What are the advantages of using [any] ABR [settings] instead of straight up 320kbps CBR in [my] mp3 files if [the resulting] file sizes are irrelevant?

Are there any? The Hydrogen Audio Wiki says the following:

"Therefore, hydrogenaudio forum members recommend that for maximum audio quality at a given average bitrate, a VBR MP3 produced with the ABR method is always more desirable than a CBR MP3 at the same bitrate." Source: http://wiki.hydrogenaudio.org/index.php?title=ABR

However, it doesn't mention if this statement applies to 320 kbps CBR mp3 files. From my understanding, 320 kbps is the highest possible bitrate setting a LAME encoded mp3 file can have correct? If so, there is no room for any given frame in an mp3 file to sound better than the 320 kbps it would have had instead of any given ABR setting used. If there is room for sound improvement over an 320 kbps CBR encoded mp3 file, that would only be as a result of the implemented psychoacoustic model as part of the encoder correct? That is, how the encoder chose to impose it's psychoacoustics as it created the mp3 file.

I have two goals:
1) I want to be able to play music, audio books, and recorded speeches anywhere and on anything. This is the reason I use the highly portable mp3 file format, and specifically, the CBR setting.
2) I want the highest quality mp3 file possible. Although I'm using the mp3 file format- and therefore sacrificing some quality for portability- I still want the best audio I can get.

I am not new to creating mp3 files. However, after doing a lot of reading on CBR, VBR, and ABR settings, this question still remains unanswered. The Wikipedia and Hydrogen Audio Wikis are great sources of information, but neither of them have tackled this question as clearly as I (and from what I have seen, others) would like. Searching the web and reading through many forums threads hasn't gotten me any further either.

Thanks in advance for your help.

You could simply have used the search feature, so here's the short summary:

If you want the highest theoretical quality totally disregarding filesize and if you are able to hear it, then use 320kbps or better lossless.

VBR and ABR is for sane people who do not want to waste space for stuff which they cannot hear anyways.

Lyx, thank you for your answer. Yes, I did use the search feature within the Hydrogen Audio Forums. I did not find the answer. Therefore, I decided to put my question out there and see who is willing to provide some aid in the form of solid answers.

Of course, it is entirely possible that I may have I missed the answer through my searching and reading. Certainly, others have come before me and asked the same questions. Even if in a slightly different way. If so, please be kind and point me in the right direction. Please do not chide my post and assume that I didn't do something so basic as searching for the answer. And please do not imply that I'm insane or wasting space since I have written file sizes do not matter for my project.

Perhaps I am looking for an answer that is right in front of my eyes. So I will ask my initial question in another- more basic- way. The highest quality and most compatible or portable mp3 files available today use 320 kbps CBR. Correct? Is there anything better than this? If so, please help me understand by saying so and proceeding to explain why. Or at least point me in the right direction by giving me a link and telling me why I should be reading its content and what I should deduce from the text (so that I clearly understand what you are trying to say).

Thank you. I don't mean to waste anyone's time. I am honestly trying to figure this out. :-)

It's true that the corresponding CBR are always of a worse (theoretically) quality than the same VBR - Because the VBR are able to gain up to 320 kbit frames when nessesary, where CBR are always capped at the given CBR.

Therefore the true statement are "VBR/ABR at the same average bitrate will always have a higher quality than a CBR file at the same bitrate, except for the highest available bitrate."

However I think it's a waste of space to use 320 kbit for a lossy file, since there will always be limitations like lowpass filtering, and I've seen many posts with people claiming that older encoders have specific problems at that bitrate. Then use around 700 kbit for a lossless file instead

But really, instead of trying to percieve a theoretically maximum quality for a compatible audio file, take a few ABX tests, to see how low you can go without noticing quality loss - You'll be amazed! (I was! )

Sometimes, what happens is that we don't know how to look for the answer, and that's what happened to you:

List of recommended LAME settings ( sticky on this subforum )
http://www.hydrogenaudio.org/forums/index....showtopic=28124



So, for your needs, you've made the right choice.

[Edit: Ok, "right choice" IF best quality, MP3 and no problem about filesize. Of course speech could be perfectly ok at mono and 160kbps.]

Audio books and speeches at 320kbps? I'm not sure that's the 'right choice'.

The OP should really do some ABX testing to determine what VBR (possibly ABR for the lower bitrate voice files?) setting they really require. Blindly using 320kbps is just foolish.

Lame freeformat can go beyond 320 kbps frames, but you will be able to play it only within a specific tool environment.
Moreover, you can't expect quality to really increase because of the fact that quality is already very, very high in the 256 kbps range, and from the fact that the little little bit lacking towards perfection is not expected to be significantly improved when going higher in bitrate.

So even cbr320 isn't wise to use for most mp3 users.

As for quality:

Using ~128 kbps already provides a very good quality for most people and most musical situations.
In this bitrate range a vbr method like Lame's VBR should provide the best quality, but it also depends on the implementation.
Same is true when going higher in bitrate, and quality is getting better and better also for most of the situations that are not encoded well at 128 kbps.
Beyond 200 kbps quality of a good encoder is so high that quality increase is only relevant for rare samples that are hard to encode, and even then it's a matter of taste whether say 256 kbps is worthwile the quality increase over say 192 kbps.

In the ~ 250kbps range using a vbr method like Lame's VBR is questionable IMO when talking about quality.
On one hand CBR 256 or ABR ~250 provides an excellent quality which is hard to improve when going CBR 320 (see above), and if it's not possible for CBR 320, so not for any VBR method.
On the other hand it's often overseen that VBR amplifies weaknesses of the underlying psy model. In those rare cases where the psy model flaws the error is amplified by using VBR cause VBR totally relies on the psy model.
In short the chances/risk relation using VBR in the 250 kbps range isn't very good.
As always exact meaning depends on implementation (encoder used).

Comparing ABR with CBR ABR is preferable in principle. In both cases bitrate variation is based on the same quality criteria, but CBR audio data bit rate variation is limited by the size of the bit reservoir.
At 256 kbps however this isn't an essential difference because of the extremely good cbr 256 quality of a good encoder. Moreover CBR does allow for a meaningful variation in audio data bitrate though it is restricted.

Taking it altogether:
As you seem to be mainly interested in quality:
My advice is to use ~ 250 kbps, and do so using ABR or CBR whatever you like best. Don't care too much about technical stuff.
A maybe more relevant question is what encoder to use, but again, at ~250 kbps there's not much difference whether you use Lame, FhG, or Helix.

Added:
I was talking about high quality music encodings.
As for speech 128 kbps should be more than enough.

I don't think it applies.
An ABR file will use variable bitrates, but will average out at the bitrate you have specified.
Since 320 kbps is the highest a standard MP3 will go it can't go under 320 since that would make the avarage bitrate below 320.


I've come across one DVD player (Harman/Kardon) that had great difficulties playing a 320 kbps MP3 file. It played, but with a lot of skipping and weird noises.

However, I think it is unlikely that you will experience difficulties with VBR or 320 kbps MP3-files on modern equipment.

Note the earlier-in-the-thread mention of "free format", which allows larger than 320kbps frames. In addition, there's the also the ability to disable the default "strict ISO" setting of Lame, which normally requires that bit-reservoir use not create frames that have an effective size larger than a 320kbps frame.

But the simple answer, assuming one isn't using any of the fancy and less compatible features mentioned above, is: 320kbps CBR is as best as lame can sound.*

-brendan

* but there's almost no reason to waste that much disk space.

I'm aware of the "free format" that goes up to 480 and with some encoders 640 kbps. But there is little support for these files and "softy" wanted to be able to play his files "anywhere and on anything".

That was my hunch. Why doesn't someone add a sentence to the above statement saying it doesn't apply when creating mp3 files with the 320 kbps CBR or LAME insane preset (both are the same) setting? This would help avoid much of the confusion I've seen out there on the Internet. Folks would be able to reference the article for the answer.

Thanks for all of the answers and ideas. Here's what I have deduced from this conversation.

Q1. What are the advantages of using [any] ABR [settings] instead of straight up 320kbps CBR in [my] mp3 files if [the resulting] file sizes are irrelevant?
A1. There are none.

Q2. Are 320 kbps CBR mp3 files better than those using the highest ABR setting?
A2. Yes. So long as file size isn't an issue. It is for many people.

Q3. The highest quality sounding and most compatible or portable mp3 files available today use 320 kbps CBR. Is this correct?.
A3. Yes. Once again, so long as file size isn't an issue. It is for many people.

Q4. What if you just want the best sounding mp3 files (headphones, speakers), consistent results (portability, ability to play on anything), and ultimately the file size doesn't matter (because you see that storage space is increasing at amazing levels)?
A4. The short answer; Use 320 kbps CBR or the LAME insane preset setting (both are the same).

No.

Even though the *theoretical* quality is highest at 320kbit, it will in almost every case not *sound* better. Chances are high that without a direct comparision (casual listening) you cannot even tell the difference between v4 and 320kbit, althouth the latter will on average need 2x the space.

People who have never done an ABX-test typically overestimate their hearing-abilities and underestimate the power of their consciousness by leaps and bounds. Many people who do regularily do ABX-tests in turn tend to focus on rare problem-samples under unrealistic listening-conditions (direct back-and-forth comparision while ABXing, very high focus) - thus inflating the perceived significance of the results.

Off course, there are those "one amoung thousands"-people for whom things are a bit different(hi guru :) ), and everyone typically asumes that he or she is one of those - nearly everyone however isnt.

My point is: the asumed "danger" of perceivable quality-loss under normal listening conditions, is mostly just a thing in peoples heads. The mood in which you are today has a higher influence on how something sounds like, than the difference between v4 and v2 - let alone 320kbps CBR. Or to put it more frontal: people are paranoid.

Glitches during non-secure ripping on the other hand...........

- Lyx

Hi Lyx,

IMO in principle you're right at least when thinking of a close-to-ideal mp3 encoder.

But looking at real world and encoding for instance Moon with the currently recommended Lame encoder at -V4 --vbr-new I'm pretty sure you will not find the encoding acceptable. No ABXing required, and Moon isn't a special problem samples. Lame 3.97 just encodes samples like these poorly when using -V4.

Using very high bitrate like CBR320 does have the advantage that even when the encoder flaws on specific samples (or the mp3 format flaws) this brute force approach makes sure perfect (or at least good) quality even in these cases.

So using a very high bitrate has the same psychological background like using a lossless codec. Encodings then aren't lossless when using mp3 but perceptionally close to it, and that's more or less guaranteed even for special musical situations.

Sure that's usually overkill but everybody does it the way he likes it best. Disc space difference when using -V4 or CBR320 isn't an issue to many people nowadays.

No it is. And it's not only a true problem sample (i.e. not something which happen on every track nor album but on contrary something linked to specific and rather uncommon signal), it's also something which doesn't last very long (max duration of artifact = 1 second). How many "moon-like" sample have you discover on your side? You should have found tons of similar cases if this problem wasn't a very special one, don't you think?


And how do you explain that pre-echo is ABXable at 320 kbps CBR with several samples? I don't call such 320 kbps encodings "perfect" or "good" when it comes to appreciate the quality on these critical parts of music: it sounds poor, unsharp, smeared... but again it only last few micro-seconds and doesn't happen every 30 seconds. The only reason persons (even some trained ones) can live with this issue is simply that it offers more advantages (hardware compatibility, space saving) than problems.


You can get the same effect with half the bitrate, spare much more space and as sole reverse side have to suffer few additional artefacts. That's probably why 320 kbps encodings aren't very popular: benefit = zero.

On most portable players it still makes a real difference... there's disk space first and also battery life.

Really? Please tell me how, I'm interested

You're saying more or less the same thing Lyx says, and you're both right. As is the OP.
The point is where to put the sweet spot, and that's a personal thing even when the additional benefit when going -V4 to CBR320 is very small.
Obviously the OP doesn't care about the additional disc space. And obviously everybody who cares will have a different view.

Your argument is based on the asumption, that objectivity does not exist at all, and everything is subjective. If you think like that, then why are you here?

To back up the above claim: i as well as guru only stated rationally valid facts. And from current knowledge fact is, that if VBR is acceptable (compatibility) and efficiency has any weight, then 320kbit CBR is stupid. According to current knowledge it is also a fact that for almost everyone there is no perceivable difference under normal listening conditions. Even for those with bat-ears, the artifacts under normal listening arent bad, but just suboptimal, and they are rare and short.

You can of course ignore the "cost" factor completely and make quality the only criteria. In that case, 320kbit CBR is the optimal choice. I stated that two times in this thread.

However, you are twisting the facts, so that they fit the conclusion better to which you want to arrive. You put the effect before the cause. Or in other words, you are trying to cheat yourself to justify to yourself what you want to believe.

1. "Here are the facts - which conclusions can we draw from them?"
2. "Here are the conclusions - which facts can we find to support them?"

"2." plain and simply sucks for anyone with a desire for objectivity, and therefore has no place on this board.

Choice and personal criteria is something different than facts. You can choose whatever criteria you want for your ratings, and its okay because ratings are relative. But dont mess with the facts. Dont try to change reality just so that it agrees with preconceived conclusions.

- Lyx

To make short:
mp3 coding introduces quantization noise; quantization noise always lowers the objective quality but not necessary the subjective one. With the same encoder, the quantization noise is necessary lower at 320 kbps CBR than with any VBR encoding. As a consequence CBR 320 maximizes the objective quality but the impact of such low quantization noise on human's (subject) perception is inexistant in most situations.

Indeed, the audibility of quantization noise depends on:
- listener abilities (natural ones + training)
- listening conditions (environment, hardware, concentration)
- kind of encoded signal

In five years nobody on this board was able to prove that 320 kbps brings any benefits over inferior bitrate encoding on normal listening conditions excepted for rare pathologic cases (I mean samples, not persons). In other words it's very unlikely that someone would find any advantage in using 320 kbps instead of good VBR encoding (even ~130 kbps encodings are proved to be very good for several listeners). Latest LAME's release improves overall quality in VBR mode but introduced a new artifact audible in very specific conditions (on very tonal signal); this artifact is almost fully corrected with latest alpha.

To conclude:
- perfect, 100% transparent and trustable encoding can't be achieved with any lossy formats and certainly not with MP3. Use lossless instead.
- excellent, transparent encodings can be achieve with ~200 kbps VBR encodings (sometimes much less, and of course with higher bitrate too).

Once transparency reached is there something to gain with higher bitrate excepted problems (compatibility issues, shorter battery life on portable devices, etc...)? No. Experience have proved that high bitrate lossy encoding doesn't bring piece of mind for a long time and that most of not all anal people are moving to lossless and focuse their full attention to the ripping process.

My advice: if you're worried about quality save your time and start using lossless, which is easy to convert to lossy at friendly bitrate for all specific needs.

Where to put the sweet spot is a subjective decision.

Facts are objective, and everybody can learn about objective facts here on HA. In a previous post of this thread I did show up myself that quality can hardly be expected to be increased when being in a bitrange range way below CBR320.

Decisions are not objective. It's always subjective how to weigh the facts, and that's what it's about when deciding. Decisions are emotional, and there is for instance the emotional target of near-perfection to some people. To me there's nothing wrong with it as long as the disadvantages of the decision don't bother them. And even if they do it's up to them.

Guess you won't argue with people using -V2 though this too is overkill most of the time.
Okay, -V2 isn't extremely far from -V4, but what about people using -V1 or -V0?
What I want to say is: these people have made up their minds to give away disc space for the sake of better quality in rare situations where this is so. Not essentially different from what the CBR320 people do (though I wouldn't want to use it either).