I understand that in the long run the vorbis format is superior to mp3 but ogg hasn't been "fine tuned" as much across all quality settings as LAME. So for archiving, am I better off using LAME/mp3 for the time being? Or is a slightly fine tuned ogg better than a finely tuned mp3? (I'm using the extreme preset in LAME right now and have been playing around with -q6.25 in ogg - but I think I've read that ogg hasn't been fine tuned that much above -q6 yet.)

B)

I wouldn't use ogg for high-bitrate archiving at this time. Ogg will treat you right for your mobile needs (64-128kbs). I find that LAME is doing a better job at 190+ (VBR, of course), but hope that ogg will kick arse in the future.

But, I don't know, and someone here will.

Hmm, I always thought Ogg Vorbis was better than MP3 in most respects?

IMHO I recommend Ogg Vorbis over MP3 for archiving.

Laters
AgentMil

If you are planning to go with ogg in the next future, you should begin yet with vorbis. Don't forget that mp3 isn't gapless (I switch to mpc for that reason).

I use ogg 6 for my files. Don't forget that lame--extreme produces much bigger files compared to ogg6. and ogg has no time-seek problems, ogg is fast, decoding works perfect, ogg has good tagging (well, id3v2 is ok too), ogg is free, ogg is open, u'll get bitrate peeling soon.. etc etc. ogg files are just perfect.

it's the first time I have the feeling that I have real CD copies on my hd. with mp3 I always had problems with gaps beween tracks etc.

Is there any progress in tweaking high -q settings?
Or is everybody occupied with some other tasks?

yes, even though ogg vorbis isnt the best in high bitrates its still good enough to beat mp3 at any bitrate. An example: a vorbis 224kb file will sound much better than a 320kb mp3, using vorbis 1.0 that is...

MP3 192kbps cuts off frequency above 18.5 Khz
WMA9 At 116 Kbps cuts off above 15.5 Khz (it sucks)
OGG 128 does not cut any frequency (??? Yes - I used spectralab). It has some problems above 18500 but it`s not a cutoff. I can send picture from spectralab to anyone (e-mail needed).
MPC 202.4 Kbps has much more problems tahn OGG 128 above 20000hz.

OGG Q=4.25 sound very good to me (SBLive, headphones music: classical, metal, newage). I don`t know why some people use q=6 or q=6.5?
There is no hearable difference between MPC202.4 and OGG Q=4.25 for me.

Using frequency cutoffs to measure audio quality isn't the right way to go about evaluating audio codec performance.

Search around the forum and you will see why. Tonal purity!

YES!
I know that, only valualbe test is hearing test!
But in WMA9 i can easily hear cutoff above 15.5 Khz
Analysing spectrum can give additional informations...

AgentMil: Can you hear difference between OGG (latest version) Q=4.25 and MPC (latest version) c.a. 200 kbps (using ABC/Hidden Reference) ?

Ah ok, sorry if I offended you.

Haven't used Ogg Vorbis since the 64kbits tests. I am a MusePack exclusive user .

I will try that out later tonight if I have time that is, been busy with typing heaps of documentation, preparing some research notes and got an assignment due in as well . But I doubt I will hear the difference, if you can't tell the difference I doubt I can. But a listening test can only prove that .

Considering most people still use BladeEnc, FhG and Xing-based encoders for making MP3 files I'd say Vorbis wins hands down if look at a greater perspective...

Couldn't agree more.

I can think of no reason why the casual listener shouldn't change to Vorbis, I know a lot of people who think that --alt-preset "200+" bitrates are too large and end up going 128-192CBR anyway. I think it's safe to say that Vorbis triumphs in that bitrate field. As far as archiving goes, I cannot hear the difference between Vorbis q5 and the original wave file.

More people need to give Vorbis a go and spread the word. Let's leave Xing on the funeral pyre once and for all.

Amen to that!

Yes, Vorbis is better than most mp3.
But it is not tuned yet.

Is 224kbps Vorbis better than 320 kbps LAME?
I don't know for sure. It could be.
The point is, even if it is better than mp3, it has potential to be even better than it is right now at -q6+.

Imagine how good it will be when it is tuned.

How to check that? I`m sure that i can`t find any difference between OGG224KBps and MP3 320kbps - in quality.

[quote=Cobra,Sep 22 2002 - 08:45 AM][QUOTE=AgentMil,Sep 22 2002 - 12:35 AM]Can you hear difference between OGG (latest version) Q=4.25 and MPC (latest version) c.a. 200 kbps (using ABC/Hidden Reference) ?[/quote]
Invert position : Can you hear difference between MPC (latest version) quality=4.25 and OGG (latest version) c.a. 200 kbps [-q6] (using ABC/Hidden Reference) ?

[quote=guruboolez,Sep 22 2002 - 03:58 AM][QUOTE=Cobra,Sep 22 2002 - 08:45 AM][QUOTE=AgentMil,Sep 22 2002 - 12:35 AM]Can you hear difference between OGG (latest version) Q=4.25 and MPC (latest version) c.a. 200 kbps (using ABC/Hidden Reference) ?[/QUOTE]
Invert position : Can you hear difference between MPC (latest version) quality=4.25 and OGG (latest version) c.a. 200 kbps [-q6] (using ABC/Hidden Reference) ?

[/quote]
Even if not... (i`ll check it!)
OGG is better in range <=96kbps and MPC is not better in range =>128kbps

So i use OGG

Maybe it's something else you can hear... Analyzing the spectrum gives you practically no useful information about sound quality. They are hardly an indication.


I can hear a difference with ogg -q 6 on many files, and with ogg -q8 on some test files. MPC standard/extreme is nearly always transparent (even there).


I'd say yes, because MP3 is only a little bit better at 320 kbit than with APS (~200 kbit), while ogg still gets better at high bitrates (noise shaping comes to my mind), or is already better (transients).
For me the question is, how low can ogg go, bitrate-wise, and beat APS?


For you, but for many members of this board, MPC is clearly better. (Try castanets.wav for example)

Where can i find castanets.wav and other sample files?

http://www.mp3dev.org/mp3/gpsycho/quality.html

http://www.ff123.net/samples.html

http://www.audio-illumination.org/forums/i...t=ST&f=1&t=2441
http://www.pcabx.com/technical/reference/

I can`t hear any difference So OGG is for me!!!
I used oggdropXPd 1.5

It`s for original castanets.wav coverted to OGG:
Length : 0:06
Average bitrate : 145 kbps
File size : 119,966 bytes
Nominal bitrate : 136 kbps
Xiph.Org libVorbis I 20020717

It`s for castanets-060.wav converted to OGG:
Length : 0:00
Average bitrate : 198 kbps
File size : 14,875 bytes
Nominal bitrate : 136 kbps
Xiph.Org libVorbis I 20020717

So you can see, OGG Encoder used 198kbps for critical section of castanets.

MPC is not better than ogg vorbis > 128 kb/s ?
Damn ! Have you ever heard about pre-echo ? It's a (big) problem with Vorbis I. Many people don't care about. They are maybe right...

But I think that a 300 kb/s encoding must be totally transparent : and I can easily ABX vorbis at -q8 (and -q8.5) on castanets.wav. I can ABX vorbis -q9 (and sometimes -q9.5) on another castanet sample... I was able to do it with mpc --extreme (--quality 6, 200 kb/s), but recently, I can not reproduce this ABX performance.

On electronic sample, Vorbis I failed at -q10 (ABX 16/16), when mpc --standard [but higher bitrate than ogg -q10 ] was totally transparent.


Be careful. At 130 kb/s, Vorbis is certainly a winner. But you can not extrapolate these results to higher bitrate. MPC --standard is a happax in audio encoding : very robust preset for most critical samples for only 170 kb/s...

You are using special test samples. I`ve encoded some CDs (Iron Maiden, Vangelis, Laser Dance, Mozart, Preisner...) and i can`t hear any difference.
Sample files are good for audiophiles and for tuning. For "normal" music OGG Q=4.25 is good for me.

the « special sample » came from regular discs. I found pre-echo problems with Jordi Savall early recording, Penderecki Symphony, Pierre Henry works...

Castanets.wav is showing difficulties that will occurs with many discs....

Yes, if you want. If you're happy with -q4.25 and some rare artifacts, I understand you. But you cannot say that MPC is not better in range =>128kbps : this is not true. AAC & MPC are better than Vorbis > 160 (at least for pre-echo).

But most of the samples are taken from real music! There is really no significant difference between test samples and music, it's not hard to find something that is difficult to encode. A guitar or a violin sound is often enough to provoke artifacts, and in electronic music, there are many more possibilities. Not taking this into account is naive.

So, MPC 200kbps is not better than OGG Q=4.25 >FOR ME<
BTW. There are many people that think 80kbps OGG is good...
I`ll try to find artifact-loaded OGG files For now I don`t know any...

PS. What audio hardware you have? Orpheus headphones or what?

Heh, this is pretty weird way to express transparency. I mean if Ogg -q 4.25 is transparent for you then even "original wav is not better than -q 4.25 - >FOR YOU<"...

I mean if something is also transparent at lower bitrate settings, it's only confusing to compare it to higher quality setting (which many people can hear sounds better) and say that it's not better...

Not to mention that you did the biggest "crime" in psychoacoustic quality testing in this thread -> used frequency analysis and made conclusions and ranked codecs based on it...
Buut you are forgiven..

Did I say that Codec X is better than Codec Y because of cut-off frequency etc.?

NO!

care to elaborate?

->

What you saw there, Cobra, can't be classified as a problem this easily. Lossy audio codecs try to cheat the ear, not the eye. Not to mention that 20 KHz is the generally accepted upper hearing range of humans.

some old results, not sure what changed after release at higher bitrates (& iam very far from golden ears):

------------------------------------------------
knocking on the heaven's door

ogg 5 (rc3)
12/16 (p=0.038) = abxed (bass and drums)

ogg 6 (rc3)
'sennheiser hd 265' headphones
18/24 (p=0.011) = abxed, wow (part after 1:55)
-----------------------------------------

We Use The Pain (guano apes)

ogg-vorbis -q4 (rc3)
'sennheiser hd 265' headphones connected to my sony md component. (had a small beer before)
12/16 (p=0.038) = abxed (part after 45s)

ogg-vorbis -q4.9 (rc3)
'sennheiser hd 265' headphones
5/16 = not abxed
-----------------------------------------

i would say q5 should be transparent for me on most samples.

Cobra:

It is true that you may be able to get transparency on your setup at those settings, but conversely, your source isn’t known for it’s musical fidelity. I can achieve near transparency at about q8, and getting it at q9, but then again I am a nitpick, and do not tolerate artifacts.

I also upsample my rips to 48K using CoolEdit 2000 to obviate forced AC97 resampling, and my next card will have ASIO 2.0 support so I can completely get kmixer out of my playloop. (And will not be an Audigy2 since the ASIO driver is still fixed at 16/48)

I haven’t listened to MPC much at all, and I’ll have to take a stab at it. Why not, I’m not married to Ogg.

To compare, I did my tests with LAME 3.92 at --alt-preset extreme, Ogg at q6 - q10 at even integer intervals, and used either my MegaWorks 510Ds or My Senn HD580s all off a TBSC. Playback was tested with CoolPlayer 2.05 or Winamp 2.81 w/ 2.01 Waveout and MAD / Vorbis Plugin and Wav playback against original Wav Rip. All done at 44.1 Sample. Final rips are upsampled to 48K using CoolEdit 2000 at 400 setting.

@Chastity
You should check out the upsampling output plugin for winamp. gets rid of the need to mess with cool edit to reach 48khz sample rate. Works fine on my sblive, and seems to lower the noise floor. Sorry, don't have the link handy though. Anyone?

http://www.blorp.com/~peter/
Winamp v2.x plug-ins
waveOut output plug-in replacement (with sourcecode, includes version with SSRC resampling)

If you have Turtle Beach Santa Cruz you don't need to resample at all, SRC quality of that card is very high.

I don't know if there would be a real audible improvement, but SSRC resampling quality is higher.

Well, I would like to add to a few points here:

1) The reason for the resampling was to address AC97 resampling on my TBSC and Audigy. (I use both cards, one for music/dvd, one for gaming) Since both resample, I wanted to get a 1-1 pass on the data to the soundcard. Kmixer (the part which does this by Microsoft) doesn't have the best SSRC routine. The SSRC plugin by Peter for Winamp is an improvement, and definately would help for the Live due to their lousy hardware routines, was mediocre when compared to CoolEdit 2000's and Nero's (if set to Accurate) Reason being is that the routines are very granular, and resamples the data to keep the original sound as much as possible, and CoolEdit even does a second pass to remove artifacts. Very quality software, and you wouldn't notice that the sample was resampled. Peter's plugin does improve the bass response, which was an issue for KMixer, but overall the audio was still muted. CoolEdit's results are transparent. Plus you can batch resample it for convience, and yes it is slow, but quality is more important to me.

2) On XP, things have changed, since KMixer was updated with SP1, and improved it's SSRC routines. Now 44.1 resamples are more transparent, and there are quite a few white papers on the subject if anyone wants them on Microsoft's Developer Site. For others, I suggest going into Window's Sound Control Panel applet, and move the quality sliders up to Best. On some of the 9x OS's, Best is not default, and KMixer will not use its best algorithms. I am very glad they redid the kmixer, since I don't need to resample anymore.

Using ASIO is the best solution, since this bypasses KMixer all together, and KMixer has a 85dB SNR limit anyways on it's best setting. Problem here for me is ASIO on the Audigy is fixed at 16/48 and no ASIO for the TBSC. (Before someone says it, Cubase installs a Wave Map and DirectX wrapper for ASIO, which isn't the same thing.)