How come no one recommends the following as the "best/purest" setting for LAME? Joint stereo just doesn't cut it for some music tracks. Sure the music may sound a little better in some areas because of the few extra bits provided by using joint stereo but at a 320 cbr will you really be missing much using true stereo? To me true stereo separation is critical and I wonder why the insane preset doesn't include it as the default. To me the insane preset with the following overrides are what should be used.

--alt-preset insane -ms -d -c

Or would it be better to define the complete command line avoiding the insane preset all together, since the preset was probably designed for joint stereo in mind and might not give the same type of quality when used with the "-ms" (stereo) setting.

just read a bit about JS and pure Stereo. Pure Stereo is just a waste of bits as you can't hear ANY difference to the JS mode Lame uses!

I would have to disagree here. I can tell the difference between joint stereo and pure stereo on some music tracks, even at this bit-rate. Joint stereo will not always give you the same results as with true stereo. As for wasted bits, at 320 cbr there's already a lot of wasted bits just because the encoding is being done at a constant bit-rate. I use my mp3s on a high end stereo system so maybe I'm over critical about the overall sound quality of my mp3s, I would use a lossless compression but I also use my mp3s with the newer iPod's so it's easier to only have to worry about one audio format. I will admit that LAME probably has the best joint stereo implementation out there but it's still not the same as true stereo.

The joint stereo mode used by Lame is Middle-Side stereo. This mode does not reduce stereo separation, unlike the Intensity Stereo mode implemented by some other encoders.

PureDreams, I am sure you can give us some examples where --alt-preset insane has worse stereo image than -h -b320 -m s.
Please encode the original wav problem samples with flac and post the links here. Thank you

Please provide us a lossless clip where you can support this statement with ABX results, as stated in the HA Terms of Service #8
Furthermore, you could try to put up the effort to read the FAQ and use the search function to find threads like this one.

First off, I wasn't expecting such negative feedback - but the one post above did provide some interesting information about the joint stereo that LAME uses. Please be reminded, I'm not saying that joint stereo produces a lesser quality sound vs true stereo but rather I can tell a difference in the stereo imaging on some music tracks. Also note, I'm playing my music back on a high end stereo system (over $5,000) so what I may hear and what you hear are two different things depending on the equipment used. But just as a side note, when I went speaker shopping for the setup I have now the difference in sound from a $1,000 speaker (each) to a $2,000+ speaker (each) can be heard at the same volume levels. Also, stereo imaging is hearable across the entire sound spectrum, it's just less noticeable at the lower frequencies.

I don't concern myself with graphs or numbers so I won't be able to provide that kind of information. But I will find a few music tracks (just the sections that are noticeable) and will post the raw wave file (in ape format), along with a mp3 joint and mp3 normal stereo setting. I'll post the samples later this week when I have more time to find a music track that has the most trouble with joint stereo, as far as stereo imaging is concerned.

But to clarify things...I'm using the following setting

--alt-preset insane -ms -d -c

and NOT these, as someone else posted and then reposted by someone else (didn't the person read the original posting?).

-h -b320 -m s

Why does it seem that every new member to this forum have to disagre with the preset in of of there 5 first post ?
and still lack the ability to do probber abx testing.

A god advise i have always followed was to read in a forum at least a month before making statments.
Many times you can learn alot by setting youself in listen mode for the first month.

Yay. You seem to have succeeded in making all the newbie mistakes all at once.

1) Not reading the FAQ or using the search function, causing you to ask something that has been asked about 1000 times already.

First of all, there's a thread marked 'Read before posting', I suggest you actually do that:

http://www.hydrogenaudio.org/show.php/act/...t/ST/f/1/t/7516

Specifically the section: "Why is joint stereo better than pure stereo ?"

That should already answer your question to a large extent.

2) Claiming a commandline is 'better' without any evidence whatsoever to support your statement. And no, what you think (which was wrong) what it does is not sufficient. Not only is it foolish and certain to provoke harsh responses, it's also in violation of the forum rules.

3) Reminding everyone that you have a really expensive setup that consequently surely must be better than what everyone else has and thus magically allowing you to hear things noone else can hear. It's the best way to never ever get taken seriously. You only forgot to mention you have a PhD and thus surely are also smarter than anyone else here.

4) Exposing yourself at never having seriously tested what you are recommended, making you look like a total fool. Hint: the -d switch doesn't work as you describe, in fact it's entirely the opposite. This was changed in LAME but not in the documentation. Of course, if you had actually TESTED what you recommend you would have figured this out. Since you haven't tested what you recommend, why should we take you seriously?

5) Recommending normalization. Normalization is in general a very bad idea to do, simply because it doesn't accomplish anything, contrary to what you may think.

You weren't expecting negative feedback? Think again.

I've been reading and searching the posts for over a week before I decided to become a member - wish they wouldn't ask for your e-mail address during sign-up, always worried about getting more junk mail. I understand that LAME uses a different flavor of joint stereo during higher bit rates. But Mid/Side Stereo is not stereo image lossless if used with a lossy encoding format like mp3. It's used to gain more bits so the encoding can be of a higher quality. The cost is loss of stereo imaging/separation on some music tracks. Unlike most, I'm willing to give up a bit of sound quality during those times for true imaging - at least if I'm using 320 cbr. But as a side note, Mid/Side Stereo can provide true stereo imaging if used with a lossless format and implemented correctly, such as with the ape file format (Monkey Audio - uses Mid/Side Stereo during it's lossless compression).

Listed below is an interesting read, for those that care to dig deeper.

Joint Stereo - The Myths ... and The Realities

You will not receive any mail unless you ask for it (e.g. resending your password if you forget it).



You do not understand and you are wrong. LAME does not support Intensity Stereo, it always uses Joint Stereo LR/MS switching, and does so at both low and high bitrates.



This is actually correct. But what you are completely failing to realize is that this is also the case for forced stereo. The loss in imaging results from the encoder failing to properly encode the Side channel. But the exact same effect exists in full stereo coding when the encoder artifacts in one channel and not in another. There is no advantage of forced stereo over mid side stereo at all here. In fact, in most cases forced stereo is more likely to do this, because it is a less efficient encoding method for most sounds.

LAME APS uses advanced switching code to determine the encoding type that is least likely to artifact, taking into account that a loss of stereo imaging is to be avoided at all costs.

It's not clear to me from your posting whether you understand this, but Joint Stereo means that both Mid-Side stereo and Left-Right stereo are used, whatever one is most appropriate for the sound being encoded.



Completely correct.



In most cases it will be exactly the opposite.

Note that I say in most cases. It will be possible to find clips where JS switching fails, but there will be more where Forced Stereo is worse, _including_ in stereo imaging.



It's a very good explanation, maybe it should be added in the FAQ. Would you mind reading it? You certainly don't sound like you did. The author is specifically recommending to use Joint Stereo over forced stereo.

Tssk...
This matter has already been researched to great extents, and it turns out you can tweak LAME much further.

Here's a quote from the relevant post (that was me):



As you will see, 48kHz VBR double-mono is the way to go. Not a single bit is shared between the 2 channels, because we don't trust joint-stereo. Also the psymodel is disabled, because... well, who needs it anyway.

Cheers

Thanks, I will have to check that commandline out ...

@PureDreams

You will be taken seroiusly the moment you provide hard evidence and test samples where stereo imaging suffers due to M/S encoding over separated full stereo.

Hard evidence means that you will encode your test sample two times ... standard --alt-preset insane and your preferred commandline with -ms ... then you will decode both samples and do some blind testing (aka ABX) on your high end equipment ... post the two samples (compressed with a lossless format) here and you will see that anyone interested will take care of your "problem" instead of giving you a negative feedback ...

> Yay. You seem to have succeeded in making all the newbie mistakes all at once.

I consider this an unacceptable response from a moderator of the group.

> 1) Not reading the FAQ or using the search function, causing you to ask something that has been asked about 1000 times already.
>
>First of all, there's a thread marked 'Read before posting', I suggest you actually do that:
>
>http://www.hydrogenaudio.org/show.php/act/...t/ST/f/1/t/7516
>
>Specifically the section: "Why is joint stereo better than pure stereo ?"
>
>That should already answer your question to a large extent.

No where do I see it mention the effects of overriding the insane preset with stereo in the section you've named above. It's well known that lower bit-rates benefit greatly by using joint stereo. But at 320 cbr you are dealing with a very high quality encoding and at that level imaging is paramount. Even Dibrom has stated that - "Now, that's not to say the --alt-presets are perfect. I certainly know they aren't. But they also don't have some massive flaw in regards to stereo image which is present to the degree some people imply. In fact, the only case I've seen which I put any credence in is the few isolated cases which Wombat has found (and provided samples for I might add). I will eventually attempt to address these few samples, but note that these are exceptional cases, not common cases, and as far as I can tell, they are completely unrelated to the other complaints being made. " I wasn't stating that the preset was seriously flawed but I have noticed on a few music tracks that there have been imaging problems.

> 2) Claiming a command line is 'better' without any evidence whatsoever to support your statement. And no, what you think (which was wrong) what it does is not sufficient. Not only is it foolish and certain to provoke harsh responses, it's also in violation of the forum rules.

I never claimed that the command line I listed was "better". I just asked why wasn't "-ms" used during the insane setting because it provides true stereo separation/imaging all the time, at the cost of a bit in audio quality. As mentioned above, even Dibrom has stated that under rare cases imaging has been effected.

>3) Reminding everyone that you have a really expensive setup that consequently surely must be better than what everyone else has and thus magically allowing you to hear things noone else can hear. It's the best way to never ever get taken seriously. You only forgot to mention you have a PhD and thus surely are also smarter than anyone else here.

What kind of response is this? If one were to listen to music on a pair of $40 computer speakers then you'll probably never know the difference. Stating that I'm using higher end equipment does matter, and yes there is a difference. If I was using cheap stuff people would say "must be your cheapo speakers" get better ones. I take my music seriously by listening to it, since that's what matters - no PhD required. Btw, would you take someone's view more seriously because they have a PhD but rarely listen to music or someone that listens to a lot of music often? I surely wouldn't buy equipment because the "specs." said it was best, I would use the product to determine it's quality.

>4) Exposing yourself at never having seriously tested what you are recommended, making you look like a total fool. Hint: the -d switch doesn't work as you describe, in fact it's entirely the opposite. This was changed in LAME but not in the documentation. Of course, if you had actually TESTED what you recommend you would have figured this out. Since you haven't tested what you recommend, why should we take you seriously?

Again, nice response from a moderator. Hmm...now I'm suppose to be psychic and know the source code by heart since the documentation doesn't mention the change. The "-d" switch isn't a part of the mp3 standard but considering that there could be an increased quality benefit I use it. Btw, there's no need to test what one actually uses for their entire music collection.

>5) Recommending normalization. Normalization is in general a very bad idea to do, simply because it doesn't accomplish anything, contrary to what you may think.

Normalizing does wonders for me - it keeps me from constantly adjusting the volume when I'm listening to music. Normalization is the only way to deal with this issue unless you have software/hardeare that scans the music before playback and adjust the volume for you. Sure, normalization can cause music to be cut off at the high or low end, but that's a very rare case and if used incorrectly.

> You weren't expecting negative feedback? Think again.

No, but now you have mine.

We should collect all these threads in a "newbie mistakes" forum, then when another one shows up, just forward them to it. Maybe once they realize how irrational the other (virtually identical) threads sound, they will realize their error.

Look, PureDreams... ABX means double-blind testing. We firmly believe you are the victim of placebo. Prove us wrong! Joint Stereo has long been considered PERFECT by this community because NO ONE has been able to ABX a difference.

If you refuse to do this testing, then this thread will be locked, and you will most likely be banned. Those are the breaks, my friend.

Does anyone need any further evidence that you have no idea what you're talking about? Please browse forum, read, learn (especially from http://replaygain.hydrogenaudio.org/ in normalisation case) THEN discuss.

Wrong. Very low bitrates can benefit greatly from intensity stereo (which is not implemented in LAME).
On the other hand, mid-side stereo brings benefits at any bitrate.
Btw the article you mentioned does contain the correct info !

If you use full-stereo you decrease coding efficiency by 30-40% on average. Again if you think you heard an issue at 320kbps, please provide evidence.


Having high-end equipment does matter, but you must still do proper (ie: rigorous) tests. A careful blind test on $40 headphones may have more value than just thinking you heard a difference in a recording studio.

Having very good equipment doesn't prevent one from being misled by his/her feelings. Especially when seeking very subtle differences.


Why even bother with source code or command-line tweaking in the first place ? If the -d switch actually brought improvements, if would be part of the --alt-preset switches. These presets were introduced for a reason.

By the way, it's possible that mid-side stereo will actually improve quality (but I don't know if this is implemented in LAME). Here's a quote from Ivan Dimkovic, about mppenc:

Umm.. I'd really abstain from this kind of declarations.
In lossy audio few things if anything is perfect. And there certainly have been people ABXing differences in certain area (for example mid/side vs l-r pre-echo etc.), but what is better and in which situation is not so simple at all.

That said, insane uses by default quite a lot of left-right frames. Anybody who thinks there's a problem with mid/side narrowing the stereo field, should first check for example with encspot if the encoder is even using mid/side frames at those positions where the listener thinks mid/side is to blame (it could be that mid/side would be infact needed..) Then start doing the standard ABXing etc...

Anyway, keep it calm people so that this thread doesn't need to be locked.

I would also like to add that if you wish to list hardware used, listing it as "$2000 speakers" also does not cut it. I have setup complete systems that hit the $1500 mark that outperform $10,000 systems in terms of quality, sound reproduction, imaging, etc. So $$$ ist not quality guarantee. If you wish to espouse your system, give a listing of the system in question.

Also, if you are serious enough to use 320 CBR, you might as well go lossless. The filesize difference is not going to be that much, and you don't sacrifice any quality.

You clearly have STILL not read the forum rules, which SPECIFICALLY address this in rule 8.

Even if there was not rule 8, I'd still consider my response entirely appriate.

But I'm obviously biased. If you still think my reaction is unappropriate and uncalled for, feel free to complain to the forum administrators (JohnV and Dibrom) and I will accept whatever they have to say on this.



I quote the first post in thread 2:



Now, that wasn't very hard, was it?

Not to mention the last thread talks exclusively about using -ms with --alt-preset extreme, which is also a very high bitrate setting.



Yes. Notice how this exactly corresponds to my statement that there exist EXCEPTIONS where forced stereo outperforms joint stereo in imaging. But they're EXCEPTIONS because they're in the general it is the other way around.



This isn't necessarily true, as I explained above. You are assuming that -ms automatically gives you a better stereo image, which is incorrect.



That is an assumption you make. I'd personally recommend headphones over anything for detecting stereo imaging issues, but people disagree here. In any case low grade equipment does not necessarily mean that you'll not be able to spot defects, as experience has indicated many times.



Good luck finding examples on the board. You'll have to look long and hard. Hint: Problems that occur on low grade equipment are generally worse on high end equipment.



Likewise, I don't care at all who you are and what your equipment is. I care whether you present evidence that can support your claim, and if your statements are correct and making sense.



Yes, this seems to be the root of the problem. You're just doing stuff _assuming_ you're right and know what you're talking about without _bothering_ to even _check_ whether you're right at all.



I see other people have already cleared up this horrible misunderstading, so I will not comment further on it.

Chastity, what always astonishes me, is how people would not trust a particular codec (ie: they use switches that the developers themselves will never recommend), yet they still use (and trust) the very same codec as soon as they "tweaked" it, and encode everything with it...

That part of it doesn't bother me, since I am not the one listening to that file, unless they upload it somewhere.

The thread might be locked if there is no evidence provided and the discussion repeats itself. But PureDreams said he would provide some samples, therefore the thread should not be locked.

But banning a user usually occurs if he/she ignores warnings repeatedly.

Just out of curiosity, is there an identified sample in which "real stereo (-ms)" will do better (one of those "exceptions" mentioned)?

In thinking Jimi Hendrix...

edited for clarity of question (not that it matters much anyway...I found the answer)

come on everybody! let's all get on the newbie bashing wagon!

This whole topic has gotten a bit off topic if you ask me. I asked a question, which could've easily been explained or redirected in a polite manner but instead explodes into "you have no idea of what your talking about attitude". Not everyone that responded was like that. But for a forum that's considered a good authority on audio formats it leaves me with doubts about the quality of the individuals based on some of the responses.

As previously mentioned, I will provide examples of the stereo imaging issues when using the default insane preset later this week, along with the original wav. What I find interesting is that even if I post the sample files what will that resolve? That LAME can sometimes handle joint stereo incorrectly in rare cases - which is already known (and stated by Dibrom). Not to mention, I was not out to prove what settings were best but only asked why a particular switch/setting wasn't used in the insane preset, maybe overall audio quality was the trade off to dealing with rare stereo imaging problems?

As for what else was said...

* Replygain is nice if it were 100% compatible with all mp3 players (dvd players, portable players, car players, etc.). Hmm...did anyone ask if all the equipment I use supports this before saying I have no idea about what I'm talking about? Also, some CDs that have 100% levels could cause clipping with some mp3 encoders without first being normalized to something lower. I don't know if this is an issue with LAME but to be safe and since all my hardware is not replaygain aware I normalize to 98%. Btw, it would be preferable if the mp3 player scanned the music and auto gained (no ID3 tag needed) - like some of the CD players do with normal CDs.

* The question was based on true stereo separation/imaging and not coding efficiency and/or audio quality. Perhaps this is the confusion - true stereo imaging is more of a concern for me then audio quality at 320cbr. Don't get me wrong, I still want audio quality but I'm willing to give up a few bits to insure I always get true stereo imaging.

* The suggestion to use encspot is a good one and I will report my findings when I post the music tracks causing the problem.

* As for the equipment I have - Paradigm Studio 100 reference series, I wanted the B&W's because they had a bit more sweetness in sound but I just couldn't afford the increase in price. I also have the Paradigm studio center channel & surrounds + a SV subwoofer but don't use them when listening to just music. I'm using these speakers with a B&K AVR505 receiver, wanted to go separates but it was too costly. I playback my music on this system using a Panasonic XP50 DVD player with DVD-Rs burned full of mp3s - I sometimes use my iPod also. The front & center channel speakers are bi-wired with 10 gauge Phoenix Gold wire.

* I would use lossless compression but want to stick with one file format since I have other devices that use the mp3 format - car, DVD player, iPod, and an older Nomad that I use at work. When quality is the utmost importance then I create a custom CD with my favorite tunes.

* The extreme preset doesn't produce the same results as the insane one. They may be similar but not identical so any feedback from what was said about extreme doesn't apply to the insane setting, even thought they both offer high bit-rates.

* "best sound quality" does not mean best stereo imaging.

* I stated that the "-ms" switch will always give true stereo imaging - never a doubt.

* A good set of headphones ($100+) would be a good way of doing ABX. But I don't own a highend pair. I like to feel my music and it sounds much different when the music breaths into the room from speakers. Perhaps that's the problem, good imaging is hard to find in speakers which may be why I'm noticing it more with some mp3s.

* In my findings, higher end gear makes defects more noticeable while low end stuff "hides" it. I can hear details in some CDs that I never heard on my older cheaper stereo system. Not to mention, if the CD is mastered right the stereo imaging just comes out at you with good equipment.

* Trying things is a way of testing the quality of something. I didn't make my speakers or my receiver but I experimented with the various settings and connections to get the best quality out of them that I can. If people never tried anything and always took someone else's word on things then how would you know what you got is any good or doesn't have problems?

* A codec is like a car, it will run with it's default and recommended settings. But just because someone wants something different doesn't make the codec/car bad...that's why people customize cars and some customize their LAME settings. Would I get heat if I override the lowpass down to 19Khz? Nope...they would just ask if you can't hear it then why encode it.

At any rate, wait for my track uploads and hopefully what I post can be heard by all. Who knows, it could possibly be my equipment or the acoustics of my room - not that anyone cared to mention that as a possible problem. But if you must reply to my response then please only do so if it's informative and will benefit others.

Before you upload anything, I recommend getting a copy of Encspot and analyzing the particular passage. Out of all the encoders, the -insane mode has the most sensitive JS switching mode I have come across. When I last tested my favorite "stereo separation" clip, all other encoders failed on it, but the -insane mode handles it with ease and goes into full left-right mode.

Replaygain for MP3 IS 100% compatible with ALL mp3 players, as it modifies gain values for each frame of the actual MP3 (it is not just a tag, and any player is able to read the files with the correct replaygain value.) Normalizing does not prevent clipping in the final MP3, nor does it make all your files sound the same volume.



Garf already explained that this is not true in one of his earlier posts.



Trying things is a good way to test the quality of something. The reason that nobody is taking your word on this issue is because people have tested -ms vs -mj for hours and hours on end, and came to the same conclusion that -ms results in decreased audio quality and little or no loss in stereo separation in the vast majority of cases. The reason that you SHOULD take the forums word on this issue is because it has been thoroughly tested, again and again and again. And again !



As for your system, are you sure that your Panasonic XP50 can play MP3 files 100% correctly? Not all MP3 decoders are created equally. Certain decoders can have problems with the LAME codec in general, and others behave poorly with joint stereo. You need to make sure your hardware is up to spec before blaming a setting in the codec. Why not use a computer to play your audio files?

Edit: Should have stated using MP3Gain to replaygain your files is 100% compatible, lo siento, my mistake B) ...

Statements made on a factual or technical basis have absolutely no correlation to the "character" of the individual. I suggest you do not continue in this line of debate.


I'd like to ask you to refrain from the "Dibrom said" comments. I don't like my comments to be used out of context, and in my opinion, this is definitely what you are doing.

Every lossy perceptual coder can have problem samples, but the conclusion you are trying to draw here is not supported by the evidence at hand. Yes, it may be possible that joint stereo causes problems in some cases, but it's also possible and much more likely and common for normal stereo to cause problems as well. Given the amount of tuning that the alt-presets have seen and the amount of samples thrown at it, joint stereo has been shown to work better almost every single time. On the contrary, as Garf said, there have only been a handful of exceptions where normal stereo (even at 320kbps) performs better. What's more is that the problems that joint stereo might have caused in these cases is not related to stereo collapse as you seem to imply. You seem to have missed that point in your repeated usage of my words.



Umm.. I think by this statement it actually does make it pretty apparent that you don't have any idea what you're talking about regarding mp3 and replaygain. The most popular replaygain implementations available modify the mp3 themselves and thus are completely hardware independent.



You seem to have missed the fact that mid-side stereo (not intensity stereo) does not degrade stereo imaging by design. Thus the whole notion of the fact that you're getting a compromised stereo field is flawed. Others have already stated that in this thread and I stated it many times in the threads you were pointed to. Did you not read them?



"best sound quality" in this context refers to the measured deviation of the encoded signal from the original. This does take into account stereo imaging, but once again, your notions of a compromised stereo image from mid-side stereo (which are very common) are baseless.



You seem to misunderstand some of the fundamentals here. Stop thinking in terms of "stereo image", etc. These things don't apply in the manner which you seem to think they do in terms of audio coding. If artifacting occurs because inefficient coding (using normal stereo instead of joint stereo), then it very well may have an audible effect on what could be interpreted as "stereo imaging". The point is that the type of artifacts being generated here are not correlated in the manner you think they are. -ms really does nothing to guarentee you a better "stereo image", unless you're using one of the Fhg encoders with a buggy/improper implementation.



Yes, testing is good as long as it is not done arbitrarily and without a common methodology.



This is a poor analogy. If you're talking about quality, and assume that everyone has a common goal of reaching the best quality, than this example doesn't work. Where it might work is if people are needing something completely different such as mono encodings or specific bitrates or other functionality oriented features which are in a completely different class from eachother.

Beyond that, the presets are designed to be transparent at a top level, not at a mean level. This means that if *anyone* can hear the difference, then the idea (at least when I was still working on them) was to tweak it until that was no longer the case. If you could actually show that your hearing is significantly better than almost everyone who has ever participated in tuning, and that you can hear these problems reliably (AND that normal stereo improves upon this), then you might be on to something. So far you haven't, and I must admit that I'm quite skeptical about your ability to ever do so.

I think the best thing for you to do at this point would be to either provide samples or not continue debating these points.

Well.. just to be exact - Replaygain is just a name for a certain method. You can utilize this method in several ways. With MP3 for example adjusting global gain value, using tag, using some player db value.

And like you pointed out replaygain utilized by adjusting MP3 global gain values (MP3Gain) is compatible with all players.

Although you're right that the tone with which someone expresses himself has nothing to do with his technical knowledge and the accuracy of the facts presented, I am sorry to also agree with PureDreams's first paragraph. Actually, the question was answered in a short and polite manner at the beginning (by B Lebsanft, Gabriel, and Sony666), but things went different with Sven Bent and Garf's posts.
I won't blame them, since forum Rule 8 says "you are likely to be receive harsh responses to your posts. The HydrogenAudio staff will not take action against any users which may post these responses", but I nonetheless think that there was no use for them to answer like this.


We will be able to discuss about something real. I'm not saying that what you hear is not real, I don't know it. But post after post, command line after command line, we have lost faith in people's claims, even audiophile or sound engineers. Remeber the story of this very experienced guy from an audiophile magazine that believed he was listening to the most expensive speaker cable of the world ("one must be deaf not to hear it !") while it was actually 24 gauge zip cord.
So we stated that ABX blind tests were the rule. They are effective, let's use them. A bit time consuming, but it's nothing compared to the time spend talking about things that don't exist.

Dibrom brought up a concern which I've had throughout this thread...

You're trying to separate "sound quality" from "stereo image". Haven't you ever thought that it's certain aspects of the sound quality that give rise to your perception of the stereo image? The two are not opposites, but related:

Very inacurrate audio reproduction gives rise to a very poor stereo image. Very accurate sound reproduction yields a very good stereo image. Some of the factors that cause this in Hi-Fi equipment are irrelevant in audio coding, but it's still food for thought...


You raised a smile when you mentioned the cable used to bi-wire your speakers, whilst normalizing your mp3s - stick around long enough and you might learn from your mistakes. Don't be angry at people jumping down your throat - if you'd seen as many people arrive at this forum, state "I'm right, you're all wrong", only to find out that they're wrong after all - well, you might take the same attitude to "newbies" who don't read all the excellent material that's available to them.

The program you need instead of normalising is mp3gain. If you don't want to make all your albums or tracks the same loudness, just use a constant gain adjustment suitable to prevent clipping, and look in the mp3 FAQ on this board - it has all about mp3gain, mp3 quality etc etc etc etc etc - have you really read through it?

joint stereo for sceptics

Finally - believe what you're being told about true stereo and placebo! If you've found some tracks where there is an audible problem, post them - developers are interested in these kind of things. But listen blind first - please - or you're wasting your time and everyone else's. (Google search for "N-rays"!)

Let's imagine you have a really wide stereo field. Not even a synthetic "pan this hard left, this hard right" type thing - a real acoustic purist two-channel recording which happens to capture a very wide stereo field. You're going to add (hopefully) inaudible noise to it - that's what psychoacoustic coding does. For it to be masked (inaudible), this noise has to fit in with psychoacoustic principles - one of the most important ones (which isn't talked about very often) is that, to be inaudible, the coding noise needs to be at the same spatial location as the original signal component which is supposed to be masking it.

If you restrict the codec to L/R encoding, then that's exactly where it can put the coding noise: in the left and right channels. If you could hear that noise, that's where it would sound like it was coming from: from two, reasonably uncorrelated noise sources, sat in each loud speaker. If you have a pure mono signal, then the noise added to each channel will be identical - but even the slightest inter-channel difference will mean that the noise added to each channel is different, and you'll get two different noise signals, located as I've described.

If you let the codec use M/S encoding, then it has another option: put the coding noise in the dead centre and at the sides (actually, out of phase in the two stereo channels).

True stereo means it can only do the former. Joint stereo means it can do the former, or the latter, and it can vary the amount of noise in the left channel or the right channel, or the mid channel or the side channel - as appropriate, to make sure it's hidden.

Joint stereo gives it the chance of putting the coding noise in roughly the same "place" as the dominant component of the original audio signal. True stereo does not. Joint stereo means that a given amount of noise will be less audible, because it's with the dominant signal, and less likely to drag the sound stage away from where it's supposed to be, because it's co-located with the dominant signal.


Of course, we're hoping that the coding noise is inaudible. If you're going to put the noise somewhere else (i.e. not in the same position within the stereo field at the dominant signal), then it's going to have to be much quieter in order to be inaudible. It's also going to have to be much quieter in order not to damage the stereo sound field. If you can put the noise in nearly the same position as the dominant sound, then the opposite applies: it could be louder before being 1) audible, or 2) changing the sound field.

That's one advantage to joint stereo - and here's another: It's more efficient. That means, for a given sound, if you can choose between L/R and M/S (as you can when using Joint Stereo), you can choose the one which will use fewer bits. Or, at a constant bitrate, you can choose the one which will add less noise.

So, there's a double gain with joint stereo: you could add more noise before it became audible, but you can add less noise, because it's more efficient!


How cool is that?


It only works if the codec switches sensibly between L/R and M/S when using Joint Stereo mode. BUT even if it completely messes up, it cannot be worse than "true" stereo on average. True stereo will be worst on samples where M/S would have been useful (that's lots of the time in most recordings!), and Joint stereo (if you're using a hypothetical stupid JS encoder: no switching, only M/S) would be worst on samples where L/R would have been useful.

So, for Joint Stereo to sound worse, it's got to be used on a sample where L/R should have been chosen, but choose M/S instead. With the tweaks in the lame --alt-presets, this is so rare as to be virtually unheard of. For "True stereo" to use L/R where M/S would have been better is going to happen much of the time on most recordings. Why make the encoder do this? Remember "better" isn't just better sound quality - it's better stereo image (sound field) as well, because the two are related.


Where you do hear Joint Stereo doing badly is where encoders are buggy: The FhG one mentioned in your link sometimes low-pass filters the "S" channel (of M/S) at 2kHz!!!!! As mentioned elsewhere, the Xing encoders don't even implement M/S - they use IS all the way up the bitrate scale - disaster!



Hope this is of interest. Have I convinced you?


If your ears are still telling you otherwise, please send us the samples!

Cheers,
David.
http://www.David.Robinson.org/

There could also be a miscommunication of terminology going on here.

It could very well be that what is being described as a "stereo separation" or "stereo image" problem by one person would be described as a "pre-echo" problem by another. Sometimes temporal artifacting makes things like cymbals sound a bit more phasey, and I can easily imagine this being identified as a change in stereo image.

Classic narrowing of soundstage is something I've really only heard at lower bitrates, or with the buggy Fhg mp3 encoder, or with Ogg at bitrates at or below about 128 kbit/s.

With samples in hand, we can at least agree that something is wrong, even if we don't agree on how to describe it.

ff123

Are you talking about peak normalization ? Or are you talking about overall lodness with dynamic compression (aka RMS normaliziation) ?

while the first won't give you audible results on most modern CD's, the second will ... with the side effect of destroying the original signal and dynamics ... might be useful for car stereo or portable use (general: on systems that aren't able to reproduce high dynamics) ... but apply this to a backup copy only ...

By the way ... if you talk about flawed stereo imaging ... I see that you are using a multichannel sub/satellite system ... room acoustics and speaker placement have a very big (if not the biggest) influence on stereo imaging (like staging of big orchestras) ...

As an addition (and as viper1313 stated before), MP3 Hardware decoders are sometimes (even when built into panasonic DVD players) buggy ... there have been reported problems on some decoders with VBR, M/S encoders and even LAME itself ... If I were you, I would indeed do some blind testing ...

@PureDreams

A quick reply on the whole "attitude" thing that's arisen - I've been with HA for about a year now. I've seen my fair share of people with attitudes of superiority and intolerance for those who don't know as much. But as was said before, this has ZERO bearing on the technical knowledge this community possesses! You'd be very hard-pressed to find a community anywhere on the internet with the knowledge and expertise this community has. You're really only hurting yourself if you decide to leave because you don't like the response your first post received.

Second, I'll once again re-hash what others have already said. The presets have gone through more man hours of tweaking and testing than you could ever possibly hope to do on your own. Why is that surprising that people are going to press you for hard facts when making claims when they've already long since come to a consensus as to what works best?

Third, for the love of God, use MP3Gain if you want to normalize your songs!

Fourth, you make a big deal about the expense of your speaker system. My first impression when I read your post was "Wow, this guy is full of himself." You're also making the assumption that the equipment you're listening on is better than that of which these presets were tuned on. Again, they were tuned on a far greater range of hardware than you could possibly imagine. That's the benefit of community! Not only that, but as was said before, errors in encoding tend to become more pronounced on lower-end equipment. This is something you'll see time after time after reading here for awhile.

In summary, stick around and you might learn something I know I sure have... Also, trust the presets. They were designed to be as tweaked out as they could possibly be. If you think they aren't as good as they could be, be prepared to be met with skepticism and demands of proof.

To add some humor to the discussion
When you mentioned cars I thought of one i saw on TV a couple a days ago. To me, this company does to cars what has been done here to Lame with the presets B)
preset tuning

Btw: I don't know alot about cars and sorry about the complete off topic.

I'm very sorry that my (misplaced) comment has made all this fuss.
It was never intended to bring alle these bad vibes.

I do hope that PD will stay and recived the same great knowledge from HA that i have recieved, but i still think he needs to listen in a debate and be willing to learn.
I know that i learned alot from Ha and more then once needed to "readjust my view at the world".*

*This probally got transleted all wrong.

ohh I hate that js vs stereo discussion, it's been discussed many times before, in many forums.

With lame you can use whatever you wish, so who cares if pure stereo is not in the preset.

I too would not use js, but i'm one of the few on this forum, and i've been around from the very beginning.

Why? Please don't say something along the lines of: "Because it's simply my preference." That doesn't add anything useful here.



How is this supposed to a bearing on anything?

Wow, someone who has eluded the limelight for so many years.
All the more we should have these 'controversial' discussion threads, to ferret out these misguided people.

Pardon me, but you've already seen so many persuasive arguments, and one that explained the role of joint stereo VERY well (by 2Bdecided), and you are still unconvinced?


Has it ever occurred to you that joint stereo is in the alt-presets for a reason?

I think we scared him/her off - Maybe he feels his PureDreams turned into a Pure Nightmare! Hope not.

D.

Sigh. someone else who goes away to ABX but doesn't come back with the results.

We should have a list of these long and pointless threads in the FAQ:

"things that people think they hear, but don't"

part 1, those who come back and say "gee fellas - sorry 'bout that - I was wrong" and
part 2, those that never show their faces again.


Cheers,
David.

3. stubborn ones still on forum... (cheers Budgie)
Disclaimer: I didn't want to hurt anybody's feelings with this post

Which newbie reads FAQ completely anyway...

I remember one of my first posts. I bashed the quality of CD's in general.

Actually its threads like this that made me lurk for 9 months before I dared register. So they're not all bad.

But also look at the wiseguy. they have to tell everything agian and agian because people is to lazy to read Faq. etc.
its just give you the feeeling thtat its hopeless to help people.

Bleh. I stand by what I said. He made all the newbie mistakes, didn't read the FAQ, and attempted to use completely wrong statements to support his point of view.

When someone makes very dubious statements and does not properly support them, expect a very harsh response whether you are a newbie or not. I can understand a newbie not finding the FAQ and asking about the JS/MS stereo issue, and I will help them, but if they CLAIM that JS doesn't cut it and provide no proof they are out of bounds (especially if they make statements about the greatness of their audio gear and that that surely must make the difference.).

If you try to argue your way out of it, you're definetely fair game.

The fact that you can use whatever preset/setting with LAME you desire is true. But if you're after audio quality, which most people here on Hydrogenaudio are, why use a setting/switch that produces inferior MP3s? Why not use what's advised and save yourself some woe?

I think it's strange that you've been around here for so long and still use settings that aren't optimal. I mean, why not stick with good old Blade while you're at it? It is a faster encoder than LAME, right? And audio quality is not that important to you, is it?

I mean if people here start thinking like you do, then there's no point in trying to improve the audio quality, because people don't care about audio quality. All they want is to see it compressed maximally so they save disk space, or an encoder that works fast (or a combination). I also wonder then what the point in buying expensive audio equipment is, if audio quality isn't that important after all?

Of course people care about audio quality. People convince themselves that their settings sound the best, and then they really enjoy their music more that way. The problem is that this method of achieving their desired level of quality is completely unscientific and totally psychosomatic. What works for one person will not work for another (unless the first person coaxes the second person to imagine the same audible benefits). These imagined benefits are not reliable and often not repeatable, and certainly not repeatable in a controlled test environment. They don't belong in an objective discussion of audio quality.

Oh, and since you mentioned Blade and audio quality, I thought I'd point out that people with strange command-line parameters are often hunting for the elusive tonal purity.

Actually, it has been proved many times that blade encoded files have more air and a better sound stage.
By the way, I put a green jpeg at the beginning and at the end of all my audio files. This way, the hard drive is more stable (even more with a uni-directionnal ide cable), which improve greatly the treble accuracy.