Inspired by all the recent discussion/confusion about the quality of Vorbis at archiving bitrates (>160) vs. the 'old' GT2 I was just asking myself this question.
I know that Musepack is clearly better at those bitrates, so please no "Use MPC!" rants...

<edit>
Damn! It should be entitiled "If I would start archiving with Ogg Vorbis now, I'd use..." of course!
</edit>

dev0

fixed.

I'm very happy with the -q8 setting myself, so i use that for archiving

Could someone please explain to me what "archiving" means? When I hear that word I have an image of storing the music in some dark place between the occasions you listen to it... In that case I archive my music in a box in my bookshelf. I store the music in whatever format I get it - CD mostly.

Do you have another definition of "archiving"?

I'd say Q1, but since it's not listed there, oh well....

-q 6 is what I'm using, and I find it OK, I haven't heard any deficiencies yet. But for "real" archiving, I'd use something else, like MPC --quality 6 or something.

To clarify things, for me "archiving" designates what is left to me, if the original CD gets destroyed.

Well, for archiving purposes I'm still using --alt-preset extreme LAME MP3's... mostly for historical (and LAMEness) reasons :-)

Like many people I started with --r3mix. After finding HA, I switched to --a-p standard and was in heaven, except that something subtle bugged me... on metal music some sharp transients sounded a tiny bit duller than the CD. Setting the same lowpass as r3mix (19.5 kHz) did the trick for me. To avoid cmd line hacking, and to get a safety margin for low-bitrate transcoding I switched to Extreme and lived happily ever after :-P

Yeah I could use the latest and greatest.. but the average transient response and 220kbps bitrates are good enough for me.

About OGG-GT2 @ 350kbps, I really don't see the point - because we don't know how much quality is gained !
When I'll switch, it'll probably be to MPC, when SV8 is implemented or even before. I like simple and efficient designs (only one complex part: the psymodel).


MadiZone I don't understand your approach ! I mean, you archive your music @ streaming bitrates ? What if something bad happens to your originals ? Will you still enjoy your music when using a good system ? Will it still be listenable after transcoding to something else (for portable listening) ?

With my brother's setup (cheap onboard audio & Cambridge soundworks) the 64kbps issues are clear. If I push it to 80kbps, I'm not annoyed by the sound anymore. That doesn't mean it's perfect.
Of course, 64kbps did sound good to my uncle... but you should know, that he wears a hearing aid and thus can hear nothing above 8-10kHz

Going a bit off-topic:
For other uses (portable listening w/ in-ear headphones) I use 96kbps AAC (last week I tried 64kbps AAC again, but it was too hissy, with cheap treble). I don't think it's an AAC issue though. The culprit is almost certainly the encoder bundled by Panasonic.

Archiving: Ripping CDs in a quality, which is transparent to me and most other people, so I don't have to mind if the original gets destroyed. The quality should be high enough to be transcoded without too much quality loss into the 64-128 kbps region for portable use.

350 kbps Ogg-GT2 makes sense to me, because if you really want to use Vorbis, this is the best quality you could possibly get. The bitrate (270-330) is not that much higher on my music (mostly Punkrock), than --alt-preset extreme (250-290) and Ogg Vorbis itself is just damn sexy .

dev0

<edit>
Made the bitrates a bit clearer.
</edit>

Dev0:
I completely agree with what you said, and I too find Vorbis is a pretty format but... I don't understand the advantage of using a transform codec at these bitrates ?

(as a side note, how can you claim up to 330kbps average with mp3? lol..)

Do you really store more significant data than you would by storing it in the time-domain ? I doubt it.

My approach is, say for archiving one might ask for (almost) guaranteed quality. That includes time-domain transient response guarantees, and accurate frequency and phase response in general.

For exemple, what does Vorbis GT2 @ 320kbps actually provide (or better: guarantee) in terms of transient response ? Is it really *that* sharp ? What about phase distortions (which, if I'm not mistaken, could introduce smearing and coloring) ?
In theory, if I got it right, with the kind of short-block flexibility that's built into the Vorbis format, it could very well beat AAC's transient response at these high bitrates. No ?

Maybe Garf could give us some insight here.

I have two 300 watt speakers connected to a $900 mixer/amplifier. I have a Sound Blaster Live.
I rip at 80 kbit, because I find it transparant. In 98% of all cases, I cannot tell Vorbis 1.0 Q1 from the original CD. In the 2% where I detect a loss, I encode at Q3 (112 kbit).

I'm 18 years old, so I don't consider by own hearing as bad. But whatever artifacts OGG Vorbis produces, it does a remarkable job hidding them from me. OGG Vorbis at Q0 sounds a bit flat, and at -Q1 it sounds like 64 kbit WMA.

That don't impress.. me much
just kidding.



I just don't understand this approach to the archiving concept.. but well, at least you exploit your harddisk space well :-)
What about transcoding though? If feel a bit worried about doubling (on average) the distorsion/noise etc.. Especially since you're "at the edge" already.

Or.. maybe you've found a Vorbis portable player yet?



Argh I can't stand 64 kbit WMA

Well, I trust Emmett when he claims that he has received "test units" from iRiver.
So I believe OGG Vorbis is coming within 100 days to iRiver.

I don't download a lot from the internet, but what I download, I make sure to download in at least 192 kbit MP3. Then I transcode it to Q1 OGG Vorbis.
90% of my music comes from original CD-source.
Until May 2002, I had mixed formats (RealAudio, MP3, WMA and VQF). ( http://www.hydrogenaudio.org/forums/index....=17&t=4387&st=0 )

Then I was hit by a virus that erased by harddrive, and it was time to re-do everything. Before beginning to rebuild my collection, I had some things in mind. I wanted transparent audio, I wanted compact audio, I wanted the same audio format, and possibly one I could use on portable players without having to transcode. I found OGG Vorbis to be the obvious choice because

1) It feels right (opensource - freedom - yada yada)
2) It sounds CD-quality at 80 kbit / Q1 ...
3) ... which makes it portable at the same time...
4) ... And great for internet exchange (P2P, e-mail, FTP, etc...)
5) I'm convinced portable support is coming very soon.
6) I'll invest in a (complimentary) Macintosh by the time I finish graphics school. OGG Vorbis is multiplatform.
7) It encodes fast (which is good, when you have 300+ CD's and a Duron 750)
8) It's frozen, and can only improve over time - and I truely believe it has a great lifespan.
9) It's just soo bloody flexible because of so many things.

Difficult to explain, me just likes

The fact that it's portable from the beginning, saves me from transcoding, and that's a plus, because transcoding is timeconsuming, quality-consuming and you have to keep track of which files are originals and which are transcodes.

Why would you bother doing this? I mean, for the few songs you would have in MP3, is it really worth sacrificing so much quality just so that you can say all your files are the same format?

Most MP3's on file sharing networks are already usually of poor quality -- if you take them and then transcode to such a low bitrate, the end result will just be that much worse. I hope for Vorbis's sake (reputation wise.. it's already hard enough to convince people that it's better than MP3), that you don't go back out and share these files again.

i did some reading somewhere on this board that things gets saturated at certain bitrate, is there a way to measure that or are there any real numbers out there? (as to what is the point where the actual quality wont really increase at all, or just so sligthly its not worth another bit) , or is that a 2 subjective thingy to evaluate at all ? (if not, i would be happy with some speculation (not proven) numbers as well and not just for vorbis).

oh, and my vote is for 'oggenc 1.0 @ -q 6' (most of my music at this moment is mpc iam afraid... )

-q1 (80kbps) is probably sufficiently low enough that the difference between encoding from a 192kbps MP3 and the original would be negligible. Transcoding may be evil most of the time, but it usually make sense when you need really far lower bitrate than the original encoding AND you don't have access to the original.

But I seriously have some doubts about -q1 and looking at the word 'archiving' in the topic....

Yeah.. perhaps if you search around, JonI made a whole load of graphs using EAQUAL and Vorbis RC3 quality modes. Trusting EAQUAL at high bitrate is not a good idea though...

Although I have produced bitrate/quality graphs in the past, I never used EAQUAL (and I forget the name of the person who did). In addition, xercist's bitrate/quality graphs for 1.0 are much prettier than mine ever were.

What a shame.

Hmm...
I'm suprised that only one person voted for oggencgt2 -b 999, which seemed to be better than oggenc1.0 -q 6 according to the tests done by guruboolez a few days ago.
Would anyone be interested in participating in a listening test involving those two settings?

-q 6 in OggEnc v1.0, for the mere fact over 7 hours of audio can fit on one 700MB CD-R. If we were still stuck with v1.0 RC3 I'd say -q 7.

25 people waiting for GT3?

Maybe I should set up a Paypal donation fund

And an Amazon wishlist...

http://www.amazon.com/exec/obidos/tg/detai...=books&n=507846



Ruairi

I already DO archive @ mostly q6! (lower for some, q7 for favorites) And I'm quite happy with it.

More tuning and tweaking by really smart people will be great.

However, only ONE THING can fully complete my Ogg Vorbis experience: *bitrate peeling*!

Why? Because I, being the type of person who enjoys reading hydrogenaudio.org, compress at a higher bitrate than I really need for casual listening. However, my friends, being NORMAL PEOPLE, have frightfully low artifact perception limits (as in, WMA @ 64 sounds ok to them).

Thus, I want to be able to take songs from my q6 - q7 collection, strip 'em down to q0 - q1, and give 'em to my low-bitrate lovin' friends (or cram 100 'em onto my iRiver device for when I'm out and about).

Just my $0.02

i use -q7 because i find it almost transparent, i can still her imporvement over -q6 which is good as well, but not enough for archiving quality.

Anyway, current aacenc at extreme quality sounds perfectly transparent, just a tad better than -q7 but good enough to make fail all abx tests. Im currently archiving my music in ape until AAC gets good hardware support or vorbis improves -q6 quality, then ill transcode and save some gbs

I really would donate!

Is there a general Xiph.org Paypal donation fund?

dev0

You could always use WMA 64kbps = CD Quality to archive...

I would probably donate to. Go Garf!!

As for Xiph, go here: http://www.xiph.org/ogg/vorbis/donate.html
You can buy Xiph merchandise too. Yummy.

Not really.

I am not 100% sure about this. Garf could eleborate not so much the transient response, but never the less physoacoustics masking curves over a set of impluse/padding short blocks that need be adjusted based on a set of samples. Last time I checked at least Vorbis allocated enough PCM samples into a buffer and uses a 10th order IIR Highpass Chevysheb Filter over a predetermined masking threshold in order to determine whether current window should or should not be switched based upon large spectral energy increases over the threshold n samples ahead detected by the filter.



not to insult anyone at microsoft (i'll get sued) or have them take this in the wrong way, but whoever designed that physcoacoustics coder should really go back and retake a course in Digital Signal Processing and or Information Theory.



Nice! I got first dibs of the Xiph T-shirt!



Yeah, and if you really wanted to send them a present (I would if I had the dough) you could bring them a bit of yule-tide cheer this holiday season.

I really don't have too many albums out there, and almost all my audio CDs are original.
As I'm elitist on the audio extracting // coding issue, I use EAC@Secure Mode to extract then encode to MPC --quality 8.5 --xlevel then apply replaygain.

If I have to use mp3, then I encode with --alt-preset insane [always from the CD as source]

As for OGG Vorbis, I really do not use it because it's tuned for being the best at low bitrates (internet streaming), while I don't share nothing on the P2P networks.
If I sometimes get something from there, it's for preview, and if it is possible I buy the CD.

But If you're using Vorbis, archival quality should stay away from the lowest bitrates. So, -q 6 is OK.

Too bad for the music labels that most of the music listeners are not like you...

I agree with sYeLtH

I voted Ogg vorbis @ q8, but I don't use it...
I fully agree with sYeLtH, that at high bitrates other formats are better suited than Vorbis.
For real archiving quality in Vorbis I would rather think of 300-350kbps...

But I voted for q8, since spending more bits is such a waste atm (Vorbis could be a lot better at high bitrates... but atm I think Vorbis's use is still in the lower bitrate range, the range it was originally designed for btw). MPC can achieve these results too at lower bitrates (200-250kbps)

Let's hope this will change (maybe Garf has some time left? )