Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Next step up from 320Kb/s MP3 is...? (Read 10931 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Next step up from 320Kb/s MP3 is...?

.
I did some searching through the forums before posting this, but I don't see much in the way of 'better than 320Kb/s'.  Is there such a beast as 640Kb/s ~ MP3 file format?

Yes, I realize that many folks will shout "why would you want that, 320Kb/s is transparent"  However, just work with me here, for the sake of discussion / learning, is it possible, with any tool currently available to the music community, to create and playback a 'super-high-bitrate' MP3 file?

Thank you in advance.

Andrew D.
www.cdnav.com
.

Next step up from 320Kb/s MP3 is...?

Reply #1 –
lame --freeformat, see lamedocs. Often unsupported by players. Useless, as with lossless you can achieve the same compression.

Next step up from 320Kb/s MP3 is...?

Reply #2 –
With Lame you can use freeformat which allows for higher bitrate than 320 kbps.
However you'll have problems playing them back at least with DAPs.
For mere PC use guess you'll find a solution, but for mere PC use you have better alternatives.
lame3995o -Q1.7 --lowpass 17

Next step up from 320Kb/s MP3 is...?

Reply #3 –
Lossless is the best next step not because you had 'golden ears', but for the peace of mind that you didn't have to make the choice to throw anything away. With lossless you stand above the arguments about which traditional format is "the best".

Next step up from 320Kb/s MP3 is...?

Reply #5 –
Even if players did support 640k mp3, I wouldn't want it. If 190k is generally transparent quality then 224~256k is the extra headroom. Above this you only carry over the deficiencies of the format and waste bits.

Next step up from 320Kb/s MP3 is...?

Reply #6 –
Higher bitrates won't fix artifacts, as higher bitrates do nothing more than reduce quantization noise.

Next step up from 320Kb/s MP3 is...?

Reply #7 –
Three simple scenarios:

1. At around 400-500kbps you still cannot be sure, that the ultra-rare artifacts get fixed, and you will get very low compatibility. A hybrid lossless encoder in lossy-mode - like for example WavPack-lossy... gives you more stable quality and better compatibility at this bitrate-range.

2. At around 600-700kbps, mp3 will still be lossy and you will get very low compatibility. A lossless codec like i.e. FLAC or WavPack however will give you lossless at the same bitrate and in addition offer better compatibility.

3. For ultra high bitrate VBR encoding, mp3 above 320kbit will improve, but usually not fix, a small amount of samples, and you will get very low compatibility. OGG Vorbis offers you the same benefits but with better compatibility.

Thus, MP3 above 320kbit is irrational and useless. The point of MP3 is its inbeatable compatibility. If you take this away, then other alternatives perform better. The only reason to use MP3 is compatibility - if your priority isnt compatibility, them MP3 is a bad choice.
I am arrogant and I can afford it because I deliver.

Next step up from 320Kb/s MP3 is...?

Reply #8 –
Quote from: Firon on
Higher bitrates won't fix artifacts, as higher bitrates do nothing more than reduce quantization noise.
Is there a difference?

I thought that everything that's "different" from the original in an mp3 is due to quantization (or worst case: zeroing) of DCT coefficients. These "differences" manifest themselves in various ways as we listen, partly due to various other processes within mp3 itself, partly due to human perception. But it's all quantization noise.

Throwing more bits at the problem will reduce its severity, and must eventually make it go away. Whether 640kbps is enough, who knows? I would guess that it is, almost always!


I agree its a pointless discussion because freeformat mp3 is so incompatible.

Lossy FLAC is a much better idea

Cheers,
David.

Next step up from 320Kb/s MP3 is...?

Reply #10 –
Quote from: Lyx on
3. For ultra high bitrate VBR encoding, mp3 above 320kbit will improve, but usually not fix, a small amount of samples, and you will get very low compatibility. OGG Vorbis offers you the same benefits but with better compatibility.


I thought that all mp3 above 320kbit were necessarily CBR?

Next step up from 320Kb/s MP3 is...?

Reply #11 –
Quote from: pdq on
Quote from: Lyx on


3. For ultra high bitrate VBR encoding, mp3 above 320kbit will improve, but usually not fix, a small amount of samples, and you will get very low compatibility. OGG Vorbis offers you the same benefits but with better compatibility.


I thought that all mp3 above 320kbit were necessarily CBR?

Honestly, i wasn't sure if it is possible and just added scenario #3 as an afterthought, to cover all possibilities.
I am arrogant and I can afford it because I deliver.

Next step up from 320Kb/s MP3 is...?

Reply #12 –
I disagree with the principle that "640 kbps mp3 is pointless since Lossless encoders are a better way to go". Lossless encoding of contemporary music generally ranges from 900 to 1200 kbps which is much larger. Compared to the 1411 kbps of CD audio itself, I would argue that Lossless is the most useless method of compression. I personally would love to use 480 kbps mp3, but like others have said I've found it difficult to find a convenient player/decoder to support it, so right now I'm still encoding 320 kbps mp3s.

If I really wanted higher bitrates, perhaps OGG or other formats which naturally support higher bitrates (I dunno which do and which don't) would be a better choice, but I would have to personally listen to any suggested format for a very long time before I were convinced to switch.

> At around 400-500kbps you still cannot be sure, that the ultra-rare artifacts get fixed...

I agree with this, mp3 could have ultra-rare samples with 1) artifacts which never go away regardless of bitrate, though I've never heard any before, or 2) artifacts which only go away as the bitrate goes very high, possibly approaching lossless. From personal experience, there are a number of "type 2) artifacts" that go away as far as I can tell at around 500 kbps mp3, so I think it's worth it. I speak mainly of 'Eig', just about the most severe pre-echo killer sample for mp3 I know of. 'Castanets' is also ABX-able I think (I didn't rigorously ABX, but to me it's obvious) at 320 kbps depending which version of LAME you use (one that disables the bit reservoir for 320 kbps mp3). Castanets is an example of a severe pre-echo test that is not as bad as 'Eig' (which is almost like a delta function pulse so it's close to the worst possible...Castanets is sharp pulses too but there is a slightly tonal aspect to the strikes and they are also bandwidth-restricted to high freqs) and it probably becomes transparent at 400 kbps to 450 kbps so I think 500 kbps is good leeway to fix almost all pre-echo problems with mp3. 'Castanets' is transparent to me with LAME 3.92 320 kbps (bit reservoir on in this version, so therefore it's really a 450+ kbps file in disguise for 'Castanets' and 'Eig') while 'Eig' is almost but not perfectly transparent at that setting.

> A hybrid lossless encoder in lossy-mode - like for example WavPack-lossy... gives you more stable quality and better compatibility at this bitrate-range.

I disagree about the stable quality thing. The 'Keys' sample that shadowking pointed out recently requires about 500 kbps for transparency with WavPack lossy. I was able to make a artificial sample (17 kHz sine wave in left channel, 18 kHz sine wave in right channel) that requires 800+ kbps to be transparent with WavPack lossy. The main problems with the hybrid-lossy encoders is actually lack of stability...the transform-based encoders like mp3 and OGG are the more stable. That's why 2Bdecided is currently working on his 'Lossy FLAC' which is like a VBR Lossy mode for FLAC (he's still developing it now though as we speak so it's probably not ready for widespread use yet).

But better compability, definitely.

> OGG Vorbis offers you the same benefits but with better compatibility.

Yeah, I should try OGG myself someday. Everyone says it is better, but until I listen critically for myself (I've listened to OGGs before, but not critically, and I didn't play around with encoders I just used the default settings) I won't be sure.

> The point of MP3 is its inbeatable compatibility. If you take this away, then other alternatives perform better.

I'm not saying you are wrong, but I'm not convinced of this until I hear for myself someday. OGG may be far superior to mp3 at low to medium bitrates, but at higher bitrates, in general I think the various transform-based audio codecs might start to perform similarly. I think it's wrong to assume that if OGG sounds better at lower bitrates that it will still sound better at extremely high bitrates. And I rarely see people executing careful listening tests at extremely high bitrates. Most people just say there is no need so they don't bother to test.

I'm most hopeful for the true VBR hybrid lossy encoders though (like 2Bdecided's lossy FLAC, I'm not sure if Optimfrog's VBR algorithm counts or not) but they are still in development. I'm also hopeful for better VBR transform-based encoders (not necessarily mp3). True VBR should be smart enough to use bitrates ranging from 0 kbps to 1000+ kbps. Many of the lossy codecs including mp3 already have VBR but their versions aren't dynamic or perfect enough, in my opinion.

Next step up from 320Kb/s MP3 is...?

Reply #13 –
I have hundreds of albums in wavpack 320~350k and never heard a problem and that is much more representative than a collection of 5 test signals.. We are getting too distracted with pathalogical cases although I do agree some wavpack could benefit from some tuning in those cases. I can abx one sample AAC at 350k vbr - Do I now start making claims that AAC is not ready for prime time ? Also if you can't hear mp3 preecho in Florida seq or emese I don't think you are sensitive to that artifact.

foo_abx 1.3.1 report
foobar2000 v0.9.4.3
2007/06/22 00:58:26

File A: H:\temp\abx tests\A03_emese.flac
File B: H:\temp\transcoded\A03_emese.mp4

00:58:26 : Test started.
00:58:42 : 01/01  50.0%
00:58:49 : 02/02  25.0%
00:59:07 : 03/03  12.5%
00:59:21 : 04/04  6.3%
00:59:30 : 05/05  3.1%
01:00:20 : 06/06  1.6%
01:00:26 : 07/07  0.8%
01:01:14 : 08/08  0.4%
01:03:08 : Test finished.

----------
Total: 8/8 (0.4%)

True stability you talk about is achievable with Vorbis and AAC, but i believe it is in the 350~400 k range.  They won't have pre-echo limitation of mp3 or the noise of the hybrids. They are also not low bitrate crippled like mp3 / mpc / hybrids. The NERO encoder has amazing potential in quality and useability.

When we add the very rare cases from each encoder the answers aren't as clear as we think. I am interested in modest bitrates for portable use and very high bitrate lossy for 'archiving' level quality [350~400k]. In some past tests i've done mp3 and freeformat didn't cut it and support is terrible. Now I am playing with AAC and it looks promising. International standard, great performance for ultra-low bitrate to very high bitrate, multichannel ready.. I started some transcoding test 350~400k AAC > 128k AAC or V5 mp3 and the results are transparent so far.

It seems today there are several options for very high quality which could be explored. Mp3 is a great performer at mid-high bitrate for such a technically limited format and its age, but clearly its not as robust as AAC or Vorbis. mp3 quality vs usability 'limits' are 128~320k. I am very interested in AAC / hybrids / 2bdecided's new thing.

Next step up from 320Kb/s MP3 is...?

Reply #14 –
Quote from: shadowking on
We are getting too distracted with pathalogical cases.

Agree. If all your music has nothing but problem samples in it, you have a bad taste in music (IMO, unless you listen to all harpsichord music or something similar). I'm willing to bet you can't tell the difference between 90+% of the music out there at 320kbps on MP3, AAC, OGG, MPC, or WMA and a cd of the same music.

You could always just use a wave editor to edit out the problem sample then encode it at 320kbps. Problem solved.
[span style=\'font-size:8pt;line-height:100%\']"We will restore chaos"-Bush on Iraq[/span]

Next step up from 320Kb/s MP3 is...?

Reply #15 –
Quote from: QHOBBES 2.0 on
Quote from: shadowking on

We are getting too distracted with pathalogical cases.

Agree. If all your music has nothing but problem samples in it, you have a bad taste in music (IMO, unless you listen to all harpsichord music or something similar). I'm willing to bet you can't tell the difference between 90+% of the music out there at 320kbps on MP3, AAC, OGG, MPC, or WMA and a cd of the same music.

You could always just use a wave editor to edit out the problem sample then encode it at 320kbps. Problem solved.


In the case of wavpack just copy the .WVC correction file and you get the lossless file back. That is great innovation and useability.  Mp3 was great innovation 15 years ago. MPC was great 6 years ago. Both pale in comparison to Nero ability now and don't have hybrid lossless restoration capability.
 

Next step up from 320Kb/s MP3 is...?

Reply #16 –
I agree that the extreme problem cases are super rare, but still that is up to each individual to decide for himself if he cares enough to try to make encodings to take into account such cases.

Quote from: QHOBBES 2.0 on
Agree. If all your music has nothing but problem samples in it, you have a bad taste in music...unless you listen to all harpsichord music or something similar. I'm willing to bet you can't tell the difference between 90+% of the music out there at 320kbps on MP3, AAC, OGG, MPC, or WMA and a cd of the same music.
WAY more than 90%, yup. Probably 99% or more and I cannot tell the difference. Still, what you do about the remaining 1% is someone's own choice to make, the decision shouldn't be made for him.

Is harpsichord music tough for transform-based encoders? I'm guessing it can be moderately difficult for hybrid lossy/lossless encoders, but I wouldn't have thought it particularly tough for mp3.

But as you noted, things can change depending on your taste in music. In my case, my favorite kinds of music (tinkly high-freq music) in all the world is exactly the same kind of music that is dangerous for hybrid lossy encoders, which is the main reason why I'm not using them as my primary encoding method yet (I may later on, if true VBR for them gets fully developed).

Regarding mp3 and other transform-based encoders, pretty much none of my music is a danger. The stuff that's most dangerous for them is not songs, but sharp noises which are preceded by dead silence (if there's also a full bandwidth song in the background that is sufficient to hide pre-echo from me in almost all cases). But depending what you are encoding with mp3/ogg/etc...you could be in danger also. If for example you encoded a movie where some guy is running around shooting guns at people (without BGM playing), you could have issues if you really care about quality (most people including me don't care though, for movies).

I've never heard florida_seq or emese yet, I've only downloaded a tiny handful of samples so far, maybe I'll listen to them later. I'll probably seriously look at OGG and AAC too, eventually (in a year or two, which will hopefully also allow time for the hybrid encoders to improve). For now I'll make do with mp3.

Next step up from 320Kb/s MP3 is...?

Reply #17 –
I would say your music is not that safe. Listen to the track with the loud violin (bad ringing) in the end and also to the tinkly keys with mp3. Sure at 256 abr~320k it sounds good, but chances of getting these problems with AAC / vorbis would be much less and they won't sound as bad if they weren't transparent.

Next step up from 320Kb/s MP3 is...?

Reply #18 –
Quote from: Porcupine on
But depending what you are encoding with mp3/ogg/etc...you could be in danger also. If for example you encoded a movie where some guy is running around shooting guns at people (without BGM playing), you could have issues if you really care about quality (most people including me don't care though, for movies).


What Vorbis movie issues with > 320 k ??? We are talking high bitrates right ?

With mp3 you can't eliminate preecho. At higher bitrates who is gonna hear pre echo on vorbis or aac or any artifact ?

Next step up from 320Kb/s MP3 is...?

Reply #19 –
Thank you for all of the intelligent replies.  What is the highest bitrate encoding that I can use for Vorbis?  I dig Open Source!

www.cdnav.com
Andrew D.

Next step up from 320Kb/s MP3 is...?

Reply #20 –
Quote from: Porcupine on
Lossless encoding of contemporary music generally ranges from 900 to 1200 kbps which is much larger. Compared to the 1411 kbps of CD audio itself, I would argue that Lossless is the most useless method of compression.

I would rather call lossy codecs useless because they aren’t able to reproduce the original even with an infinite data rate (by definition of "lossy"). On the other hand, lossless codecs are able to save an infinite amount of space if you compress enough data with them.
FLAC.

Next step up from 320Kb/s MP3 is...?

Reply #21 –
Quote from: digital on
Thank you for all of the intelligent replies.  What is the highest bitrate encoding that I can use for Vorbis?  I dig Open Source!
Well, unlike MP3, Vorbis doesn't really have a target bitrate. See the Recommended Ogg Vorbis article on the HA Knowledgebase. -q 10 is the highest Vorbis quality ever... and extreme overkill.

I think 99.999999% of humans on this planet will find -q 6 to be transparent for 99.999999% of their songs.