Hello, I've been reading articles on this forum for some time now and have learned a lot.
I also welcome the debate and clarifications that come out of it.

My question is NOT about what is transparent, but about STANDARDS used today in MP3 format in terms of best possible quality.

Ripper:
I've heard that EAC is the best ripping program to use, hands down.
Anyone disagree?

Ripper Settings:
I have heard of things such as cdparanoia and comparison ripper settings, etc.
Is there any settings for EAC that would improve quality?

Encoder:
We most all agree that LAME 3.97 is, as of now, the best encoder to use for the MP3 format.
Anyone disagree?

Encoding Settings:
I have heard that you can extend MP3s maximum range beyond 320kbps, but that it breaks compatibility with almost every device, player, software, etc, so that nobody does that. Any comments about that?

As 320kbps CBR is the maximum bit rate, there is only one other option I know about, Stereo.
I have seen that JOINT STEREO is ALWAYS better than any other form of Stereo for encoding. Anyone disagree? I've read the threads very carefully, I seem to come to the conclusion that Joint Stereo is always superior.

Other people might want to bring ID3 tag versions and album art format/resolution into this topic, but I'm only asking about the audio quality settings in MP3 formats.

SURPRISINGLY ENOUGH, even professional websites that sell MP3 format downloads are using 3.96 version of LAME and encode in NORMAL STEREO, which is odd (bleep.com, beatport.com, etc).

Am I correct or incorrect about my conclusions?
I base all my findings on what I've read here.

As far as I know, 320kbps CBR LAME3.97 Joint-Stereo MP3, ripped with EAC is the best audio quality file one can produce for the MP3 format. If I am erroneous in any way, please share your findings and insight.

My take on your questions:

Ripper - EAC is one of the best along with DMC and foobar2000. I use foobar2000 because it's easier to use than EAC and unlike DMC, it's free.

Ripper settings - As long as you use secure mode with EAC, you should be fine. Test and copy with matching CRCs is fine as well. Only difference is the speed of extraction depending on the CD and drive.

Encoder - LAME 3.97, since it is currently recommended, is the safe choice, since it is most tested. I myself use 3.98b4 not because I can hear the difference but because I like the higher version number.

Encoder settings: preset insane IS the strongest setting for MP3. Consider lossless if that's not enough for you.

Generally you are correct. With every year I disklike standards more and more- do your own testing. I would say 3.98 might be better in terms of bugs fixes and quality. EAC might be better only on damaged cd's. Foobar ripper is just as good for me and never had a problem with cdparanoia.

Anyway, there has been bloody battles for many years regarding mp3 standards etc. Just remember two things:

-The reason why people still use STEREO, CBR etc is that they each try to outdo each other and create their own standards. They believe in a single magic end-to-all setting.

They don't know or care that:

- Mp3 is flawed in terms of quality and it can never be fixed as long as the format follows the mpeg1 layer 3 specs. No matter which setting or encoder, mp3 is abxable on many samples. There is no 'safe' encoder or setting - I am happy with CBR 140k FHG on my portable.

Ripper accuracy/recovery of good data in the event of an error (in order):
1. dBpowerAMP R12 Reference with drives that support C2 pointers
2. PlexTools (provided you have a true Plextor drive and aren't having trouble with conflicting hardware and software)
3. EAC
dBpowerAMP R12 w/Powerpack, foobar2000, CDex, Easy CD-DA extractor or any non-windows program: all inferior to the first three, but still better than other programs not listed (I think EAC just edges out dbpa R12 w/powerpack because of the way it deals with re-reads, but could go the other way depending on the disc and drive).

As I've said elsewhere, as the DAE quality of your drive improves, the differences between the programs shrink. With discs in excellent condition, which program you use does not matter; so long as it has a true secure mode or is capable of performing test and copy.

Ripper Settings:
Depends on the drive and the state of the disc being ripped. There is no silver bullet.

Finally, AccurateRip is far and away the best guarantee that your disc was ripped correctly but only if it has accurate data for the its exact pressing. Next best guarantee is to get matching results between two different drives (better make sure the offsets are calibrated to one another!). Other ways to improve your confidence that you didn't suffer from consistent errors is to get matching results using different ripping algorithms.

Yes, these are generalizations based on personal experience as well as trends I've seen in various tests (many are well documented either here on this forum or over at cdfreaks); YMMV, of course!

Pio2001, spath, spoon, JeanLuc, you've been involved in quite a few detailed and technical(!) discussions going several years back, what do you guys think?

Ripping:
dbpowerAmp R12 for me too: fast and secure - AccurateRip is great as is secure mode.
But I had one or two CDs I could not read using dbpowerAmp which were no problem to EAC. So I keep EAC as well but usually don't need it.

Lame encoding:
I'd use 3.98b4 which has fixed bugs that were hidden in previous versions including 3.97 for several years.
3.97 suffers from serious and not extremely rare tonal distortions like in Birds. These are overcome with 3.98. I started worrying about these things when listening to the seriously distorted trumpet sample with 3.97 (b1 then). trumpet was improved in later 3.97 versions and is overcome totally with 3.98.
There is still a tremolo issue with both 3.97 and 3.98 when using VBR. Other than that 3.98's VBR mode seems to be more robust than that of 3.97.
In case the eig sample behavior can be at least partially generalized towards pre-echo behavior 3.98's pre-echo behavior has improved as well.

The beforementioned problems are rather severe and to the extend they are audible they're pretty Lame-specific. However as you consider using CBR320: even though 3.97b1 was very poor with trumpet, at CBR320 this was not annoying (though easily abxable). In general a brute force strategy like using CBR320 is able to significantly soften even severe encoder problems. To me this is mainly an argument for using 3.98. 3.98 has overcome serius problems 3.97 has, and for the not-so-much-tested-as-3.97 argument a high bitrate to me is enough for me to feel secure.
As for joint vs. plain stereo: I don't see any experience that stands against the basically superior joint stereo mode.

Using very high bitrate IMO it's sufficient to use CBR 256 or ABR ~260. You save ~ 20% of file size against CBR 320 while not having to fear a really lower quality. Because of 3.98's improved VBR behavior -V1 is worth considering too IMO. From previous 3.98 versions I know -V1 can improve quality against -V2 while -V0 isn't worth considering against -V1.

Exactly what I found too.

>Depends on the drive and the state of the disc being ripped.

A relatively high percentage of secure rippers are configured incorrectly!, for our next version we will default secure mode to a sizable cache value, it will have no downsides with our ripping method and will ensure secure ripping out of the box so to speak.

>There is no silver bullet.

If only there were better following of the MMC drafts, the silver bullet for us would be by default to enable c2, if unsupported in drives it would all be 0, ie off, but there is such a variety of drives not all would even work in this simple respect.

Will you be changing the pass/fail criteria for re-reads without C2 pointers as well as the pass/fail criteria for re-reads with C2 pointers? From what I remember, 10 matches out of 1000 tries without C2 wasn't exactly a good idea, and with my PX-716A, I could have saved a lot of time and still gotten a accurate rips if re-reads would have ended as soon as data came along that wasn't dropped by C2 information (rather than waiting around for another 9 matches). The option of only re-ripping bad sectors and combining them with good sectors from previous rips (again, determined by C2 information) would have really improved matters with this drive as well.

...but then again, Andre's philosophy might be relevant: is it worth all the additional complication/development just to save one or two discs out of hundreds?

halb27, I thank you for your thoughts.
although 3.98 is superior to 3.97 in several areas, one could use CBR320 brute force to soften 3.97's problems.

A question: which sample can you easily ABX with CBR320 LAME3.97? I am interested to hear what a LAME3.97 CBR320 cannot easily make transparent.

For the above comment, I agree in your application of lower bit rates to save space without realistically affecting quality. However, I have a somewhat radical view of sound formats in regard to the DRM movement. The following may be a bit off topic, but I think you'd find it very interesting:

I personally feel that DRM will be shoved down our throats in future years. Protected WMAs, DRM AACs, etc. There may even be completely new future formats (5.1, 8.2 channels) that use DRM to protect people from burning discs, uploading to their players, and using information in the file to track the digital movement (who bought it, when, where it went, etc). I fear this.

The only 2 truths we can all agree on completely would be:
1) the human ear will not be capable of hearing better in 5 years
2) the size of our hard drives and speed of our internet will be faster in 5 years

Hence, I have no qualms at all with using CBR320 and FLAC (I simply use CBRMP3 because it makes life easier for MP3-DJing, uploading to my iPod, etc).

I believe the issue of file size will be completely moot in 2-3 years.
With a 750gb hard drive and 900kb/sec download speed, I never worry whether a file is 220ish kbps VBR or 320CBR. I might as well take the 320CBR...

However, if LAME still cannot encode some audio as completely transparent at 320CBR (as halb was saying), that is why I also use FLAC.

The day will come when downloading a FLAC file will take about 5-10 seconds and our iPods (and other devices) can hold our entire libraries in FLAC, portably. But, as I must use MP3 for software reasons and compatibility with browsers, cell phones, applications, iPods, etc, I try to stick to 320, even though most of my library has been encoded -V1. What are your thoughts about 320 MP3 vs future DRM formats?

I also worry programs like iTunes, etc. will STOP coming with MP3 encoders, and only FutureDRM formats. Perhaps a revolution of people ripping everything in 320kbps lame 4.xx b/c they don't like DRM? We'll see...

320 CBR mp3 ?? . Try 160k AAC. Quality on par or better than 320k mp3. Try Hihat.wav, EIG, elctrodrip.. If you can't hear it you can't hear pre echo. Mp3 is useless above 256k IMO. The other encoders are not worth using above 256 k either, better at 160~190k. If your DJ gear supports MP2 then try 320k cbr.

I find it very hard to believe that a 160kbps AAC could possibly be BETTER than 320kbps MP3.
HydrogenAudio's own Wiki describes this.
What makes you say 160 AAC is better than 320 MP3?

I cannot abx over 190 kbps VBR mp3, but nonetheless, I am still interested in abxing samples that people say 320kbps MP3 lame 3.97 cannot properly do (easily differentiable from original CD). Anyone?

Just tried a few samples.
eig is easily abxable as was said by shadowking already. However I personally wouldn't call it real annoying (I'm not very sensitive to pre-echo).
I tried typical tonal problem samples that are easily abxable at moderate bitrate using 3.97, and yes, brute force CBR 320 helps. I couldn't abx them in my quick test.

But IMO you should do some listening tests on your own, for instance with eig. After all it's your choice whether deviations from the original bother you or not.
May be you find out going the shadowking's proposal way is worth while.

Actually, my terrible ears are of no importance...
I am interested in distribution of MP3 files for pay, such as bleep.com and beatport.com use - however, they are using lame 3.96 and encode in Normal Stereo........

Of course, the files are all CBR320.

In 4 or 5 years, it will be the difference between a DSL user downloading a picture that is 15k or 28k. Our transfer speeds will be so fast & the hard drive capacities so huge that that we won't care - ?

Lame 3.97 still has problems, as halb is saying.
Even at 320kbps, LAME3.97 cannot reproduce some sounds well enough to be transparent?
My...

iTunes PLUS just came out, offering you that same $0.99 track for $1.30, but 256kbps AAC instead of 128kbps AAC.

Bleep charged $1.49 or more for 320kbps MP3, but they use Lame3.96 and encode in Normal Stereo...

Even the experts say one way, the big business does not listen??

here's a vote for dBpoweramp R12 reference!

Why do sites like:

www.bleep.com
www.beatport.com

use LAME 3.96?
And even though they encode CBR320, they use Normal Stereo mode.
That makes no sense, right?

Maybe they started mass encoding of their mp3s around the time 3.96 was released, or maybe they didn't bother to upgrade to a newer LAME version, or maybe the are using a 3rd party program that was shipped together with an older LAME dll and maybe they don't even have control over all LAME options and simply using some default settings in some program.

Did you try to ask them directly?


Regarding stereo mode: it has always been the strongest misunderstanding among vast majority of mp3 users that "stereo" should always be better than "joint stereo".
Sometimes -- yes, maybe, depending on the encoder/version/options.
With LAME it is the other way around.

What I have been told in the past (by a certain elitist group) is that C2, even if supported by the drive, is unreliable and to not enable it at all. Is this still true, even for recent modern drives? Could you shed more light on the faulty C2 issue?

Thanks

That is true for EAC (although in my extensive tests, ripping with EAC and c2 enabled on a few drives was better than not).

We designed our ripper, so any implementation of c2, no matter how bad would help the ripping process.

Using 3.96 is no real problem. It was only about 6 month ago that it was still the recommended version here. They probably just encoded a library with what was best at the time.

As for the joint-stereo versus normal stereo (more correctly MS stereo versus LR stereo) it's still an issue with lots of misinformation out there. Maybe they went LR stereo so as not to alienate the many mis-informed who are still skeptical (and who usually who dont understand) about MS Stereo. In any case at very high bit rates the need for joint stereo's efficiency improvments are not so important.

Ahem, Lame's joint-stereo mode uses both MS and LR frames.