I'm nerved, because 90% of MPCs I've downloaded so far are encoded with qaulity 6, but quality 5 is already 100% transparent.

Already discussed before. Any argument claiming waste of bits is unfounded unless you're using a very inefficient algorithm. That's why some people use lossless and others use nonperceptual codecs at quite high bitrates. Every bit is made use of, and we assume people decide based on their needs on HA. That's why it's none of my or your business if someone else prefers q10 with their archive (hence the title of this thread is annoying).

• Some of the samples have been observed to be non transparent at q5 and q6 recently with MPC. So q5 might not be as safe as one would want. Check this link.
• One might require quality higher than transparency levels if you use DSP or plan doing post-processing / transcoding (check this link). Transparency is usually a subjective term (unless you mean transparent to every human being in the history of humankind and the future). ABX testing is for personal results and do not reflect to everyone. That's not the level of certainty I would like to have with my lossy archive. Thus I use q8 --ms 15, 70kbps difference with standard is negligible for me.

Blah! People should use the quality setting that is transparent to THEIR ears.

But that's practically impossible. One track might be transparent to you with -q x today but you might realize some difference later on. You cannot meticulously listen to all your music for possible artifacts before choosing a quality. You might even decide to have a robotic ear transplant in 30 years from now that will improve your hearing to 25kHz. That's why the argument which tries to generalize everyone's needs (which goes like "everyone should ...") is absurd. Everyone could do what they want with their music, and I respect it if it's based on evidence and reasoning. It annoys me when someone comes and says "everyone should use q5". I don't think even Frank Klemm or Andre Buschmann is in the position to determine the needs for everyone using MPC. That's why they included those quality flags, for people to have a choice.

What people should do is weigh the quality and space that a codec provides. In terms of quality and space, nothing has been able to beat -q 5.

If you have the space to spare and you don't have confidence that -q 5 is right for your needs (or you were able to find problems with -q 5) , there is nothing wrong with using -q 6.

The people using -insane and -braindead are in theory probably crazy, but if they think they need that kind of "quality" then I have nothing against them. It is their music and if given the facts they believe those presets are what they should use then I have no problems.

Hehe, that's exactly why some people use lossless and others are perfectly satisfied with whatever is transparent to them

I would say, be thankful you are able to download music that is 'beyond transparent' rather than the filth that's infesting KaZaA and other networks

Sorry, I can't figure out what to answer a question like this.

Btw: What do you think of apples?
(I) Apples no doubt is the best fruit in the world
(II) Apples tastes like #¤%#@

upnorth this one is easier:
How tall is Mt Everest
1. True
2. False

Answer:
Who cares? (except climbers)

Because bandwidth is finite, particularly when uploading from cable modem and DSL-connected home PCs, I think there is a strong case for making -q 5 the standard (as it's profile is called) for album transfers. Could a higher quality setting be more transparent? Yes, but we all know it involves trade-offs. If you are using lossy codecs for size convenience go for the best bang for the buck. -q 5 is the most compelling choice if you are going the lossy route.

Of course, I don't think anybody has room for complaint if you are transferring copyrighted material. Beggers can't be choosers.

That´s wrong. The difference between --q5 and --q8 is (on average) around 90 kbps (~ 50 % more compared to --standard). And for sure it´s even bigger with this "--ms15"-tweak. --q5 had around 176 kbps, while --q8 had around 265 kbps on average. For further details see my test-statistics HERE I made last december.

And I agree completely with mithrandir: The problem meanwhile is not the harddisk-space of every user. Nearly everyone can run in the stores and buy a 160 GB HD for around 120,- €. But at this argumentation-chain the "narrowest point" is always forgotten: The internet! No one of these "extreme-encoders" has a DSL-line with 1 or 2 MBit upstream (most of them have only 128 kbit), to trade effectively. So, as long as the internet is so slow, yes, it IS important, in which quality-level everyone encodes. And for sure it´s worth to be discussed.

BTW: When is V1.15r alpha declared as V1.16 beta ?! - I think it would be time to do it. If this would be done, we all could wait for an even longer time for SV8.

i use -q6 for the warm fuzzy feeling. plus i have noticed that the file sizes are smaller than --alt-preset standard. for my type of music, anyway.

Which is also the reason I don't like these guides that advocate some arbitrary quality level.

for me q5 is perfect. never found anything that is not transparent.

Answer: Because it's there.
-----
And BTW...I'm not voting because, IMO, the concept of the question in the thread title is flawed (as many here have implied as well).

I agree with everyone who says that transparency is subjective. I'm a newbie, but that was one of the first concepts here that I learned. Asking "What quality level is transparent to people" is the same as asking "What underwear is the most comfortable to people". 100 different people will give you at least 10 different answers.

The closest I could envision to proper advise I guess would be to use an ABX tool to test samples of several of your songs and determine what is transparent at least 99% of the time. Then add one (1) quality number to it as a "safety margin". If -quality 5 is generally transparent to your ears, then encode your stuff at -quality 6. The bitrate increase will usually be marginal, but worth the extra "insurance" in my opinion. But again, even this approach is subjective. It's how I would do it, but not how everyone should do it. Each person has to decide for themselves.

Most recently, I have adopted the philosophy that atici pointed out about transparency not being a static entity for a person, especially over time. So, I encoded all of my music (that I could fit onto my HDD) in FLAC. I'm a fairly...OK, a *severely*...neurotic person, and I particularly enjoy the extra warm, fuzzy feeling I get from listening to lossless, even though I have *never* been able to ABX high-bitrate lossy that would save me a lot of HDD space. I don't even pretend to think FLAC sounds better to me than high-lossy. For me, it's kind of the same thing as tweaking my car to go 170mph. I'll never experience that speed, but just knowing that it's "there" means something to me, however illogical that may seem.

So use ABX and pick your own quality setting. To each his own...

That's not at all the concensus reached in the recent discussions on this subject. In fact, I'm sure the majority of HA users will still tell you that going much higher than --quality 5 or 6 is a waste of bits, simply because it serves no practical purpose.


You very conveniently left out the fact that in many of these problem cases, using a higher quality index still doesn't yield transparency.


I guess you should seriously re-think your reasoning here. What gives you the idea that the level of certainty is any higher using your braindead setting? It gives you no guarantee that all tunes you'll ever encode will come out transparent, far from it. The risk of a sample not being encoded transparent always exists when dealing with lossy audio compression, no matter how high the bitrate (didn't guruboolez once post a sample that even killed --quality 10?). If you can't put up with that, going lossless is the only real solution to the problem.


Well... in the thread you linked there, you said the very same thing, but I don't see anyone actually confirming that reasoning.

Someone (was it Gecko?) once posted ABX results comparing MP3 files transcoded from different sources (original WAV file, MPC --quality 5 and 6, perhaps 7 also, IIRC), and the outcome was that the file transcoded from --quality 6 did sound better than the one made from the --quality 5 file, but that there was no ABXable difference between the file made from --quality 6 and the one encoded straight from the original WAV file (someone please correct me if I'm wrong... I'm trying hard to remember, and I'm too lazy to search right now ). So we can agree that when it comes to transcoding, a little overhead can't hurt, but this still doesn't at all prove the necessity of insane settings like you are suggesting.

There's also no evidence that proves your point about DSP processing, it's purely theoretical speculation - even though it does make sense, there's nothing to suggest that DSP processing can reveal flaws in --quality 5 files that would not be present in --quality 7 files. Why don't you encode some samples with various quality settings (perhaps 5 and 7), decode them, pass them through some DSPs, and then try to ABX the difference between them? That would help everybody a lot more than these unconfirmed assumptions of yours.

I couldn't care less what the majority would say because noone can reasonably argue in favor of "waste of bits" without providing impartial evidence (partial evidence=their needs generalized to everyone else's). Quality 5 is at least as arbitrary a selection as Quality 10.



That's simply not true! Check guruboolez's or Pio2001's post in this thread:





The encoder noise is lower which is a verifiable fact and the basic assumption in lossy encoding process! But there's no certainty as to whether or not with any quality setting you'll be safe with lossy encoding (as you also pointed out) when it comes to audible artifacts. But the amount of compromise I'll make is up to me to decide. I can use lossless or not, I don't have to give an account to the rest of the HA community here as long as I believe my argument makes sense.



Check user's post in this thread dated Jun 30 2003, 04:27 AM. I don't understand what you want me to prove. Because all I am saying is as you increase the bitrate the resulting file approximates the original better, which is a verifiable fact and the fundamental assumption in lossy encoding. As soon as you admit this information theoretical fact everything I say is a simple consequence. And how much loss you'll tolerate is up to you to decide (for yourself not for me).



I am not suggesting anything! You do. What makes you think I encourage everyone to use my settings?



Oh yeah! Theory is pure speculation anyway And we should forget about the entire science of mathematics too, right? Oh omniscient Volcano the great...

Edit : Fixed link

Untrue. There are various samples that fail even at q7 and above.

http://www.hydrogenaudio.org/forums/index.php?showtopic=4084
http://www.hydrogenaudio.org/forums/index.php?showtopic=5060
http://www.hydrogenaudio.org/forums/index.php?showtopic=4697

No one is saying that mpc doesn't get better at higher bitrates. (Note that EAQual thinks different.) Artifacts are milder and people who push the Incredible 3D Surround ™ button will get a more enjoyable experience. But under regular listening conditions once a sample is transparent it doesn't get any more transparent. And fact of the matter is that for the broad majority and even the majority of critical listeners, mpc q5 is transparent on the broad majority of samples. The way people hear isn't all that different from one person to another and the amount of quality mpc provides is nearly constant for every sample.

So for the people starting out with mpc, q5 is a very good choice, but these people will also read your post and/or the * guide. You are an authority here and people will take what you say for granted. You give them the impression that because you are so expert you choose exactly this setting and they will adopt it. The same problem applies to the * guide. It's an authority that people look up to and they will conclude that anything below q7 is not up to the job, which is simply false.

Q5 is NOT arbitrary. It's exactly the point where almost all artifacts disappear. If you go below q5 you will surely find lots of problem samples. But if you go above q5 you will not notice a difference on the very broad majority of samples.

You can use whatever setting you wish, I don't care, but when you post it and talk about it you should put up some sort of disclaimer. Saying that everyone should use what he feels right is not enough. The standard newbie will not run any extensive tests. People will tell him that q5 is fine but then he stumbles across your post and others like it talking about loosing less information and more or less claiming q7 as absolutely transparent (as suggested by the posts you linked to) and he will think: "boy, that guy sure knows what he's doing. I don't understand but it sounds good... I'll just use his setting then." This is real. The target audience of this board includes these newcomers and you should allways keep them in mind when posting.

You haven't given any good reason for using your command line other than that you are throwing away less information and are thus closer to the original. You have not proven that it is in any way neccessary to preserve this particular amount of information. Instead you are confusing the newbie. Like I said, you can use whatever you want, but instead of reasoning about it you should admit that it's just for warm fuzzy feeling that you have lost less information.


MOD: * no links or names to ripping group guides please.

Not at all. --standard is tuned (AND AIMED) to be transparent on all (ok, most) tracks. Quality levels above just leave some information, that can be thrown away (because psymodel MAY give wrong results). These levels have their names (insane, braindead) for reason.

-Eugene

Very well said. Those are my sentiments exactly, I agree 100%.



Is this so hard to understand? In all threads you have linked so far, all I see is people simply claiming that they need extra headroom for further DSP processing on the decoded material. What I don't see is anybody actually proving that this extra headroom is necessary, i.e., proving that using (for example) --quality 7 over --quality 5 makes an audible difference when files created with those different settings are passed through a DSP. Should that not be the case, one would get nothing in return for using a higher bitrate, and your claims would be meaningless.


Of course. But you also ought to consider the practical side of all this, which involves extensive ABX testing to prove that the user does get something in return for using such insane bitrates (be it better results when transcoding to another format, better results when applying DSP processing, etc.). Someone (I'm still not quite sure who it was ) has already done the transcoding bit, and proved - at least for the samples he tested - that using anything higher than --quality 6 is overkill because it yields no practical advantage, now it's your turn to prove your point about DSP processing. Encode some samples and process them, fire up ABC/HR or WinABX, and test away!


Actually, for issues like these, yes, you *should* forget about the entire science of mathematics. It doesn't matter whether setting X provides a more accurate representation of the original signal than setting Y - what matters is whether or not this difference is audible (i.e., whether the two files are ABXable from each other or from the original, not that the "difference files" sound different - jeez, ever since you had this idea of "listening to the difference file", the meaning of "audible difference" has to be made extra clear ), or whether setting X *really* provides audible advantages which are revealed after transcoding (which has been taken care of already) or DSP processing. And this you have yet to prove.


Are you running out of arguments, having to resort to this kind of childish crap?

I don't understand the reason for all this confusion. All I was saying is let people have the choice to use what they want. Because as well as there're samples for which higher quality settings do not help, there're others (and even more of them based on guruboolez's quote) which higher quality settings help. Moreover, the masking effect of lossy encoders do not assume post processing and therefore if one uses DSP bitrates that normally provide transparency might not be sufficient anymore. Because the encoder noise could not be masked anymore and some problems might surface. Therefore all quality levels are as legitimate to use as quality 5. I don't need to prove anything further than this: "If you go higher in the quality scale, the result is going to be a better approximation of the original" which is a clear and obvious fact. I am not suggesting people to use any specified setting, on the contrary I encourage them to pick their own based on their needs. If people tend to pick my settings because my post number is high, that's really their misinterpretation.



That is entirely your personal viewpoint. Why are trying to impose it on everyone else? How do you define "audible"? That's a pure subjective term. Check the post here dated Jul 30 2003, 07:18 AM. What if someone in 25th century beams you a bionic implant which gives you the most amazing hearing on the planet? Then everything will all of a sudden become audible. However using higher quality will still get closer to the original, because it is an objective fact.



You always get something in return for using high bitrate and that's exactly the point. You need no tests for that purpose because that's how the lossy algorithms work (a mathematical truth just like 2+2=4). The difference might not be immediately audible to you or to me, but could be audible to someone else as well as observable for people who do transcoding (den already tested this once you might search for the test results, so I do not need to prove anything about this either) . That's why nonperceptual codec WavPack which preserved some inaudible data did a better job although at much higher bitrates. High bitrate always brings additional quality, if you deny this you are basically denying all the information theory facts about lossy encoding. All you say is difference is not audible, and therefore subjectively overvalue audible data (you don't even define audible and I doubt you can but I guess you mean audible to most of the population) over inaudible in case of no post-processing. You go on with that attitude when you demand I do listening tests. My initial point was: there could be no objective argument favoring any quality setting over another. Let people have choice. Is that so hard to understand?

This makes no sense to me. Is anyone stopping (you or) anybody from having the choice to use what they want?

When somebody says "is X what people should use?" then AFAIC they mean "is X good enough for the majority, as a basic standard?" As with Lame --alt-preset standard, my answer would be "yes, that's what people should use."

Now that's a question about english language and the meaning of "should". When you say "should" I understand it as demanding everyone to use that setting. That's what many other people also thought and commented similarly before in this thread.

Wait a minute... if you re-read my post, you'll notice that I did actually consider the issue of post-processing, and that I consider leaving a little headroom perfectly OK, should it be *necessary* to achieve transparent results (which is what I'd like you to prove by means of an ABX test):




Anyway. If I were to reply to your post in full length, I'd basically have to repeat all I have said previously (just like you did), so I'd better not. This discussion is leading absolutely nowhere, I'm outta here.

i usually stick to q6, and q7 for favourite songs for the added satisfaction that its of higher quality!

My long post will be a small picture.

Vertically, the perceptual transparency
Horizontally, the --quality setting used
In blue, the result acheived

I must agree with atici. From what I've read from people that do extensive testing like guruboolez, -q 7 will solve most problems at -q 5. And if not, will make the artifacts more difficult to hear. I also support the freedom to choose whatever setting you want.

I have an upstream of 1MBit on DSL, so !

edit: I just changed mbit to MBit so it would look nicer.

The curve would probably be a little more rounded than that. (sorry to be nitpicky)

I've just post another sample, ABXed with success at --insane setting... Nevertheless, problem at --standard is easier to hear than with insane.

http://www.hydrogenaudio.org/forums/index....t=0#entry121408

And that makes all the difference; considering the time it would take to ABX every track on every CD I have. --braindead (a very nice comprimize between --standard and lossless) works for me and my 120GB HDD.

This is more like what I have in mind:

Much better atici! Ah! the use of a parabola to express the law of diminishing returns as it pretains to audio compression. I love it!

For me, MPC q 4 is near transparent (4.88 out of 5.0, averaged over Roberto's 12-sample test). This is compared with lame --alt-preset 128, which I rated 3.35 over the same 12 samples.

I certainly would be happy to encode and use this setting for my personal use.

ff123

I think the question is flawed

No, why should I? So would that be:

But it also depends on what you want to with the music, just listen to while doing something else (much like a radio) --quality 5 is very adequate and bitrate friendly.
For archiving a precious CD I would use a higher setting, just to be sure. But I don't want to force that on other people. If I had a bigger HD I would consider lossless for that.

Indeed, it's like: Do you prefer apples or the Mount Everest
--
Ge Someone

is there way to scale the bitrate range like... from lets say 128-256kbps? cause from the "MPC recommended settings" thread i get that each preset has a range of 40-60kbps, like for --quality 6 its "168 ... 212"

@atici: Are you sure you didn't miss this part?
It's not that strange that you get another graph when you change the meaning of the axis.

The weird thing is, the rational and irrational come together nicely for most of us in this issue.

We all know that q5 is more than good enough almost all of the time, and much better (and consistant) than we can do with most other formats.

But we all know that there are a few samples where q5 is not transparent for some people. And some of us think that maybe one day we'll use some kind of processor to bump up the stereo signal to 5 or 7 or 10 channels. Or maybe we'll need to EQ a track. Or maybe we'll need to transcode. So we use q6, or q7, or q8, or q9 or whatever, just hoping that going "just that little bit better than standard" will save our ass in the future.

Then I think: No - don't do that David - there's absolutely no proof that using a higher quality is actually going to help me in this or that situation. Chances are, it will help, but it won't make things as good as lossless. In any of these hypothetical situations, I just won't use my MPCs!

So, what I'll do is keep my CDs safe as a lossless back up, and encode my mpcs at -q5 for use now. That's the sensible, rational thing to do. I'll just enjoy the music.

And then I think "maybe I'll use q6 - just for a little extra bit of security?"

ARGHAGRHAGRHAGRHAGHRGARH!!!! ;-)


Cheers,
David.

*LOL* That seems to be the critical mass that this type of topic races toward.

I found my own solution. I've made a hobby out of re-encoding my music. I re-encode a large portion of my collection (~400 CDs) once every two weeks. I'm on my fifth cycle now since about May. I won't quite call it an obsession, but the novel I was reading two months ago has been gathering dust on my nightstand, I can't remember my girlfriend's name, my hair's pretty long now, and what's that smell?!? It's because you people here at HA just confuse the hell outta me, and I can't seem to settle on a codec... format... bitrate... lossy... lossless... transparent... psymodel... lowpass... compression rate... file size... quality setting... VBR... CBR... ABR... ABX... bah! (Just kidding...I really like it here!)

Seriously, here's my recent history...
MP3 (FhG) 128cbr, then
MP3 (FhG) 192cbr, then
MP3 (FhG) 256cbr, then
MP3 (LAME) --alt-preset insane, then
FLAC,
...and right now I've just started re-encoding all of it into Ogg Vorbis Post 1.0 CVS -q 3.

See, I don't have to worry about whether my hearing may one day improve or degrade, or what kind of audio equipment I'll end up with, or what my needs will be in a year, or in six months, or in one month...
...because the weekend after next it's all changing anyway...to MPC -q 4.37, baby!

Another way of seeing it would be to say: "Hey, I could use --quality 7 and still get a lower bitrate than I would with lame --alt-preset standard (actually ture for my music), so why not?"

In fact I keep lossless backups of my CDs and encode them to whatever I need them using foo_clienc/foo_mpcenc/foo_vorbisenc.

dev0

Ah, but you simply MUST try MPC -q 4.38! That .01 of quality makes all the difference!

Seriously though, I find q5 to sound very good. If I was feeling neurotic about getting every last bit of transparency I would switch to lossless. What works for your equipment and ears might be different. q6? q7? Use whatever you feel you need to gain transparency with your music...



ff123, I haven't heard many others say that q4 was good enough. Interesting. I haven't tried ABXing it, but I suspect that for many 'average' (non-trained) people it would indeed be good enough, especially since people new to audio encoding are used to poor quality 128kbps cbr mp3s...



Very good point, if the concern is file size compared to MP3, then indeed MPC q7 is a great sounding alternative that will give that warm fuzzy feeling of using a bitrate higher than standard. (I still like q5 though, hehe.)

It's a different graph than Pio2001's. My y axis is not perceptual quality (but objective approximation quality?) and x axis is not MPC q setting (but bitrate). There's no formula for the curve that I can write down and I do not intend to prove this because I know it's vague It's just how I interpret the way lossy codecs work. If a codec manages to preserve audio quality at low bitrates (like MPC) it's bound to miss some other information (though inaudible or masked) because of the regularly high information content (around 700 kbps) of average samples. That's why I intend to set my safety region based on bitrates (until I can afford and conveniently store lossless archive).

Atici: How do you "set [your] safety region based on bitrates" (using MPC)? That seems vague. I mean, once you go beyond transparency (at your current level, before bionic implants, of course), how do you judge what bitrate is adequate? e.g. Would you always maintain over 500 kbps just because 350 "just doesn't seem like enough"? Yes, 500 will be somewhat closer to your original 700 kbps sample in terms of pure data, but since it won't sound any different, the only way to justify it is with some vague notion of using a huge bitrate to drown the psymodel's few flaws or else the good old Warm Fuzzy Feeling™.

Oh, and to answer the question and stay On Topic, I think yes, the general populace should definitely use MPC --quality 5. Since many people continue to use MP3 between 128 and 192 kbps, this would be a huge improvement. However, going beyond that would be unreasonable right now, since q5 offers the best quality/file size ratio for people with limited bandwidth and storage. Think of it this way: You're welcome to spend your money on a private school to obtain the education you desire, but that doesn't mean we shouldn't try to raise the standard for public education. However, you can't expect the public to go from 128 MP3 to MPC q8 or lossless in a snap, just as you can't expect public schools to immediately perform at ivy leage level. A user will download a song to their laptop equipped with poor speakers, think "gee, this sounds the same as all my other files, why is it 5 times larger?", and go back to MP3. So, I'd be ecstatic if more people could make the leap and realize the quality benefits of MPC q5.

Atici, as many other people, had space enough for braindead encoding, but not for lossless. He just took an insurance. If you want to pay more a complete insurance, covering some improbable risks, where is the problem ? --standard is more economic. On the other side, it's less sure. I sometimes found and ABX slight problems with Q5, corrected at higher setting. I used mpc for a very good level of transparency, and complete this choice with an adequate setting. If this amount of quality wasn't a concern, I probably stayed with --preset standard (not exactly true : there was the gap issue).
I can afford some waste in space, if some problems are corrected, even if some others remains.

(sorry for answering for atici, but I'm an insane mpc user, and for some reasons. I spent some time in ABX test, and I'm mature enough for chosing myself my own setting).

Nice post, actually. It reminds me of headphones, and how something like the Sennheiser HD580 or HD600 achieves something equivalent to Q5 with MPC... yet some people do spend $4000 on headphones like the Sony MDR-R1O (R10?) to get the additional 5% closer to "perfection." In the case of audio codecs it's just bits, so no crime in it if people want to use "braindead" bitrates.

Gee, and free shipping too ;-)

I'm fed up with such discussions - I'm using FLAC and nothing will change it...
(maybe a new, fast (Pentium 100 for normal background playback),
more efficient (like 10%) lossless codec)

Ah, yes, insurance. I understand his situation and apologize if I seemed to be saying anything above Q5 is ridiculous, because I know it's not for everyone. What I really meant to be asking is: Why is a certain level of insurance acceptable where another is not? I suppose that goes well beyond the realm of audio encoding. However, I find it interesting in this case, since you guys represent the borderline between die-hard lossless users and hard-set lossy preset abiders, choosing instead to max out the lossy options. I was just mostly curious why you aren't comfortable choosing one end or the other

I personally fall into the camp that uses q6 "just to be a fuzz better than q5" (yes, roll your eyes), so it's not like I'm completely decisive on this.

From what you guys have said, space is still a slight factor, since the difference between braindead and lossless can be a megabyte or two per track. If a lossless encoder could compress, say, another 15% further, would you choose that over MPC? I think a lot of people would. So I wonder if another format will eventually be created to fill this lossy-lossless gap and satisfy the need for "lossy with insurance", or if perhaps the switches in MPC will change in the long run to have "--insured" or somesuch to represent q8 and beyond, to the point where someone can force their MPCs to be nearly (bit-wise) indistiguishable from lossless. (sorry for the long sentences, it's after 3am)

Not sure. For most people, lossless compress at 50-55% (700-800 kbps).
As a classical listener, most of my discs are near 40% (< 600 kbps). Here are the additional 15%. Nevertheless, there's a big gap between mpc at 240 kbps and lossless file at 550-600 kbps. The only lossless discs on my two HD are temporary files, waiting for burning.
I'm not waiting for algorithm improvments (they will probably never come), but for very high storage capacity, at low price. In two years, lossy encoder for archiving purpose/high quality jukebox will be less interesting. For the moment, and in my case (> 1000 discs), I need bitrate < 300 kbps.

Here you can see a sine which runs in 20 seconds from 100Hz to 22050Hz.
The point is, that the mpc sounds even with quality 5 transparent, but even with quality 10 it's everything else than "scientific usable" or as good as lossless.

English please... So what are you trying to say? Use what you like is the general consensus in this thread. If you have a good argument against that, which I doubt anyone can impart anything further, please go ahead.