Your opinions about m MP3 settings Options

[Bourti]

Hi all,

Well, I would like to have your opinions about my MP3 settings. First of all, my audio converter is Xrecode (see here) that is excellent and free. Considering I'm a musician, I wanted to have the best balance between quality and size (with a preference for the first one).

Here my settings:
VBR (new method)
Min Birate: 256
Stereo (not Joint Stereo)
Quality: Highest (0)


Thanks for your comments ,
Cheers

This post has been edited by Bourti: Mar 30 2009, 12:35

[pdq]

Hi all,

Well, I would like to have your opinions about my MP3 settings. First of all, my audio converter is Xrecode (see here) that is excellent and free. Considering I'm a musician, I wanted to have the best balance between quality and size (with a preference for the first one).

Here my settings:
VBR (new method)

Depends...which version of LAME are you using?

Min Birate: 256

No

Stereo (not Joint Stereo)

No

Quality: Highest (0)

No

Thanks for your comments ,
Cheers

[Bourti]

Hi,

Xrecode uses the following version of Lame: 3.98.2

Could you be more "specific" about my settings?

Tia

[Remedial Sound]

Use one of the VBR presets ("-V"), they exist so ordinary blokes like you and I don't need to futz with swtiches, etc.

I suggest reading this first:
http://wiki.hydrogenaudio.org/index.php?title=LAME

[pdq]

VBR new is the default in 3.98.2

By setting a minimum bitrate of 256 kbps you have negated any advantage of VBR - you would be better off then with CBR (which is less efficient but makes better use of the bit rteservoir)

Joint stereo provides better qualitu/smaller file size than forced stereo

The Q settings of 0 and 1 are identical to the default setting of 2

[Bourti]

Hi,

I had already read it (moreover very useful!)
I wanted to use the Stereo mode and not the Joint Stereo. It seems that below 192, the Joint Stereo mode is useful but not above: that's the reason why I set the minimal bitrate to 256. Finally, I tested with my own records.

[Slipstreem]

Considering I'm a musician...

It makes no difference what you are. You still have human ears.

I wanted to have the best balance between quality and size (with a preference for the first one).

It has to be a standard (ie, a non-modified) VBR setting to satisfy both criteria.

Here my settings:
VBR (new method)

That's the default.

Min Birate: 256

As said previously, you're stopping VBR from being VBR as defined by LAME internally. If the software forces you to choose a minimum bitrate value, choose the default which is 32Kbps.

Stereo (not Joint Stereo)

Joint Stereo is a lossless process with LAME and is advantageous when encoding stereo source material at any bitrate. All you achieve by not using it with LAME is wasted bits. It is also the default.

Quality: Highest (0)

As also said previously, all values of q from 0 through 2 are now identically mapped internally. A value of 2 (which equals 1 or 0) is the default.

If the software you're currently using doesn't allow selection of -V values for VBR directly, dump it and use one that does. If you think there is an audible difference between, say, -V2 and what you're doing now, carry out a blind listening (ABX) test between the lossless source and the encoded file.

Thanks for your comments ,
Cheers

You're welcome.

Cheers, Slipstreem.

This post has been edited by Slipstreem: Mar 30 2009, 14:22

[shadowking]

For quality preference over size, I suggest V1 (200..260k), V0 (220..290k) even b320.. Don't bother at all with anything lower or any joint stereo myths.

[Slipstreem]

For quality preference over size, I suggest V1 (200..260k), V0 (220..290k) even b320.. Don't bother at all with anything lower...

As quality is in the ear of the beholder, I don't think that's a safe suggestion to make personally if efficiency is still to play a relevant part in the equation.

The OP may not be able to distinguish, say, -V3 (~175Kbps) from a lossless source for all we know. If that turned out to be the case, any additional bits used by a higher setting would be totally wasted and would skew the required quality versus required typical bitrate ratio heavily and compromise efficiency with no gain whatsoever in perceived quality.

I'd like to see the OP being encouraged to carry out ABX tests in Foobar2000 personally before he goes wasting his bits unnecessarily. At least it would encourage him to learn with a scientific approach rather than just making blind assumptions.

Cheers, Slipstreem.

This post has been edited by Slipstreem: Mar 30 2009, 14:54

[Sound-of-Muzak]

I suggest V0 encoded with LAME in foobar2000 for encoding. Put this in the converter output settings for typical tagging %album artist% \ %album% \ %tracknumber% %title%

Honestly, space isn't really an issue if you have an average sized collection. I ripped all my CD's to FLAC and came in at like 56 GB (about 175CDs), converted to mp3 V0 in foobar and my collection is only 16GB. You may not be able to tell the difference between V0 and V2 (I can't), but if you're obsessive like me, you'll encode to the highest mp3 VBR setting, even though you can't tell the difference. If you have a massive collection, I suggest V2 or V5, but an average collection like mine, encoded in V0, will fit on a 20GB ipod. And if you ripped to FLAC, you can always re-encode.

[Bourti]

Thank you all for your valuable help

Joint Stereo is a lossless process with LAME and is advantageous when encoding stereo source material at any bitrate. All you achieve by not using it with LAME is wasted bits. It is also the default.

Do you mean that the Stereo mode implemented in LAME (contrary to the Joint Stereo mode) is always lossy?


Here are the settings I'm gonna use in Xrecode:




I'm gonna carry out some listening tests and I'll keep you informed,
Cheers

This post has been edited by Bourti: Mar 30 2009, 15:25

[Slipstreem]

Do you mean that the Stereo mode implemented in LAME (contrary to the Joint Stereo mode) is always lossy?

No. What I'm saying is that Joint Stereo is, in itself, a lossless process and will always save bits when compared to Stereo. This translates into smaller files for a given -V setting in VBR, or to higher quality for a given fixed bitrate in CBR. It's just more efficient at no cost to quality.

I'm gonna carry out some listening tests and I'll keep you informed.

They will need to be blind comparison tests against the lossless source to be truly meaningful. Foobar2000 has a blind listening test (ABX) comparison tool. Blind comparisons are the only way to eliminate placebo effect.

I'd highly recommend using Foobar2000 for your MP3 conversions too as it's guaranteed not to be fiddling with any potentially damaging settings behind your back when using the supplied presets. If you need any further advice on how to set up and use the ABX comparator tool in Foobar2000, just ask. We're here to help.

Cheers, Slipstreem.

[sld]

Do you mean that the Stereo mode implemented in LAME (contrary to the Joint Stereo mode) is always lossy?

Stereo mode is what its name means, audio that is in two channels is encoded exactly like that. Joint Stereo is poorly named, because what it does is re-encode the left/right channels into mid/side only if the encoder determines that the process is lossless. Otherwise, if the left/right channels are always different enough, you will still get pure Stereo when you set Joint Stereo mode.

In summary, the setting is dynamic. I hope your worries about JS are authoritatively dispelled!

This post has been edited by sld: Mar 30 2009, 15:49

[uart]

Thank you all for your valuable help
Do you mean that the Stereo mode implemented in LAME (contrary to the Joint Stereo mode) is always lossy?

No they're both lossless (actually "stereo" mode is not really a process but if you like you could call it lossless). The difference is that joint stereo encodes via a lossless transform, L/R to M/S, which can give lower bitrate for the same quality or higher quality for the same bitrate. There is no reason not to use joint stereo, it's very safe.

Here are the settings I'm gonna use in Xrecode:

Yes good, that's generally the best way to do it. Just use the "pre-defined" setting at the top and leave it at joint-stereo and it will be as good as you'll get. You might want to compare "pre-defined" levels of V0, V1 and V2 and see if you can tell any difference. Few people can notice any improvement at settings above V2.

[Alex B]

This has been said over and over again, but it might not be clear to everyone that joint stereo* does not affect channel separation. It does not combine part of the information so that the stereo effect would be lessened.

The M/S stereo coding process itself is lossless. All usual lossless codecs use this kind of joint stereo automatically. When a lossy format is used it is possible that some artifacts may exist that may not be present when the files are encoded without joint stereo, but that is not typical. More often the opposite is true.

I once did an ABX test in which -V 2 (using -m j) was better than -V 2 -m s: http://www.hydrogenaudio.org/forums/index....mp;#entry490936

Actually that whole thread is good reading for anyone who wants to understand LAME's joint stereo better.

EDIT
*as LAME uses it. There are also other "joint stereo" techniques that can combine channel infomation. Some of them are explained here: http://en.wikipedia.org/wiki/Joint_stereo

This post has been edited by Alex B: Mar 30 2009, 16:55

[halb27]

Some more words on joint stereo emphasizing the 'joint' part:
The stereo signal can be represented by the left and right channel signal (L/R representation) as it comes for instance from CD, or it can be represented by the equivalent mid and side signal (M/S representation) as it comes for instance from FM radio. The different representations can easily and without problem be converted into each other (apart from a scale factor the mid signal is the sum and the side signal is the difference of the L/R signals).
Joint stereo means that the encoder which works on blocks (frames) choses for each block whatever representation is more appropriate. For a solo singer singer or a solo instrument the M/S representation is expected to be more effective for encoding than the L/R representation, for a big orchestra it may be the other way around.
Joint stereo just meets intelligent decisions in this respect.
While the L/R and M/S representations of the stereo signal are equivalent and losslessly transformable into each other, in the end we get an mp3 encoding of either the L/R or the M/S representation. These are not equivalent, it is not totally correct to talk about the entire process as a lossless procedure. Moreover switching between L/R and M/S representation can be an issue. That's all theory however, epecially Lame does take good care of joint stereo, and it's better to be able to move more bits into the more appropriate of the L/R or M/S representations than having no space for bits because of a fear of theoretical problems. Joint stereo is highly recommended.

This post has been edited by halb27: Mar 30 2009, 18:23

[[JAZ]]

Heh... If "joint stereo" is a bad name .. "canaux joints" is even worse! (although it's a plain translation of the former).

The others have already given broad and detailed information about it.

[kornchild2002]

The OP may not be able to distinguish, say, -V3 (~175Kbps) from a lossless source for all we know. If that turned out to be the case, any additional bits used by a higher setting would be totally wasted and would skew the required quality versus required typical bitrate ratio heavily and compromise efficiency with no gain whatsoever in perceived quality.

Now that the misconceived notions regarding Joint Stereo have been shot down, I think this should be another point made to the OP. -V 0 is more than likely overkill for the OP unless they can conduct blind ABX tests showing otherwise. Even when space constraints aren't "really an issue," the encoder setting that the OP determines to be transparent (through blind ABX testing) should be used. We don't know the OP's library size, the amount of hard drive space they have, the amount of storage space that their portable has (that is if they even have a portable device), or the number of songs that they want to carry around. For example, I have a 120GB iPod classic and could have stuck with -V 0, iTunes AAC at 256kbps VBR, or Nero at -q0.65 simply because I have such a large capacity player. However, I can't use bitrate settings that high because my portable player would be filled with my lossy music only not allowing for other media. I really think that the OP should conduct a blind ABX test to determine what setting is right for them.

Honestly, space isn't really an issue if you have an average sized collection.

It depends on what you define as a "average sized collection." I have a bunch of friends whose libraries take up about 15GB of space when using the 128kbps VBR setting with the iTunes AAC encoder. Their libraries would increase to nearly 30GB when using -V 0. I don't think that there is a definition for the average size (as in amount of songs) of libraries. I also know many other people who have 15,000+ songs in their collections. So I have friends with about 3000 songs and ones with 17000 songs.

[timcupery]

Honestly, space isn't really an issue if you have an average sized collection.

It depends on what you define as a "average sized collection." I have a bunch of friends whose libraries take up about 15GB of space when using the 128kbps VBR setting with the iTunes AAC encoder. Their libraries would increase to nearly 30GB when using -V 0. I don't think that there is a definition for the average size (as in amount of songs) of libraries. I also know many other people who have 15,000+ songs in their collections. So I have friends with about 3000 songs and ones with 17000 songs.

My collection is over 80gb, most of it mp3 at V4 or higher quality. And I have friends with much larger collections. But most people on HA probably know people with enormous collections. I was just surprised that your example was people with 15gb collections.

Defining an "average sized collection" would mainly be an issue of defining the population we're interested in. Anybody who owns their own computer? Or people who use their computer as the main source for listening to music? Anybody who uses a DAP? That's the tricky part. Once we have a population definition, the measurement is easy

[Bourti]

Hi all,

I noticed there were several presets and I would like to know what they mean. In Xrecode, you have several choices, for instance:

V0 up to V9


Extreme


FAST EXTREME


VERYHIGH_QUALITY



I made the following test: I encoded from WAV to MP3 with three different presets:
- Extreme
- Insane
- V0

Here are the results:




You can notice that the sizes are different between the V0 and the Extreme presets ... but I thought it was actually the same setting but with a different name!
If not, what are the differences?)
And what about the sound quality?

Thanks for you details,
Cheers


P.S: I currently use the V0 preset.

This post has been edited by Bourti: Apr 5 2009, 14:31

[shadowking]

Extreme is vbr-old and V0 is vbr-new. 'fast extreme' should be V0

[Bourti]

Ok, thanks for the tip!

I guess you would recommend to use the V0 preset (as recommended above).

[Slipstreem]

If you need to be encoding at that high a level of quality, yes. If you've carried out the ABX tests as suggested above and don't need -V0, no.

Cheers, Slipstreem.

[Bourti]

Hi Slipstream,

I've carried out some tests, I'll post my results.

Cheers

This post has been edited by Bourti: Apr 6 2009, 08:36