I have a feeling most of these crappy "MP3 sites" come from Germany - here's another one:

:puke: http://www.franksmp3.de :puke:

LAME isn't even mentioned, Blade is claimed to be the highest quality encoder... and so on.

Another big mail is coming up... And here my usual request to German-speaking users: Please help! :argue:

CU

Dominic

Hrmm.. I need to fix the smilies

Anyway... I really don't understand why people seem to think Blade is the highest quality MP3 encoder. Perhaps it's some of the totally misleading (and false) statements made on the Blade page here:

http://bladeenc.mp3.no/skeleton/quality.html

It seems as though the author tries to play off Blade's weaknesses as some sort of "advantage".. heh.

A few fun excerpts:



He implies that things like masking "depend on the music and bitrate". This is incorrect and misleading. Psychoacoustic effects themselves are a constant, and if you were to acheive optimal efficiency you would always want to exploit them no matter what the bitrate was. In addition, joint stereo does not "cut down the stereo effect" per se.

QUOTE

On the other hand,  most developers have chosen to concentrate on the segment around 128 kBit since that is the most popular bitrate and in the process (knowingly or unknowingly) made design decisions that hamper high bitrate performance. A good example was an earlier version of Fraunhofer's encoders that left joint stereo encoding on at 256 kbps, seriously limiting the quality. Some tests have even hinted that some mp3 encoders might contain bugs that lowers the quality at high bitrates!

So we have vague statements that imply that low bitrate tuning somehow effects high bitrate tuning. This is not necessarily a given at all and there does not have to be any particular tradeoff.

Now for the most hilarious part of all:

QUOTE

Heh. So he tries to say that somehow by allowing pre-echo, Blade has more "tonal purity". That is total BS. There is no excuse for allowing pre-echo and "fixing" it has nothing to do with "tonal purity" either. Simply a totally misleading notion that further supports the idea that somehow Blade is superior in "special ways". For that matter however, I can almost guarantee that Blade suffers far more from dropouts and other tonal distortions than LAME and probably almost all Fhg encoders, so even with that alone the whole "tonal purity" argument is thrown entirely out the window.



Again you have some totally misleading and incorrect statements about quality, mp3, and joint stereo. I guess MPC, Vorbis, and AAC all must only be "mid-fi" since they all emphasize the use of channel coupling in some form or another by default. Right...

QUOTE

Filtering out  the higher frequencies is generally a good idea when dealing with low bitrates and depending on circumstances might also be a good idea at medium bitrates, but is a bad idea when dealing with high bitrates.  Some encoders therefore have different filterlevels for different bitrates.

More falsities... especially when dealing with MP3, disabling the lowpass entirely, even at high bitrates, can lead to worse quality in some cases.. usually do to more bit expenditure on the extreme high frequencies and less elsewhere (even roughly around 16khz). That can increase the chance of ringing and in fact is one of the main reasons LAME should have an adaptive lowpass (which might actually eliminate ringing entirely).

QUOTE

Just like the mp3 encoders, different mp3 players use slightly different algorithms and methods, giving them all an individual sound. This has of course the side effect that there are better and worse combinations of mp3 encoders and players. I've noticed differences in the same mp3 file (encoded with BladeEnc) when running it through WinAmp (pleasant sound, but with clearly noticable pre-echo), Sonique (a sharper sound which I wasn't too fond of, but with less noticable pre-echo) and XMMS (Linux player. Pleasant sound, less pre-echo).

This is a bit amusing also. As if a truly compliant MP3 decoder is going to change the amount of pre-echo which is, in effect, [b]hard coded into the file via window sizes and the like. Heh.



I totally disagree with this also. I've seen people use this argument before and it just doesn't hold up. In general, a more well tuned psychoacoustics engine will produce less artifacts for everyone if it is done right. There is no way that there is going to be the extreme variance that he seems to imply.

An example is MPC. This format has some of the most well tuned psymodel I've seen (perhaps the best ever so far?). Just about 100% across the board, everyone I've seen has agreed (or has not been able to prove otherwise) that it is significantly less artifacted than almost all other encoders/formats. This difference is universal, there is not some imaginary variance there. Then he says "that's why we will never see a perfect encoder". I don't believe this is because of the reasons he gives exactly, but because we cannot possibly account for every single signal that is possible to create or every encoding accuracy needed. I don't think it has anything to do with people simply "hearing it totally differently". That's not to say that some people can't hear better than others, especially for particular artifacts.. but that's different. If you improve that artifact for one person correctly (without just shifting the distortion somewhere else), it should be improved for pretty much everyone who can hear it.

Anyway.. enough of that. Back on topic though, I think the kind of statements you see made there, along with some other ridiculous claims about quality made on other pages, are the reasons people keep thinking that some how Blade is the best encoder out there when in fact, it might be the worst. As long as misinformation like this is so prominently displayed, there will always be people who simply believe it as you apparently seem to keep finding out.

Btw, I have nothing personal against Tord or anything either, but I do think it is irresponsible to spread such misinformation when he probably even knows better.

Dibrom, does this all mean that (at least theoretically) it's possible to create a perfect universal psymodel, which can later be used in any lossy codec?

Hrmm.. probably not a psymodel per se. The reasons why could get pretty technical, but not all encoding approaches would even be compatible with certain aspects of a psymodel.

However, psychoacoustic phenomena should be universal (though sensitivity to them will vary by person) which is why I say an improvement according to those guidelines to one person should also be an improvement to another person (the reduction of pre-echo for example) assuming that no other difference was introduced simultaneously.

All psychoacoustic based lossy encoders should share similarities then which may be what you are asking about. They should all exploit masking effects of some kind. So in general, a better understanding of the principles behind psychoacoustics would certainly allow for better applied psymodels universally.

Great, Dibrom, instead of explaining all those details why Blade is crap, I'll just give that git a link to this thread (Probably I'll then find out that he doesn't even speak English, and if he does, he won't be happy about me calling him a git )

What were those other facts that make Blade so bad? Gabriel once posted a nice list on the r3mix board, the only points I can remember are:

- no joint-stereo
- no lowpass filter
- those legal issues because of the ISO code (not necessarily Blade's fault)
- some issues with "block switching" or whatever it's called.

I'll do a search and see if I can grab that thread, there were lots more issues (probably you could list them here and now, Dibrom).

<OT>

*scratches head and looks ashamed* It's about time I and Benjamin got this newbie website started that I have been promising for nearly half a year now... seems even worse than the (in)famous "Vorbis RC3 is a week away". (If I think about it, if I added up all the time I have spent at the Chip.de forum writing looong explanations and arguing with clevercloggs - I could have had the site online in that time, easily...)

BTW, while we're at it (this is getting quite silly, but quite amusing IMHO): I have been invited to a party for the weekend (where MP3 music will be played of course), BUT have been sharply warned: "ONE WORD about the selection or the quality of the music, and you can go home! We don't want to hear a single 'that song rings' or anything like that from you!" (Silly but true, this really happened today, some people take my concern for quality a bit too seriously, hehe )
</OT>

CU

Dominic

Hmm I couldn't at all find that thread about Blade's flaws at r3mix anymore.

I copied Gabriel's list of Blade flaws to my page at:

http://ff123.net/notrecomm.html

But the reason I personally don't use Blade is not because of that list, but because it sounds like shit at 128 kbit/s, and I notice a quality degradation in normal acoustic music (for example, Filburt's favorite, 41_30sec.wav), even at 256 kbit/s. It has to be pretty bad for me to notice defects at that high of a bitrate.

ff123

There is an important reason for ISO/Blade to sound really bad:
it's the quantisation selection.
As you know, an mp3 encoder usually uses a kind of iteration loop in order to choose the best quantisation. If it is able to find a quantisation with no distortion, then use it, otherwise use the least damaging one.
But as ISO code is not really smart, if there is no distortion-free one found during the iteration, instead of using the best quantization it uses the latest one. Too bad because usually the latest one is one of the worst.

A note about Tord:
I wouldn't flame Tord like that. Yes, Blade is as bad as ISO. But Tord is aware of this. Keep in mind that as most of us, is playing with mp3 during his spare time. And sometimes you really don't have so much spare time.
The quality page was written a long time ago, and 'has not been updated, probably for a time reason. He's not the only guilty for the joint-stereo section. I think that I indirectly contributed to it. A few years ago I had a wrong understanding about js, and then put some "not totally acurate" info about it on my website. As this info was ripped in a lot of places, ideas such as the fact the in js the low freqs are encoded in mono are my fault.

There is also something that you should not forget. Yes Blade is really bad, but keep in mind that Blade really contributed to the popularity of mp3 and to its widespread use. Tord was one of the few to bring us an encoder a lot faster than the iso code, and also was the only one to "resist" against FhG. With Blade, although the quality is bad, Tord definitively contributed to the mp3 history. Let's not forget that.

this is very true, i recall myself using Blade long way before i was aware of Lame. i think the mp3 scene owes a lot to Blade (even if it has now been left out in the dust).

My only objection is that the Blade website tries to convince people that the deficiencies of Blade are good things:

It reads as if the intention was to design the best quality mp3 encoder possible (for classical music), and that the optimum way of acheiving this is by disabling many of mp3's best features!

This is plain misleading. The truth is that Blade is taken from intentionally crippled code which was designed to sound awful. If the bitrate is high enough, and the music is simple enough, then it will just manage to encode a reasonable sounding mp3.


I've had emails from people who believe everything Dibrom quoted from the Blade website - they want to know why people would use compromised encoders like FhG or Lame which include so many features that reduce sound quality.


I hadn't noticed the bit about mp3 players before - LOL!!!
http://www.mp3decoders.org/

To confirm: the information about difference players in the Blade website is all hype and spin. From the best to the worst decoders, none of them to ANYTHING to reduce (or increase) pre-echo. And a decoder is either compliant, or it isn't. If it's complaint, it gives the same output as every other compliant decoder (down to about -90dB FS). Audible pre-echo is typically at -6dB relative to the signal, so the -90dB differences certainly aren't causing any "less pre-echo".


In a school project, you often try to write about your work such that the deficiencies become positive aspects. You shouldn't do that when appearing to offer a (free) product to the public.

Cheers,
David.
P.S. (after writing that)
DISCLAIMER: replaygain doesn't really work (even if it sounds like it does) and the mp3decoder tests are flawed (even where they confirm what others have found). There - you can't get me now! ;-)

Having done much experimentation with LAME in this area, I totally agree. In fact, LAME's quantization selection still isn't "perfect" either. It seems to be a tricky prospect to find a perfect balance between being too aggressive on easier signals (allowing too little distortion) and not being aggressive enough on harder signals (allowing too much distortion). Of course, if masking calculations are perfect in the first place (which they aren't ) that would help things along also..

QUOTE

A note about Tord:
I wouldn't flame Tord like that.

The intention was not to flame him on a personal basis. That's why I said I have nothing against him as a person.

QUOTE

Yes, Blade is as bad as ISO. But Tord is aware of this.

And there is the issue. Like David said, the problem is that the website tries to spin off the deficiencies as being some sort of "features", in effect allowing Blade to be higher quality. This is totally false and misleading and if Tord knows the ISO code is bad, he must also realize that Blade is not high quality, and there is [b]no excuse IMO to leave false information such as that online, free time or not. If you don't have time to update it, then take it down or put a big fat disclaimer at the bottom that the information is out of date and likely incorrect.



Well like I said, I don't have anything against Tord or even Blade itself. What I take issue with is the blatantly misleading information from which it is obvious that it exists only to make Blade look better when, in fact, it is quite inferior.

Like David said:



And I totally agree..

Hi,

there you can read at the moment about MP3 and followers like MP3plus, MP+ (haha, isn't it MPC now since a longer time ?!), Ogg Vorbis, AAC.


They tell:

MP+ is private, only one developer (Andre Buschmann) and not better than MP3........


AAC would be much better than MP3 and it would be supported by industry.

That's correct.


Ogg for them not better than MP3, not important.....


MP3 plus : better than MP3, industry

correct


WMA : Microsoft, worse than MP3 .....................



Interesting is, that industrial formats are described quite correct, AAC, MP3plus, but they deny free formats like Ogg and MPC !!!!!!!!!!!!!!

That's a bad tendency. CHIP is a quite important PC-magazine, like ct.


Could somebody please write here, what's correct, to tell the truth, not the CHIP propaganda:

Eg. is WMA 8 really worse than MP3 ?

I think I know which report you mean. Look at this thread:

http://www.hydrogenaudio.org/forums/showth...s=&threadid=214

Yes, it tells the same online.

In the paper magazine they have written nearly the same.

Unfortunately a lot of average users will read that crap article.

this is only one point where industry is more important then what the consumers would really need. Just take a look at the energy production, why do we use nuclear power, gas power, crude oil power if we could have solar energy or wind enery? Because industry would not make the big profit, it would be the consumers who would save a lot and of course our planet. But money is still the leading aspect of life! Sadly.