[blessingx]

Quote:

Originally posted by MacDEF This is a bit misleading. The fact of the matter is that if you take a joint stereo track and a stereo track that are otherwise identically encoded, the stereo track will sound significantly better -- but the stereo file will also be bigger.

This is a bit misleading

Doesn't it depend on the encoder in respect to how say VBR (or ABR) uses file size? For the hell of it, just did some tests. Encoded "One" from the Magnolia soundtrack. With Lame encoder set to alt preset extreme I got 3,683,923 bytes. Doing the same but forcing stereo I got 4,013,385. Significant increase. However if I used iTunes Fraunhofer encoder, at 160 VBR (no preset options) the stereo came out 3,771,631, while the joint actually came out slight larger 3,826,488. Leading me to believe the Fraunhofer at least aimed at the same size, but shared the space differently.
Also what recent stereo/joint stereo tests have shown "significantly better" sound? Just curious. Some new formats like AAC don't even give the option to force stereo.

 

[MTRH]

As always, LAME MP3 encoder/decoder should be used.

 

[PodMan]

Quote:

This is a bit misleading. The fact of the matter is that if you take a joint stereo track and a stereo track that are otherwise identically encoded, the stereo track will sound significantly better -- but the stereo file will also be bigger.

But you would still need joint stereo to get maximum quality from mp3, which is a 320kbps joint stereo mp3 or LAME alt preset insane. You simply cannot get bigger or better than that for mp3.

Quote:

I doubt the "engineers" know much about audio quality, they just base everything on paper.[/B]

I am under the impression that LAME development is pushed forward by A/B testing conducted by many active members of the hydrogenaudio board. LAME didn't evolve into what it is today by "paper specifications".

This is the latest thread regarding LAME 3.94alpha13 development. Notice the request by the developers for user feedback and the user feedback based on listening tests.


http://www.hydrogenaudio.org/index.p...T&f=16&t=8520&

Probably everyone at hydrogenaudio will agree that joint stereo is the mode to use for LAME in all cases (except maybe for compatibility reasons). LAME was internally optimized for joint stereo after all. Joint stereo implementation may vary for other MP3 encoders such as Xing or fhg, which may have a poor implementation of it, causing normal stereo to sound a lot better.

 

[blessingx]

Just to verify what Podman said. Just ripped Orphan Girl by Gillian Welch. Api and api -ms. It came out the exact some byte count of 9,495,298. Either forced stereo was ignored (though it read in my terminal window), or we have to assume the the same space was used duplicating information, thus lowering the bitrate available for each channel for everything else... thus likely lowering the overall quality.

 

[MacDEF]

Quote:

Originally posted by PodMan Quote: This is a bit misleading. The fact of the matter is that if you take a joint stereo track and a stereo track that are otherwise identically encoded, the stereo track will sound significantly better -- but the stereo file will also be bigger. But you would still need joint stereo to get maximum quality from mp3, which is a 320kbps joint stereo mp3 or LAME alt preset insane. You simply cannot get bigger or better than that for mp3.

I don't think you understand what I'm saying. Joint stereo gives you "better" sound ONLY because it frees up bits to be used for non-identical data. In other words, if parts of the signal are "identical" in the left and right channels, it only stores one "copy" instead of two (and then in the decoding process, this single "copy" is restored to both the left and the right channels). By doing this, you're freeing up space for OTHER parts of the signal, resulting in a compressed file that doesn't sound as "lossy."

At identical bitrates -- meaning the same amount of data per second -- a joint stereo MP3 will sound better than one that isn't joint stereo because fewer of those bits are "wasted" encoding identical data twice. I was simply pointing out that if you disregard bitrates and compared a joint stereo MP3 with a standard stereo MP3, and EVERYTHING else was encoded the same, the stardard stereo file will sound better, but will also be a lot bigger.

 

[Audio-Me]

I doubt the people at hydrogenaudio have the right "ears" for me to accept as good judgement, especially considering I disregard most of head-fi members hearing. At low bitrates I'm sure it helps, but for quality at 320kbps, the statement that joint stereo is better is absurd to me.

 

[PodMan]

Quote:

Originally posted by Audio&Me I doubt the people at hydrogenaudio have the right "ears" for me to accept as good judgement, especially considering I disregard most of head-fi members hearing. At low bitrates I'm sure it helps, but for quality at 320kbps, the statement that joint stereo is better is absurd to me.

somehow I knew this was going to be the response. Have you personally done A/B listening tests to support your claims? The people at hydrogen audio have audibly confirmed that joint stereo sounds better than full stereo in all cases. Better because it allows more bandwidth to encode non-joint channel frequencies. That's why joint stereo is the default encoding mode. Refer to my first link for the discussion. There are many more, this question comes up countless times over there.

Until you can find a physical sample to pick out that full stereo sounds better than joint stereo with LAME, what you are saying is completely based on paper and not physical evidence. If you do find such a sample, please post it, I'm sure it would help further LAME development.

And I wouldn't overgeneralize and say that everyone on that board do not have the ears to pick out flaws. Results are A/Bed with the original wav file and then a concensus is taken. Certainly, some people out of the majority will have the ears and equipment. Especially when they are comparing audible flaws with the original wav side by side.

The stance of LAME development has always been driven by what you can hear, not based on a belief that full stereo sounds better than joint stereo just on principle. I am aware that the a version of the fhg professional encoder has terrible joint stereo, this may be why joint stereo has had a bad rep amongst some users.

EDIT: Also keep in mind that even at 320kbps, there is still not enough bitrate to go around to encode problematic samples. That is why joint stereo must be used. There are numerous samples available that will not encode well even with the insane setting of LAME. Everyone who develops LAME will agree that mp3 will never be fully transparent based on its limitations. In fact, many will advise you to move to other codecs for transparency.

 

[Audio-Me]

You know, many people just say what they want to believe, not what they actually hear. I am very distrustful when it comes to peoples findings, they are inately liars and/or completely inaccurate as in most of them can't even acknowledge what they notice anyway let alone register what their ears pick up.

I may be extremely stubborn, but I am not closed minded, I will download the latest LAME and compare --alt preset extreme vs --b320 -q0 -k -ms
In the meantime, check out my earlier impressions:
http://www4.head-fi.org/forums/showt...threadid=11289

Quote:

there is still not enough bitrate to go around to encode problematic samples. That is why joint stereo must be used

Are you implying that js uses higher than 320kbps bitrate? This is the reason I use a superior format such as mpc, which I also force stereo in. I've found a good setting which results around 450kbps, I sitll notice a difference between the original wav, but I still need to think about file size a little so going to 600+kbps is kinda crazy.

 

[PodMan]

oh no, just that joint stereo allows better allocation of that 320kbps bitrate limit. I myself don't rely on mp3 much anymore ever since I sold my ipod. I still use it occassionally if I want music played on the computer though.

If you can find a sample, that'd be great. Because LAME is still in continual development and any more testing would be helpful. LAME 3.93 was released without much public testing to rather disastrous results.

 

[xcalibur]

hey guys..thanks for all the info..and do let me know how the forced stereo at 320 CBR compares to insane...i'm very interested to know the results of it...i would have to say i think it does seem weird to me that joint stereo is better than stereo...

 

[blessingx]

BTW, something for this debate- some of the newer formats like aac/mp4 eliminated the option for 'stereo' completely as far as I know.

Also from this page titled Popular MISCONCEPTIONS about mp3-

mp3 192 kbps CBR stereo is the best and optimal quality/filesize option

NOT TRUE! This is simple. Such opinions were spread after many users realized that 128 kbps mp3s were not perfect and tried the existing encoders with higher bit-rates (CBR and VBR). But at that time implementations of VBR and joint stereo algorithms were far from perfect too (especially Xing encoders which still do not shine, but also Fraunhofer VBR -- refer to ff123's evaluation). So this misconception arose, saying that only CBR full stereo at high enough bit-rate can provide good quality.
Now, as we have Lame developed virtually to the edge, with tuned-up ABR and VBR alt-presets and smart joint stereo, there is no good reason to use either CBR or full stereo any more. It's just a waste of bits. Standard VBR alt-preset averages at the bit-rates around 192, and the sound quality is better than CBR 192, and better than any other encoders at similar bitrates. It was proven with numerous listening tests by many (those with better ears than average!) of the members of HydrogenAudio community.

mp3 320 kbps CBR Stereo is the upper limit quality you can get with mp3

Not exactly true. Yes, 320 kbps is the maximum possible bit-rate according to mp3 standard. But Joint Stereo can be effectively utilized also at 320 kbps, thus providing more bits for encoding audio content. If an audio source has low stereo separation, encoding it in full stereo 320 kbps is simply inefficient.

VBR makes mp3 sound awful.

NOT TRUE! A properly designed VBR algorithm can only improve sound quality, not wasting bits for "easy-to-encode" pieces but saving them for "difficult" ones.
The spread of the misconception was probably caused by the older buggy implementations of VBR in FhG encoders, and too imperfect Xing VBR. There is no problem of any kind using VBR with recent Lame encoder, because its VBR algorithms were tuned considerably to handle even "difficult" samples. Please refer to HydrogenAudio mp3 fora for more up-to-date information on VBR alt-presets.

Joint Stereo degrades the stereo image and overall sound quality and kills off "surround" information.

NOT TRUE! The state-of-the-art encoders (Lame, Musepack, PsyTEL AACenc, Ogg-Vorbis) exploit the so called smart joint stereo algorithms, which practically never damage stereo image at higher bit-rates. Please refer to this more technical explanations by the Lame alt-preset developer 'Dibrom' and also to this discussion of safe js.
Even more so with MPC, as Frank Klemm says here. Restricting the mpc encoder to full stereo will actually degrade overall quality.
There can be certain stereo problems with Ogg-Vorbis unless Vorbis lossless stereo-coupling is used (which is in fact a kind of smart joint stereo).

Regarding surround: this is a somewhat contraversial topic. In theory, lossy encoders are not obliged to take care of surround data, because surround stereo systems perform intensive digital sound processing (DSP) to make the sound "surround", and lossy compression is not intended to be used in any DSP. However, in practice, a proper joint stereo implementation is usually capable of preserving surround info.
Please refer to this discussion, and this discussion of surround in musepack.

NOTE: At lower bit-rates, all encoders use a more "aggressive" joint stereo modes. If you plan to listen to encoded files on a surround sound system, then you might need a higher bitrate preset: --alt-preset extreme or insane for mp3 (these presets make use of "safe joint stereo" switch); --quality 6 (--xtreme) or higher for musepack.
Those presets activate less agressive joint stereo modes, preserving the signal components that are not audible on an ordinary stereo system, but might be audible on a surround system.

 

[xcalibur]

so which encoding rate do you use blessingx? i'm thinking of which rate to rip all my cds for my new ipod...i don't have that many songs so i was thinking of --alt insane but i'm now even considering wav! yup..that's right..i know it might be overkill...but if i have the space i might as well...unless the difference isn't that much...running on .wav will drain my battery much quicker... -alt insane would be around 1/3 or 1/4 the size of wav..is that right?

 

[blessingx]

Quote:

Originally posted by xcalibur so which encoding rate do you use blessingx? i'm thinking of which rate to rip all my cds for my new ipod...i don't have that many songs so i was thinking of --alt insane but i'm now even considering wav! yup..that's right..i know it might be overkill...but if i have the space i might as well...unless the difference isn't that much...running on .wav will drain my battery much quicker... -alt insane would be around 1/3 or 1/4 the size of wav..is that right?

Which equipment are you using? If the Sony's and Senn's phones in your sig, it's likely overkill at this point. Question is how often do you want to encode your CD's, and if you plan on upgrading equipment? I swear over the last year (before discovering this site) I've done a good chunk of my CD's in 160 cbr Fraunhofer (iTunes) mp3's, lame r3mix, lame extreme, a little lame insane, and the last 10 albums I've used Toast Extractor and iTunes 320 kps aac. If you have space go wav/aiff. Not hurting anything and it's the quickest to copy/extract (and you can always compress further in the future). Since I have a 5 gig iPod, space is a bit of a concern (would like around 40-50 albums on it), so I'm going with the highest possible setting (at the moment with Mac iPod compatibility 320 aac). Though to be honest, with my ears and equipment, transparency with CD's hit at lame extreme. Just would like to hold off on encoding again in the near future for new equipment. As for space comparisons, the last album I did, Miles Davis' Kind Of Blue, extracted (aiff) was 556.9 meg, 320 aac (lame insane should be similar) was 126.9. So we're talking between a quarter and a fifth in size. Also when testing lame standard v. lame extreme I found the latter increased file size by a third.

Then there's the pychological element of knowing what you're hearing is uncompressed, so if there's a problem it's not the file. Just need new phones, amp, etc. Course your wallet will more likely suffer with that scenario.

And by the way, I think the Sony MDR-7506(&v6)/iPod combo is great. I couldn't believe the first time I heard them together. Actually considering selling the much beloved Koss 35's to pick back up a pair of v6's. Keep thinking of the sound.

Good luck. Post what you decide.

 

[clintf]

Just to add my couple cents, after a lot of ABX testing for LAME 3.94alpha listening tests on Hydrogen Audio I can say that the Joint-Stereo mode in the --alt-preset VBR presets are optimal ("--nsmsfix 1.5" IIRC). I haven't heard of anyone proving any flaws tied to the --alt-preset VBR joint stereo settings via ABX tests.

As for the default Joint Stereo modes in settings such as --r3mix and the --alt-preset ABRs, there's one sample ("creeking", do a search at HA for it) that trips them up really bad which is pretty clearly related to Joint Stereo. So the Stereo vs. Joint-Stereo argument isn't so clear with the exception of the --alt-preset VBR presets (due to the lowered --nsmsfix value) which are excellent and one shouldn't worry about noticeable artifacts or general quality loss caused by Joint Stereo. Stereo Phase Information is completely intact and you will allow the encoder make MUCH more efficient bit useage with Joint Stereo which in return will cut down on truely annoying artifacts which may very well be there when you limit the encoder with Forced Stereo. With Stereo you're relying on the bitrate per frame, the bit reservoir and a tiny amount of bits that can be shared between the channels. With Joint Stereo you're allowing the channels to share mid frames (on tracks without a lot of stereo separation the efficiency bonus can be huge), bit reservoir, and presumably the channels can borrow a little from each other. Joint Stereo is what lossy encoding is about and forced Stereo doesn't make sense at all. And forced Stereo should definitely not be used for VBR, Lame VBR was designed with Joint Stereo in mind.

 

[xcalibur]

argh..this is pretty confusing..anyway...looks like AAC development for the PC is moving along pretty well...maybe 320 AAC is the way to go...or wait..AAC doesn't have max out at 320 right?