I thought I would try some encodings with OGG but there are no stickies with recommended compiles or settings like there are for MP3. I briefly read through some posts and I downloaded the GT3b1 version from Rarewares as it appears to be recommended for settings -q5 and above.

I briefly did some tests using foobars ABX plugin to test some encoded files against the original WAV. My problem is that I was able to easily ABX all the way up to the highest encodings I did at -q7.5. Is this normal? Are my switches possibly set incorrectly? Are there other swithces that I need to add? The only option I was using at the time was the -q switch.

On other codecs I have a more difficult time ABXing over 128 kbps so it seems unlikely that I should be able to ABX OGG so easily at even higher bitrates.

Check ReplayGain in foobar2000...if it's turned on in your Vorbis decoder (and assuming you may have ReplayGain info in your Vorbis-encoded files), then there may be a volume differential between the Vorbis files and WAVs at playback.

If you have the hearing abilities of someone like guruboolez, then you may be able to ABX it up to that point even if the playing field is level. But that would be *very* unlikely IMO. What kind of samples are you using? Problem samples, or "normal" music? Make sure EQ and all other DSP effects are not active.

To my ears, and for most of my music, Post 1.0 CVS with -q 4.25 is transparent. Then again, I'm not well artifact-trained. If you are, your threshold for transparency would be higher more than likely, but that seems extreme. And I'd only recommend using the quality setting...no other switches for the Vorbis encoder.

If you can ABX up to -q 7.5 (and you're not Superman), there *has* to be something turned on that shouldn't be. Check all the player settings again.

Edit: Also, It would help if you could describe the differences you're hearing, to determine if you're "hearing Vorbis" or hearing something else, like a setting in the player.

There are some samples where OGG can be ABXed up to very high quality settings. To my ears it often manifests itself in gritty vocals and hf noise is somehow changed. The "air" is different.

See here.

As a rough guide, take the approximate bitrate you've been using for MP3 and translate it to the nominal bitrate for a particular Vorbis -q setting. Here is how the CVS bitrate/-q values break down:

-q -1 should average roughly 48 kbps
-q 0 should average roughly 64 kbps
-q 1 should average roughly 80 kbps
-q 2 should average roughly 96 kbps
-q 3 should average roughly 112 kbps
-q 4 should average roughly 128 kbps
-q 5 should average roughly 160 kbps
-q 6 should average roughly 192 kbps
-q 7 should average roughly 224 kbps
-q 8 should average roughly 256 kbps
-q 9 should average roughly 320 kbps
-q 10 should average roughly 498 kbps

... so if you've been satisfied with MP3s encoded at/around 128 kbps, you would probably do well to start with -q 4.

- M.

@ ScorLibran - Both files were replaygained to the same level (foobar gives you a warning if they are not). What I was hearing were clearly artifacts on the vocals that could be heard with or without replaygain. This was on regular music, not "problem samples", the album in question was Creed - "Weathered". All EQ settings were turned off and I had previously been ABXing Lame APS and Quicktime 128 with much difficulty. When it came to ABXing OGG -q5 through -q7.5 it only took a few seconds with 100% accuracy each time (though it was getting a little more difficult at -q7.5), always with the telltale vocal artifacting.

Again I am unfimilair with anything Vorbis and I was using the GT3b1 compile on rarewares. Are there different decoder plugins? Recommended switches? I will try the 1.0 CVS as well as some other players and see if I can't figure out if I do indeed have "Superman Hearing" or more likely that I am doing something wrong.

Is there a general consensus on what -q level is transparent in most cases (similar to -APS for lame)?

@M - My OGG encoded files correspond very closely to yours.

Actually, if he's satisfied with MP3s @ 128kbps, I would suggest starting around -q 2, and go up from there if noticeable.

[rant]It's not Oh-Gee-Gee!!![/rant]

Damn you must have good ears. Usually, I recommend -q 6 since that's when channel coupling becomes lossless.

But it's people who have ears like you that keep Vorbis from really having a "standard" setting.

Was that reply made towards me? If so, could you please elaborate?

Vorbis sounds transparent with quality setting higher than 6 so I guess you should use 6 as your base point.

Laters

AgentMil

Yes, it was. Though I didn't mean to offend you. Sorry if I did. That was more a comment of jealousy, but I forgot my emoticon

"Standard" settings usually are set when 99.9% of the users find them to be indistinguishable from the originals on most of the samples. However, both Garf, Monty, and the Xiph team have said that you have to find out the best setting for yourself. I would say that this is because of people like you, who can actually hear the difference above -q 6, when I'd say 80-90% of listeners cannot (I can't hear a difference at -q 3 for most of my personal music).

No, you didn't offend me... I was just confused. I recommended -q2 as a comparable quality to 128kbps MP3. But M suggested that bfourney used the same quality setting that hovered around 128kbps that would give better quality at the same bitrate. I recommended a smaller bitrate; I'm not sure where you picked up that I've good hearing (not that I don't).

I haven't really tested Vorbis deeply, but on most of my music q4 or q5 is good enough. However, I did do extensive testing of MusePack (MPC), and 6.5 is where it sounded lossless to me, not 5 like many people say.

Maybe I should consider reading the thread at some point, so that I don't make such huge cockups next time.

I'm sorry, my mind mixed up pseudoacoustic for bftourney. bftourney, with his excellent hearing, breaks any chance of a standardization happening in Ogg Vorbis. I had always thought that -q 6 was transparent for just about everyone, but he comes in here and has ABXed all the way to 7.5.

pseudoacoustic on the other hand.... I think you're the first person I've encountered who found Vorbis reached transparency at a lower q level than Musepack.