Just need to clarify a few points.
I have been using Lame alt-preset-extreme to encode my rips,being new to .ogg was wondering what would be a comparable setting (say,for file size) that would yiels similar or hopefully better quality sounding files.
And is .ogg format of such a better quality than lame Mp3?
Basically what I'm after is a simple answer to this question,what setting can I use for .ogg that will represent a similar file size to lame mp3's,and will it result in a better quality file for the same size.
I know I should conduct my own tests but time is a bit limited at the moment so thanks to any who can help me out.

ps,using EAC

Ok simple answer :

q7 - q8 will give you the same or better quality than -a-p e
q5 - q6 will give you the same or better quality than -a-p s

These settings are not tested by me, but these are some of the values I remembered from reading numerous threads about ogg vs lame

I guess you should test it for yourself, as this will give you the best proof of what suits your needs.

Probably -q6, maybe you can even get away with -q5.

Quality will be better than MP3...

--
GCP

What about the test samples we had, where -q 6 was needed for Ogg to have a quality comparable to LAME aps or MPC --standard? Are those issues fixed now?

It's always possible the codec the fails on individual samples. He asked for a general recommendation.

There will be samples where -q5 and -q6 outperform LAME aps or LAME ape as well.

--
GCP

Thanks guys,much appreciated,I have done a few tests now but my listening setup (everyday setup that is) is not of an audiophile quality so major differencies I cant pick up.
I'm happy enough with Lame a-p-x at the moment but just thinking of trying differing formats as they develop.

what about the magical number 4.99
Is that close enough to aps?

When -q 5 barely is, why should 4.99 be? Nope. But again (nod to Garf), depending on how you test it.

4.99 will get you lossy coupling. I haven't heard about any issues with it lately, but if you want to be on the safe side...

--
GCP

No major issues with lossy channel coupling... but it wrecks the stereo imaging.

If you have a quality system I would recommend (to be safe since that's the issue) at least -q 6.

edit: typos

That would be a major issue...it should be inaudible, after all.

--
GCP

That brings us to the ABX issue but I don't wanna rant about that no more.

How is lossy coupling wrecks the stereo imaging?

I thought aps uses joint stereo, by your argument, that wrecks the stereo imaging also?

SAFE joint stereo.

I guess that's why the switch is called -nssafejoint (Naoki Shibata's Safe Joint Stereo)

Regards;

Roberto.

I am just a newbie so I am not flamming and please point it out if I get it wrong.

I thought dibrom didn't implement --nssafejoint in aps in order to keep file size down, or is it implemented in the other preset before he came up with the standard?

So
apx uses --nssafejoint
aps only uses --nspsytune
to my understanding those two are two different things, correct?

I am not familiar with lossy channel coupling or any type of Stereo Channel Coupling in the Vorbis at all. The doc on the xiph.org is outdated as someone has pointed out before in the forum.

Plus that didn't answer the total wrecks the stereo imaging question either. I just want to know how it wrecks the stereo image.

And does this mean that it is not possible to obtain the surround phase info if it is downmixed from a 5.1 source using lossy channel coupling?

Joint Stereo is lossless coupling (*).

(*) It can lose a single bit due to roundoff errors, but that's completely inaudible.

--
GCP

??? MP3's joint stereo (mid/side - l/r) is not lossless. Are we talking about MP3 or Vorbis here?

kxy: A bit off topic for Vorbis forum but, --nssafejoint is just an alias for --nsmsfix 1 which uses the kind of threshold value which doesn't give mid/side frames as often as the default threshold (--nsmsfix 3). All VBR alt-presets use tweaked nsmsfix value (not the default 3).

The MP3 L/R -> M/S transformation is lossless except for a single bit due to rounding. It's what the encoder does with the data afterwards that isn't.
i.e. The normal errors/artifacts when encoding the new channels are what could potentially harm the stereo image.

Vorbis lossless is just M/S without the rounding error.

The channel coupling itself will only harm the stereo image when intensity stereo/respectively lossy coupling is used. But just Joint Stereo/MS will not destroy the stereo image.

--
GCP

Eh, but isn't that what the encoder does with the data kind of the point?...
At least with mp3 there's slightly different psychoacoustics with m/s so the end result if you would encode with vbr (lets say hypotetically totally flexible) would differ more than 1 bit.
Also the quantization error with mp3 joint stereo (l/r - m/s) with m/s -frames spreads to both channels. This is not the case with Vorbis' lossless coupling.

You said joint stereo is lossless coupling that may differ only 1 bit rounding error from stereo. This isn't exactly correct. You can't just look at the M = (L+R)/2, S = (L-R)/2 -transform, you gotta take the whole process into account. In audio coding the end result is what counts.

With Vorbis' lossless coupling the end result should be bit to bit identical to stereo coding.

I'm contesting the statement/impression given earlier in this thread that using channel coupling automatically wrecks the stereo image.

That just isn't true. The transformation itself is as good as lossless. The only issue it has is that _if_ the encoding results in audible errors, they will affect the stereo image more. But there's no reason this automatically has to be the case, in fact, there's generally less chance for audible errors *at the same bitrate* when using stereo coupling. Just there being different psychoacoustics also in no way means that the stereo image will automatically be wrecked, in fact, it's probably quite the reverse!



I don't understand at all what your point is here?? Of course there will be *differences*, it's lossy audio coding we're talking about.

As you've stated it now, I'd interpret this as you implying that the M/S coding routines are inferior to the L/R ones, which I don't think is true.

I stand by my point. Joint Stereo is a (practically) lossless transformation. The fact that it _may_ increase the risk of the rest of the encoder introducing an artifact in the stereo image is irrelevant to this - you have the same risk with stereo coding as well.

I'll have to ask in #vorbis about some of the stuff you say about Vorbis lossless coupling - I don't understand how it's possible to take advantage of the redundancy and not spread the quantization errors.

--
GCP

Differences to stereo coding.

No, I didn't say anything about this issue at all. I only gave reasons why you can't say joint stereo-coding is lossless coupling. If you would ask should you use Lame's joint-stereo coding with your mp3s or not, I would definitely say you should use it.

Joint Stereo is not only the transformation. It's a general name for a channel coupling method.
ttp://www.xiph.org/ogg/vorbis/doc/stereo.html
"Using polar mapping and/or channel interleaving, it's possible to couple Vorbis channels losslessly, that is, construct a stereo coupling encoding that both saves space but also decodes bit-identically to dual stereo. OggEnc 1.0 and later offers this mode."

The joint stereo MP3 uses cannot produce bit identical end results compared to stereo-coding even if it was totally flexible vbr. Audible quantization noise will spread to both channels with m/s-frames.
But Vorbis' lossless channel coupling can produce bit identical results to stereo-coding -> thus it's called lossless channel coupling...

The reason why I call Joint Stereo lossless is that the stereo transformation it uses loses no information. That's lossless in the strict sense of the word.

You are interepreting lossless are meaning 'producing the exact same results as if the file were encoded in stereo mode'. This is the Vorbis definition, and it's got no bearing at all to the point here, namely that Joint Stereo would wreck the stereo image.

I am talking about 'will not produce additional artifacts in the stereo image'.

I repeat, the point of this thread is if Joint Stereo automatically wrecks the stereo image. That does *not* follow from the fact that it doesn't produce exactly the same results as stereo mode does, because the encoding process is different.

There is a bigger *risk* of artifacts happening, but with a welltuned encoder, it's
small, and in *no way* does it follow automatically.



I'm referring to M/S mode, since it's the only difference with stereo coding in the first place.



I got the causes partly worked out on IRC. But now you tell me, how does this in any
way imply that joint stereo automatically wrecks the stereo image?

The only thing you can conclude from the above is that the Vorbis lossless stereo coupling
method can't possibly wreck the stereo image. But the reverse doesn't follow.

--
GCP

Well, I haven't implied anywhere that joint-stereo automatically wrecks the stereo image. Infact I don't remember talking about stereo image at all, just that you can't call joint-stereo lossless. But this is what Ivan said about this:

"Quantization is lossy for both L/R and M/S, but the coding errors introduced in M/S frame are slightly different (in psychoacoustic terms) than in L/R - for a human.
M/S coding errors are spread to both channels. When encoder runs out of bits, for example in CBR coding, or low-quality VBR coding, the encoding artifacts in M/S frames are perceived as loss of stereo image or something like that, while L/R artifacts in the same condition would be chirping, wobbling, etc.. but the loss of stereo separation is very rare in good M/S coding"

But it's what the thread was about before you joined in.



By your perverted definition of lossless, no, you can't. I'll call this Vorbis-lossless from now on. You're saying a stereo mode can't be (Vorbis-)lossless unless it produces exactly the same results as if the file were encoded in stereo mode in the first place, and that after all encoding steps have taken place.

IMHO, this definitions is perverted, because it implies stereo coding (not Joint Stereo) itself is lossless. This is not true. They both will have errors, but the errors will be different.

It's like defining a codec lossy if it doesn't produce the exact same results as an MP3 encoding. By this definition, even Monkey's Audio is lossy. That isn't a good definition, IMHO.



Which is exactly what I've been saying for the entire time.

--
GCP

Hmm, I think it's pretty clear that certain FhG m/s -coding implementations do hurt the stereo-image as ivan implied. In quantization process with m/s frames there might be problems which appear as stereo-image problems, especially since MP3 uses frame-based m/s. It's just that Lame has better and correctly working channel coupling. In my opinion it's not totally impossible that there will be slight loss of stereo-image even with Lame with m/s-frames in some very rare situations. Of course you can tweak the thresholds to make this possibility even smaller.

So, I don't think joint-stereo can be said to be absolutely 100% lossless regarding stereo image. But Vorbis' lossless channel coupling is totally lossless.

I agree - I've stated that several times.



No, Vorbis channel coupling is Vorbis-lossless

Just as it's possible for the joint-stereo modes to go wrong occasionally, it's possible for the stereo modes to create stereo image artifacts (but it's much less likely than with joint stereo). Think about what happens when a normal artifact is different in magnitude in each channel - this is perfectly possible in stereo mode or with lossless coupled Vorbis.

Because of this, it makes little sense to say joint stereo isn't lossless and stereo is. And because of it, I defined a lossless transform as 'will not produce additional artifacts in the stereo image' below.

By this definition, *both* stereo and joint stereo *are* lossless. Intensity stereo and lossy channel coupling aren't.

--
GCP

Hmm, anything is possible but I really doubt stereo coding would create "loss of stereo image" ie. otherwise perfect sounding, but just not the feeling of space. Of course with stereo coding there can be artifacts which seem to cause single artifact type of distortion affecting to perceived stereo separation of certain signal/instrument.
But somehow I don't think stereo coding would cause general "loss of stereo image" type of problems. I haven't ever heard that happening.

Anyway, the important thing to remember is that loss of stereo image would be very very rare with Lame's excellent joint-stereo coding. There's no reason to use only stereo-coding with Lame. Quality will be better with joint-stereo.