format yet (AFAIK), since it's such a great technology?

problematic track: crystal method - cherry twist

Reply #14 –
What's worse, many of the people chanting "wavelets, wavelets!" often have no idea what they are at all. I certainly don't; I tried searching for articles from my university library, and I couldn't find a good introductory discussion of them. All I found was a whitepaper from some private company that says "when we implement wavelets into our audio codec, it will be the best!" and goes on to show lots of pretty pictures and colorful graphs.

From what I gather, wavelet theory is pretty new stuff, and it hasn't been investigated very thoroughly. It also hasn't been successfully integrated into any useful technology, AFAIK. I have little doubt that it will eventually be used to achieve some gains in psychoacoustic audio compression, since it looks like people who are "in the know" have thrown their weight behind wavelets as an important new addition to compression. Still, until I see some sort of demonstration of wavelets in action, either for audio or video, I'm not going to blindly believe it wil be the next "wave" of digital media compression.

problematic track: crystal method - cherry twist

Reply #15 –
I have an image compression program that uses wavelets. But as for audio programs none yet. So they are not unheard of. What worries me more is the fact that after reading this thread I wanted to get my copy Vegas out and listen to it. But all I could find was the case! I panicked and freaked for a while. Finaly though I came to the realization it has to be in one of these piles somewhere. Summer cleaning anyone? Shame to destroy the leaning tower of CDs.

problematic track: crystal method - cherry twist

Reply #16 –
That's why all my CDs reside in a box, in the corner, under my 4-track.
They only come out when the rippin' is needin' doin'

=)
gnoshi
happiness comes in brown paper bags.

problematic track: crystal method - cherry twist

Reply #17 –
Quote
Originally posted by Neo Neko
I have an image compression program that uses wavelets


Jpeg2000 uses wavelets & most new image compression codecs do, I heard (don't remember where) wavelet tech has proven most useful with picture compression than audio but if it's good for image it should at least be good of video.

problematic track: crystal method - cherry twist

Reply #18 –
I've seen a quick wavelet primer at:

http://www.public.iastate.edu/~rpolikar/WA...veletindex.html

According to it these are the main differences:

The continuous wavelet transform was developed as an alternative approach to the short time Fourier transform to overcome the resolution problem. The wavelet analysis is done in a similar way to the STFT analysis, in the sense that the signal is multiplied with a function, {it the wavelet}, similar to the window function in the STFT, and the transform is computed separately for different segments of the time-domain signal. However, there are two main differences between the STFT and the CWT:

1. The Fourier transforms of the windowed signals are not taken, and therefore single peak will be seen corresponding to a sinusoid, i.e., negative frequencies are not computed.

2. The width of the window is changed as the transform is computed for every single spectral component, which is probably the most significant characteristic of the wavelet transform.

Hope that helps.
Jazzist

problematic track: crystal method - cherry twist

Reply #20 –
Quote
Originally posted by Ardax
Okay, here we go:

http://ardax.dyndns.org/tmp/CherryTwist/

Has the flac and a few encodes.  Lame 3.92 --alt-preset 128, Ogg rc3 -q4.0, and Ogg 1.0 -q{4.0, 4.50, 4.99, 5.0}.

It sounds like the quality scale has slid a bit between rc3 and 1.0.  The ogg 1.0 -q4.0 encode is noticeably worse in quality than the rc3 -q 4.0 encode, but the file is much smaller.  Encoding with 1.0 -q4.50 yields a filesize within 1k of the rc3 -q4.0 encode, and they sound very similar.  Should still be fairly easy to abx against the original though.

Edit: looks -> sounds.



I've grabbed it.  It does indeed illustrate the point well.  I'll be using it to tune after we push the release out.

problematic track: crystal method - cherry twist

Reply #21 –
Quote
Originally posted by xiphmont



I've grabbed it.  It does indeed illustrate the point well.  I'll be using it to tune after we push the release out.


Actually, I did a cursory check as I've heard this specific effect myself in other samples, but this one illustrated it better than any tests I had previously.

The problem (or at least the change from RC3 that results in the perceived problem) is clearly the new dipole stereo.  This was a new feature added just before rc4 in an attempt at more 'correct' (at least in an objective sense) stereo.  The objective improvement has clearly caused a subjective degredation. 

As dipole stereo has clearly placed itself in the 'rushed feature' category, I'm reverting it back to the rc3 method (elliptical stereo) right now, as it's clear the old method produces a better result.

Monty

problematic track: crystal method - cherry twist

Reply #22 –
Quote
As dipole stereo has clearly placed itself in the 'rushed feature' category, I'm reverting it back to the rc3 method (elliptical stereo) right now, as it's clear the old method produces a better result


What low level library source would I be able to find this in? What advantage did you percieve from changing to dipole stereo? Are there any other methods in a simplimatic way that you could add that would give better results?
budding I.T professional

problematic track: crystal method - cherry twist

Reply #23 –
Quote
Originally posted by rjamorim


And - more important - why hasn't it been implemented in any other format yet (AFAIK), since it's such a great technology?


A) It's not a silver bullet.  It's one more useful tool that must be applied properly.  Anyone who claims wavelets are a silver bullet (And, BTW, Fractals are just a special case of wavelets; that was only recently discovered) is trying to sell you something.

B) Wavelets are literally brand-new math, 'invented' in the eighties.  The other techniques being used by modern 'cutting edge' compressions have been known for anywhere from fifty years to as long as a century or more. 

C)  Rehashing A, company press releases lie.  Did we really need to point that out explicitly? :-)  Belive it when you see... err... hear it.

Nothing goes into Vorbis just because it's buzzword compliant.  I have a use for wavelets, but they're not entirely proven even in my own mind.  Quite a lot of basic research would yet need to happen before they find a home in Vorbis.  I can say, however, they were promising enough to devote alot of time to primary research related to their use alongside the [now] tried-and-true MDCT.