IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Ogg Vorbis Historic, Evolution of Vorbis Tags
Guest_OggZealot_*
post Nov 14 2003, 12:20
Post #1





Guests






Hi all, first post here, so first congratulations to Hydrogen Webmasters ...
I started using Ogg Vorbis with version 1.0 Final so I don't know anything about earlier versions & I didn't found much informations on Ogg Vorbis History on the net so I made my own one but it's very incomplete, can you all plz fill the holes very carefully, so that this post filled with your answers could become an easy reference on Ogg Vorbis History, & even -maybe- be added to the important topic of the Ogg Vorbis forum if answers are very good, thx.

Official Versions:

N1: Alpha Versions ??? Does it even exist ?
Tag (Vendor String) = Xiphophorus libVorbis I 20000508 = V1.0 Beta 1 & Beta 2 ==> N2: How good was Vorbis at Start (Compared to Lame MP3) + N3: Beta 1 to Beta 2 major/minor improvements ?
Tag (Vendor String) = Xiphophorus libVorbis I 20001031 = V1.0 Beta 3 ==> N4: Beta 2 to Beta 3 major/minor improvements ?
Tag (Vendor String) = Xiphophorus libVorbis I 20010225 = V1.0 Beta 4 ==> N5: Beta 3 to Beta 4 major/minor improvements ?

Tag (Vendor String) = Xiphophorus libVorbis I 20010615 = V1.0 RC1 ==> N6: Beta 4 to RC1 major/minor improvements ? (Garf said: RC1 = Beta4 + complete decoder, RC1 = Point of decoder compatibility with future versions but has a little bug so compatility really started with RC2, already better than MP3, Garf joined)
Tag (Vendor String) = Xiphophorus libVorbis I 20010813 = V1.0 RC2 ==> N7: RC1 to RC2 major/minor improvements ? (Garf said: RC2 = first with channel coupling + bugfix in decoder, sometimes Beta 4 perfoms better than RC2)
Tag (Vendor String) = Xiphophorus libVorbis I 20011231 = V1.0 RC3 ==> N8: RC2 to RC3 major/minor improvements ?

Tag (Vendor String) = Xiph.Org libVorbis I 20020717 = V1.0 Final ==> N9: RC3 to V1.0 major/minor improvements ?
(Edit1: V1.00 Final's Lossless Channel Coupling start at Quality 6 with a little jump on average bitrate while in older Vorbis versions it starts at Quality 5 with a noticable jump of 20kbps on average bitrate, & that's the reason why you may find some RC3 Vorbis files encoded at quality 4.99, but encoding at 4.99 (or 5.99) to gain space is now obsolete")

Edit3:
Accidental Version:
Tag (Vendor String) = Xiphophorus libVorbis I 20011217 = V1.0 RC3 ==> Xiphmont said: "The info version was accidentally not updated in one of the windows builds, so that piece of software had the wrong version for the rc3 version"

Unofficial Versions:

Rarewares Compiled Versions:
Actual Version:
Tag (Vendor String) = Xiph.Org libVorbis I 20030909 = V1.01 ==> N10: V1.0 to V1.01 minor bugfixes in the code source, right ?
==> This unofficial Rarewares V1.01 tag may evolve if bugs are found before the official Xiph release of V1.01, right ?
Old Version:
Tag (Vendor String) = Xiph.Org libVorbis I 20030308 = V1.01 ==> Rarewares homemade compil of V1.01 (Read 20030803 ?)
N11: Anything special to say about Rareware Compiled Versions ?

Garf Tuned Versions:
Tag (Vendor String) = Xiphophorus libVorbis I 20010816 (gtune 1) Based on V1.0 Beta rc2 ==> GT1 Garf Pre-Echo Tuned for 180NKbps & some others unknown fixes, right ?
Tag (Vendor String) = Xiphophorus libVorbis I 20011014 (GTune 2) Based on V1.0 Beta rc2 ==> GT2 Garf Pre-Echo Tuned for 180NKbps & some others unknown fixes, right ?
Tag (Vendor String) = Xiph.Org/Sjeng.Org libVorbis I 20020717 (GTune 3, beta 1) Based on V1.0 Final ==> GT3b1 Garf Pre-Echo Tuned for 180NKbps & up, right ?
Tag (Vendor String) = Garf said: No Tags ! Deleted Just After Its Creation Based on V1.0 Final ==> GT3b2 Garf Pre-Echo Untuned for 180NKbps & up, Experimental Version Deleted Due to Bad Results
N12: Anything special to say about Garf Tuned Versions ?

Future Versions ? :
V1.1 The Perfect Codec for Video (Theora) (Q2): Low Bitrates AAC Killer, Pre-Echo fixed, When it's Done, 6 months to 1 year, so April 2004 (Optimistic) - October 2004 (Pessimistic), right ?
V1.2 The Perfect Codec for Lossy Audio (Q5-6): High Bitrates MPC Killer, The sooner the better, let's say middle of 2005 smile.gif (Complete Extrapolation), right ?
N13: Anything special to say about next Monty Versions ?

Version Time Travel:
Beta 2 to Beta 3: 5.5 Months
Beta 3 to Beta 4: 4 Months
Beta 4 to RC1: 4 Months
RC1 to RC2: 2 Months
RC2 to RC3: 4.5 Months
RC3 to V1.0: 7.5 Months
V1.0 to V1.01: 14 Months
N14: Personnaly I don't think 14 months is "so" long, considering that if the codec itself didn't evolved much, the support for the codec evolved much during that year which is the same thing for me ...
If you think Ogg Vorbis development is slow, just have a look in Xiph IRC channels & stay there a couple of day ... you'll soon realize how hard the Xiph developing team is working to deliver its best ...

Thanks Xiph.

This post has been edited by OggZealot: Nov 16 2003, 22:42
Go to the top of the page
+Quote Post
sld
post Nov 14 2003, 12:28
Post #2





Group: Members
Posts: 1015
Joined: 4-March 03
From: Singapore
Member No.: 5312



Not that I'm being a spoilsport here... but wouldn't Xiph.org know best about the history of the Ogg codecs? It would make more sense for them than for HA.org to dedicate a few webpages to this. However, I am quite sure that there are a few Ogg users and developers here who will be more than happy to contribute.

This post has been edited by sld: Nov 14 2003, 12:31
Go to the top of the page
+Quote Post
kjoonlee
post Nov 14 2003, 12:43
Post #3





Group: Members
Posts: 2526
Joined: 25-July 02
From: South Korea
Member No.: 2782



By the way, I personally think Xiphophorus libVorbis I 20011217 should be seen as a variant of Xiphophorus libVorbis I 20011231, both of them being V1.0 RC3 proper.

You can download the old sources and binaries at the Vorbis source archives if you want to check out the sound quality of older versions of the encoder. If you're not unfamiliar with compiling from source, you should be able to build the programs easily. If unfamiliar, there are some (if not all) binaries to check out too.

Beta 1 encoded files can't be decoded by current decoders, but you can decode beta-2-or-older files fine.

Have fun. smile.gif

PS. There's a Vorbis encoder older than oggenc, but I forgot what it was called.

This post has been edited by kjoonlee: Nov 14 2003, 12:44


--------------------
http://blacksun.ivyro.net/vorbis/vorbisfaq.htm
Go to the top of the page
+Quote Post
Guest_OggZealot_*
post Nov 14 2003, 14:27
Post #4





Guests






==> sld, if Xiph didn't took the time to build an historic
maybe it's because they have more interesting things to do ...
maybe an Vorbis historic already even exist somewhere & I just didn't find it ...
so if anyone know good links on Vorbis changes, it's always welcomed.
I personnaly think Xiph developpers are multimedia visionnaires,
& so should be encouraged with our unfailing support ...
Edit4:
This topic is NOT the right place to complaint on Xiph: Xiph Rulz !

==> Kjoonlee, thks for the old Vorbis archive links,
Unfortunatly I am still a very basic WinXP user so I will not be able to test anything
that does not look as an "oggenc.exe" compatible with EAC, but the link is good.
Your post make me noticed that people might answer to my topic by subjective listening tests & this topic is not meant to speak about ABXing the quality of different Ogg Vorbis encoders so :
Edit5:
This topic is NOT the right place to discuss audio subjective opinions,
This topic is the right place to share audio objective informations.

What I want to achieve with this historic is mainly to be able to see the difference "on paper" between the different Ogg Vorbis encoders, then only I might begin listening test to see if these theorical upgrade to the code are really hearable to me:
If you own anything that looks like a ChangeLog plz post it !

Have Good Posts smile.gif

This post has been edited by OggZealot: Nov 14 2003, 17:29
Go to the top of the page
+Quote Post
sld
post Nov 14 2003, 15:00
Post #5





Group: Members
Posts: 1015
Joined: 4-March 03
From: Singapore
Member No.: 5312



QUOTE (OggZealot @ Nov 14 2003, 09:27 PM)
Plz, sld go whin & spoil elsewhere, I have nothing against you but...
I am tired of people saying Xiph is slow & that kind of bad thing ...


... ...

Edit: Damn, I forgot I'm not supposed to feed trolls.
Edit2: Maybe it's just the language barrier.

This post has been edited by sld: Nov 14 2003, 17:59
Go to the top of the page
+Quote Post
kjoonlee
post Nov 14 2003, 16:16
Post #6





Group: Members
Posts: 2526
Joined: 25-July 02
From: South Korea
Member No.: 2782



QUOTE (OggZealot @ Nov 14 2003, 10:27 PM)
Yes,
Tag = Xiphophorus libVorbis I 20011217 = Xiphophorus libVorbis I 20011231 = V1.0 RC3
as far as I know the sourcecode is the same only the tags differ by mistake,
I thought it was clear with the Monty quote on it, but nevermind.


Well, why did you mention 1217 as a "Work In Progress" release, then? 1217 is 1231 in every way except for the vendor string. (It isn't called the tag, it's called the vendor string.)

QUOTE (OggZealot @ Nov 14 2003, 10:27 PM)
Kjonlee, thks for the old Vorbis archive links,
Unfortunatly I am still a very basic WinXP user so I will not be able to test anything
that does not look as an "oggenc.exe" compatible with EAC, but the link is good.
Your post make me noticed that people might answer to my topic by subjective listening test between different Vorbis encoder versions & this is absolutly not what it is intended, so plz if you plan to reply to this topic, try to give mainly objective information like:


The name's kjoonlee, thank you very much. Well, the only way to test audio codecs in a meaningful way is to take a listen. Sound quality is subjective, you know. From the Xiph.org Fidelity measurement terminology page:

QUOTE
Perceptual codecs are most concerned with subjective, not objective, quality.


QUOTE (OggZealot @ Nov 14 2003, 10:27 PM)
What I want to achieve with this historic is mainly to be able to see the difference "on paper" between the different Ogg Vorbis encoders, then only I might begin listening test to see if these theorical upgrade to the code are really hearable to me.


I'm of the opinion that differences on paper are pretty useless unless you can hear the differences yourself. IMHO the listening should be done first.


--------------------
http://blacksun.ivyro.net/vorbis/vorbisfaq.htm
Go to the top of the page
+Quote Post
Guest_OggZealot_*
post Nov 14 2003, 19:09
Post #7





Guests






Thanks for trying to help sld & Kjoonlee & sorry if my post wasn't clear from start,

==>Kjoonlee I know very well the "Trust your ears" sentence & that kind of answers, but if trusting your ears is fundamental, it's only a part of the job for me ... what I want is to go further than that easy kind of answers I often read on audio forums, sorry but I noticed that "Trust your ears" was often used to hide a lake of real "on paper" codec knowledge & as much as possible I want to avoid that kind of answers.
Why should the fact that I trust my ears would reduce the fact that I also have to trust my brain ? sorry but I trust both my ears & brain when it comes to lossy audio
& knowing Vorbis History is important as I wanna understand "theoricaly" WHY when I trust my hears I may hear difference between encodings.
I don't wanna offend you by telling that the "Trust your ears"like'answers are wrong, but this kind of answers are only partially true, I mean when it comes to lossy audio what is important (IMHO) is not pure quality but quality/size ratio,
if we would all only "trust our ears" we would all use lossless codecs ...
so yes even in audio there is a time when you have to stop listening & start to thinking (sacrilege !!!), I would have wished my topic could have covered the "boring" technical issues which are generally poorly explained ...
To my mind difference on audio tests & on paper are complementary,
audio tests are the easiest part as I can ABX it all by myself, but I definitly miss the vorbis ChangeLog Info which I can't guess ...

This was my opinion post, no more opinions posts plz smile.gif I want fact & knowlegde !!!

This post has been edited by OggZealot: Nov 15 2003, 08:20
Go to the top of the page
+Quote Post
ScorLibran
post Nov 14 2003, 20:06
Post #8





Group: Banned
Posts: 769
Joined: 1-July 03
Member No.: 7495



Don't forget OggSquish. (Initial development, pre-alpha, 1999-and-earlier.)

smile.gif
Go to the top of the page
+Quote Post
kjoonlee
post Nov 15 2003, 03:26
Post #9





Group: Members
Posts: 2526
Joined: 25-July 02
From: South Korea
Member No.: 2782



QUOTE (OggZealot @ Nov 15 2003, 03:09 AM)
I don't wanna offend you by telling that the "Trust your hears"like'answers are wrong, but this kind of answers are only partially true, I mean when it comes to lossy audio what is important (IMHO) is not pure quality but quality/size ratio,
if we would all only "trust our hears" we would all use lossless codecs ...


I don't get what you mean. I think that if only we were all to trust our ears and hearing only, almost everybody would be using lossy. Lossless can be overkill (or unwise) for many occasions, especially if you have to consider quality/size ratios. Have you heard about the placebo effect? It can be very powerful sometimes.

We all need to trust our ears and hearing.

This thread is about "Ogg Vorbis History." IMHO, testing older encoders and sharing the experiences would count as history, and would be relevant to the topic regardless of what you want in the thread.

By the way, if you want to know about Vorbis history, you can see for yourself on the xiph.org cvs mailing list archives. I wonder if it's possible to fetch the CVS changelogs.

This post has been edited by kjoonlee: Nov 15 2003, 03:27


--------------------
http://blacksun.ivyro.net/vorbis/vorbisfaq.htm
Go to the top of the page
+Quote Post
sld
post Nov 15 2003, 05:11
Post #10





Group: Members
Posts: 1015
Joined: 4-March 03
From: Singapore
Member No.: 5312



QUOTE (OggZealot @ Nov 15 2003, 02:09 AM)
sorry but I noticed that "Trust your hears" was often used to hide a lake of real "on paper" codec knowledge & as much as possible I want to avoid that kind of answers.

Well, with a pair of good ears and some reliable audio hardware, amost anyone who "trust your hears" will easily tell you that Microsoft's "on paper" "CD-quality at 64 kbps" is pure rots. Of course Xiph.org doesn't indulge in such deceptive marketing (AFAIK), and of course by "trusting your hears" you can also tell that Vorbis at 64 kbps sounds different from the original.

QUOTE
I mean when it comes to lossy audio what is important (IMHO) is not pure quality but quality/size ratio,
if we would all only "trust our hears" we would all use lossless codecs ...


Those who use lossless codecs do it for the 'warm and fuzzy feeling', and they have the money to spare in getting extra storage space. If we were all to "trust our hears", we'll be using Vorbis and Musepack from -q4 to -q7, AAC at about 192 kbps or -normal (encoder-dependent), or LAME from -aps up till -api.

This post has been edited by sld: Nov 15 2003, 05:12
Go to the top of the page
+Quote Post
xiphmont
post Nov 16 2003, 22:03
Post #11


Xiph.org


Group: Developer
Posts: 169
Joined: 24-September 01
Member No.: 16



Let me throw a few tidbits out :-)

Comparing against LAME is a decent idea, but don't forget that LAME had also just entered a period of rapid improvement at the time Vorbis was getting started. Both were moving targets and the projects cooperated, so there was cross-pollination and more than a little friendly rivalry. Those were some good days with the LAME and Ogg (and Green Witch) folks going out for sushi together :-)

Vorbis of today and LAME of today are both *way* better encoders than they were in 2000, on that we can all agree.

> Alpha Versions ??? Does it even exist ?

Only in CVS. We didn't announce or release anything prior to first beta, and only released first beta at the time we did because of all the attention we were getting from slashdot. My feeling is that first release was a few months too early, but first release is always kind of arbitrary.

>Beta 1 & Beta 2 ==>: How good was Vorbis at Start (Compared to Lame MP3) + N\x{00B0}3: Beta 1 to Beta 2 major/minor improvements ?

I always felt we were either significantly better or at least had an edge over LAME at all times. Of course, LAME continued improving as well and eventually considerably surpassed how far I thought mp3 could be pushed. Eventually, though, Mp3 did find its wall.

I honestly don't remember the major improvements between 1 and 2... most likely the only changes were added features and better tuning.

> (Edit2: Kjoonlee said: "Beta 1 encoded files can't be decoded by current decoders, but you can decode beta 2 or older files fine."

Not true. Everything produced from beta-1 forward is decodable in a modern decoder. Some hardware players do neglect older beta files because decode of the techniques used in older beta files are far less memory/CPU efficient.

> Beta 3 ==>: Beta 2 to Beta 3 major/minor improvements ?
> Beta 4 ==>: Beta 3 to Beta 4 major/minor improvements ?

I'd have to go through CVS to pull out specific things. The tuning improved each release, as did encoder flexibility. The most relevant changes were moving toward smaller codebooks, lower memory usage and faster decode (these things are fixed in the decode requirements of a specific file. Older betas used gargantuan codebooks, and even today the low memory version of Tremor requires ~70kB to decode an older beta file while a 1.0 file decodes in as little as 20kB).

Beta 4 used modern codebooks and residue encoding, but still had no channel coupling, no bitrate management and still used floor 0.

> RC1 ==>: Beta 4 to RC1 major/minor improvements ? (Garf said: RC1 = Beta4 + complete decoder, RC1 = Point of decoder compatibility with future versions but has a little bug so compatility really started with RC2, already better than MP3, Garf joined)

Garf more or less nailed it.

> RC2 ==>: RC1 to RC2 major/minor improvements ? (Garf said: RC2 = first with channel coupling + bugfix in decoder, sometimes Beta 4 perfoms better than RC2)

Sometimes but not often; RC2 was the most radically different Ogg encoder since beta 1, and was more different than the changes of all the other releases combined. Had we had infinite time to sit around and wank, this is the release that should have been beta 1. It used coupling, modern codebooks, floor 1 and was in general the prototype of Vorbis as it is now.

> RC3 ==>: RC2 to RC3 major/minor improvements ?

rc2 was lots of new tech for testing, rc3 took most of the rough edges on the new encoding strategies in rc2. rc3 also finally added bitrate management, but it turned out not to work very well yet.

rc3 also added low-bitrate encoding down to 64kbps, again, as an iffy prototype sort of feature that 1.0 got right.

There was also a labelled but unreleased rc4 that became 1.0 with minor tuning changes.

> V1.0 Final ==>: RC3 to V1.0 major/minor improvements ?

a) new noise/tonaity estimation techniques (LAME adopted this strategy too)
B) 'noise normalization' to preserve noise energy in very low bitrate modes
c) reworked bitrate management (far superior to rc3)
d) CD rate stereo down to 45kbps and eventually low-rate 8kHz down to 6kpbs
e) improved stereo metrics

(A couple folks in HA will look at )e and snicker because even 1.0's stereo seems a little weak now, and at low bitrate, it was still the weakest point in 1.0. Stereo improvements are probably the biggest quality improvement priority for 1.1)

> (Edit1: V1.00 Final's Lossless Channel Coupling start at Quality 6 with a little jump on average bitrate while in older Vorbis versions it starts at Quality 5 with a noticable jump of 20kbps on average bitrate, & that's the reason why you may find some RC3 Vorbis files encoded at quality 4.99, but encoding at 4.99 (or 5.99) to gain space is now obsolete")

Correct

> Accidental Version: Xiphophorus libVorbis I 20011217 = V1.0 RC3 ==> Xiphmont said: "The info version was accidentally not updated in one of the windows builds, so that piece of software had the wrong version for the rc3 version"

Correct. The win32 build team jumped the gun by about ten minutes, and version info is the last thing updated before a release, so the tag on the win32 build was wrong. It was the only difference between win32 and other platforms.

> Xiph.Org libVorbis I 20030909 = V1.01 ==>: V1.0 to V1.01 minor bugfixes in the code source, right ?

Yes, but a few did affect output.

> This unofficial Rarewares V1.01 tag may evolve if bugs are found before the official Xiph release of V1.01, right ?

Correct, but it hasn't changed much/any since then. It's not been released officially only because we've been short win32 release engineers to finish the process. Oddsock is now working on it as I type this.

> Old Version:
> Xiph.Org libVorbis I 20030308 = V1.01 ==> Rarewares homemade compil of V1.01 (Read 20030803 ?): Anything special to say about Rareware Compiled Versions ?

I don't track them, so I don't know. However, they usually build from CVS HEAD, which is always considered stable... so the later the version, the more bugs that are fixed. 1.0.1 had some start and end of stream pop bugs that got fixed in later versions (these bugs had always been there). I'd have to check CVS to find the exact date, but 0909 should be considered strictly superior to 0808.

> Personnaly I don't think 14 months is "so" long, considering that if the codec itself didn't evolved much, the support for the codec evolved much during that year which is the same thing for me ...

We've been concentrating (too much I think) on helping third party vendors adopt and implement Ogg and Vorbis. This succeeded, but it's pulled us far away from core development, and we haven't really gotten back to it. I'm not going to make any ETA estimates or promises, but I'm itching to get back to coding the core. Vorbis is nowhere near final potential yet, we've just not been working on that.

Sorry for not using the typical quite style, but I hate writing markup if I can avoid it :-)

Monty

This post has been edited by xiphmont: Nov 17 2003, 06:37
Go to the top of the page
+Quote Post
Guest_OggZealot_*
post Nov 17 2003, 00:40
Post #12





Guests






Thks a lot for all these good answers Xiphmont, it's really nice biggrin.gif

I am very satisfied with your answers, as it is good enough to help newbies like me to see clearer through the technical differences between Ogg Vorbis versions & it's not so much complicated that anyone can't understand it blink.gif
Thks to you, I will now consider RC2 as the real start of Ogg Vorbis in its mature age.
The weakest parts of your answer are for the very early Beta versions but it doesn't really matter anymore as you told us RC2 was what Beta 1 should have been, so I forgive you for not reminding these very old changes happy.gif
I think I globally achieved what I wanted with this topic, only details can be added now, & I think that so few people have the knowledge on these minor changes that it would be cleverer to ask Xiph himself to consider adding 2 new pages on Ogg Vorbis official website, one for a complete Ogg Vorbis historic & one with links to dl the old versions of Vorbis Tools (for those who want to test them) I know it ask some work ... but I think it would help a lot of people.
Maybe a little word of John33 & Garf on there own Vorbis versions would be welcomed if I missed anything ... but I don't use non-official versions personnaly so ...
I am glad to learn that the development will re-focus on the codec itself now that we will all have an I-River portable for Christmass biggrin.gif , I also would like to thks Oddsock for his work on the win32 V1.01 ...
Now I can go back to ABXing RC2/RC3/V1.00 to see at which point of Vorbis developpment I will start using Vorbis for my personnal use, so thks for all.

wub.gif Long Live to Vorbis wub.gif
Go to the top of the page
+Quote Post
kjoonlee
post Nov 20 2003, 10:50
Post #13





Group: Members
Posts: 2526
Joined: 25-July 02
From: South Korea
Member No.: 2782



QUOTE (xiphmont @ Nov 17 2003, 06:03 AM)
> (Edit2: Kjoonlee said: "Beta 1 encoded files can't be decoded by current decoders, but you can decode beta 2 or older files fine."

Not true. Everything produced from beta-1 forward is decodable in a modern decoder. Some hardware players do neglect older beta files because decode of the techniques used in older beta files are far less memory/CPU efficient.

Sure enough, I was mistaken. Building beta 1 on a Debian system let me find out that a beta 1 file can be played back using modern decoders.

I guess I had done something wrong when I had built it on Windows with cygwin. Mea culpa.

This post has been edited by kjoonlee: Nov 20 2003, 10:51


--------------------
http://blacksun.ivyro.net/vorbis/vorbisfaq.htm
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 25th April 2014 - 05:12