Hi everyone.

I am not a regular visitor of this forum but keep come back from time to time to see if there is anything new i should be aware of.

Today I learned about the development problems of LAME. But putting this topic aside, someone mentioned of some limitation of the mp3 standard. I was wondering if mpc and ogg are all better mainly at lower bitrates why can't one (and please apologize my next expression) just copy-paste the good things from mpc aac and ogg into the framework of mp3 to make it better sounding. As far as I know (but its possible that i am totally wrong, that's why I am asking), each of them is using a psychoacoustic model of the human hearing. Taking this into account i don't see why one can't take an idea from an other implementation and by tweaking the parameters of the current model make the mp3 sound better.

I really really beg you pardon If i wrote something that would seem totally brainless to the experts in the area.

Can someone explain this issue with a reasonable technical detail?
Thanks a lot,

Matyas

Matyas, you have saved me starting a thread which will almost certainly infuriate some of the more 'sophisticated' members.

I was going to ask whether mp3 is capable of more, and whether anyone of sufficient capability has any interest in addressing the question. No doubt the question has been asked countless times before, but given the widespread hardware support, I wonder if there is any chance of the developers settling differences and getting back to business.

I'm afraid that I can't provide much technical detail, I'm more of an analogy person...

My car and a Boeing 747 are both methods of transport. They both use the same model for their action (Newton's laws, "every action", etc.). So why can't we take the fastest bits of the jet and 'cut&paste' them into my car?

Windows and Linux both run on i386 CPUs. Linux is much better than Windows at security, so why can't we take the good bits from Linux and cut&paste them into Windows?

You see the problem? The various methods aren't necessarily using the same psychoacoustic model of hearing (because there isn't one 'right' model), they make different assumptions based on their model (from simple things such as what frequency limits to use, to more technical decisions I don't pretend to understand), and then they implement those assumptions in different ways. I'm sure one could learn _something_ from another, but the work to convert it across would probably be way more than just tweaking what you've already got.

Cheers, Paul

Paul is right. Adapting a psychoacoustic model to another format is a PITA. Of course, you can adapt (at least parts of) a boing engine to your car. But, in the end, the result might be even worse than the original. In this analogy, for instance, it would probably be very cludgy to steer around. And you would probably need a fuel tank that can hold some thousand gallons of kerosen. :B

That said, it's not impossible. Some time ago, a developer was interested in porting the MPC psychoacoustics routines to MP3, creating an all-new encoder from it. But I'm not supposed to tell much about it, JohnV might kill me. Anyway, it seems the developer has given up this idea and was searching for other methods of improving MP3 quality beyond the level currently offered by Lame.

He will .....

Nah, I think he's traveling or something.

That should give me some more days until I die

Thanks for the answers. Although I hoped for a bit more technical explanation, the replies started to give me some image. So If I get it right there is no technical limitation in mp3 which could not be worked around (at least theoretically) to catch up to mpc. Am I right?

And regarding that analogy with a Boeing and a car. I gues there is a lot of differences between a car and a plane. Are the differences between mp3 and MPC that big too? Is Mpc a supersonic plane and mp3 a good car?

Matyas

Matyas,
I guess most users (including me) are just too lazy to list all tech differences between the various audio formats, since they did that many times before (Try the search function).

Additionally you could fr example read pages like Buschmann's explanation of MpegPlus, which is rather old and mpegplus is nowadays called musepack or mpc.
Or this one about temporal resolution (related to pre-echo effects) of various codecs. This is one of the theoretical limitations of the mpeg layer 3 format.

No, there ARE fundamental technical limitations. To make an analogy based on other computer file formats, PNG supports 24-bit color plus an 8-bit alpha channel (actually, it supports 48-bit color with a 16-bit alpha channel if you want, but that's not to common so I won't say more on it). GIF files only support 256 colors with the possibility of using exactly one of those for transparency. No massaging, copying, pasting or anything else will create a program that can make GIF files that have 24-bit color or an alpha channel. Likewise, GIF files can never gain the superior compression ratios of PNG, because they use LZW compression, which is a completely different technique than zlib (the compression used in PNG).

An example of such a limitation in mp3 is maximum bitrate. The maximum bitrate of a block in mp3 is 320kbps. That may seem like plenty, however, mpc occasionally uses more bits than that for short periods (I've heard as high as 700). This limitation of mp3 is a fundamental technical limitation of the format. The best you could hope for is to create a file format that is similar to mp3 but doesn't have that particular limitation. It would be a kludge though, and it wouldn't be backwards compatable with existing mp3 players, which defeats the whole purpose. You might as well create a whole new format, or use one of the existing formats that doesn't have the limitations of mp3. Effectively, you've turned your car (mp3) into a not-very-useful mutant half-car-half-jet thing that nobody lets you drive or fly anywhere. You'd have been better off either sticking with your car or getting your pilot's license and buying a nice jet.

That's just the tip of the iceberg when it comes to the limitations of the mp3 file format, but I think it demonstrates that what you propose is impossible. I'm sure others who have more technical knowledge about the different formats could go into more detail about some of the other serious issues related to sub-band vs. transform codecs, sfb21 stuff, time vs. frequency domain, etc.

Which is basically what MPC is, isn't it? Tossing out one of MP3's methods of masking and compression, and doing a whole lot of tweaking... (=

My understanding was that MPC is a subband codec more closely related to MP2.

Thanks phong, you made it a bit more clear to me. From what you wrote i start to see that what i'm really interested in is that subband/sfb21 stuff you mentioned. It may seem that i want to much but would somebody be so kind and make some note on these stuff too? And don't be afraid, I have finished telecommunication so i know what frequency and time domain is. I just would like to know which of them is used where and what limitation they present in addition to the bitrate limitation you mentioned. Also what is meant under subband and sfb21?
Thanks once again.

PS: Taking into account the progrees of this topic, if it would end well, it could turn out to be pretty much of a "chapterized" discussion getting more and more into detail. Hence its target group would be wide, because everyone could stop at the point he considers deep enough for him.

Limitations:
*no scalefactor for sfb21
*unable to toggle M/S on a sfb basis
*too long short windows
*too small long windows
*too limited max frame size

An other point is that some encoding schemes are featuring additionnal coding tools:
*PNS
*SBR
*TNS
*LTP
...

Hey, Roberto, if you die can I inherit your Pink Floyd collection? Or at least Shine On?

---

Concerning the MP3 encoding format, aren't most of its limitations because of ISO compliance requirements? The upside of which is hardware makers willingly developing MP3 players like the format will *never* die? I've heard ISO standards are a serious constriction which effectively prevents MP3 from being "expanded" furthur, but how extensive are these restrictions overall? Is there more to blame than just ISO?

Thanks Gabriel. This was what I wanted to know.