Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Joint Stereo Reverting to Stereo (Clarification) (Read 8904 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Joint Stereo Reverting to Stereo (Clarification)

Hi,

This is my first post and I have tried to find the answers to the following questions on my own. Unfortunately, I wasn't able to find a good explanation. I also wasn't sure if this was a general question or a technical one. The two MP3 forums share a common description.

Repeatively I've read in here and elsewhere, that Joint Stereo produces higher quality results then Stereo. If Joint Stereo produces higher quality, then why does Joint Stereo default back to Stereo when the complexity of the stereo image gets tough?

Wouldn't that inherently suggest that Stereo produces higher quality and that maybe size might be really Joint Stereo's purpose instead of strictly quality as is so often mentioned?

I'm currently using Lame version 3.96 with "--preset extreme" as the only option and have noticed that the vast majority of the frames are encoded in Stereo. Being that this is the case, it appears that just using Stereo would add slightly higher quality.
Ronn

Joint Stereo Reverting to Stereo (Clarification)

Reply #1 –
Mid-Side stereo is a lossless procedure that sometimes uses bits more efficiently than simple stereo.  At other times regular stereo is more efficient.  That is why LAME's joint stereo can switch between them frame by frame I believe.
Nero AAC 1.5.1.0: -q0.45

Joint Stereo Reverting to Stereo (Clarification)

Reply #2 –
Joint Stereo == Mid/Side Stereo + Left/Right Stereo

Joint Stereo gives the encoder the option to pick whatever is most efficient, and hence get the best quality at the lowest bitrate. Forcing Stereo (==Left/Right stereo) removes that choice.

Joint Stereo Reverting to Stereo (Clarification)

Reply #3 –
Garf, the often used term "full stereo" when talking about LR stereo misleads many people to think, MS stereo is "not full stereo", e.g. of minor quality.
(I know that you do not)

Joint Stereo Reverting to Stereo (Clarification)

Reply #4 –
Fixed

Joint Stereo Reverting to Stereo (Clarification)

Reply #5 –
Wouldn't it be nice if they had named it
"advanced stereo" or sth instead of js?...
Wanna buy a monkey?

Joint Stereo Reverting to Stereo (Clarification)

Reply #6 –
Quote
Joint Stereo == Mid/Side Stereo + Left/Right Stereo
[a href="index.php?act=findpost&pid=244475"][{POST_SNAPBACK}][/a]


Joint Stereo can also include Intensity Stereo frames - not supported by LAME, though.

Joint Stereo Reverting to Stereo (Clarification)

Reply #8 –
Most of the time using mid/side is improving compression. It means that for a fixed size, you will in this case have an higher quality using mid/side than left/right.

Joint Stereo Reverting to Stereo (Clarification)

Reply #9 –
saved bits can be used for other tasks. So if we can save bits on the stereofield, without degrading perceived quality of the stereofield, then we can spend those bits to encoder other stuff at higher quality.

Concerning "advanced stereo" - i proposed something like that to the lame devs in another thread. IMHO, simple-stereo (so, 2 seperate channels) should be ignored as a switch and always "joint-stereo" be used - and to make up for that rename joint-stereo to something like "advanced stereo". Simple stereo besides of CBR is the most often made mistake when people encode audio. Since they dont >know< what it means, they just think "oh, the >word< stereo sounds better than joint-stereo..... so stereo has to be better". It gets especially frustrating if you download releases from net-labels and artists and they're encoded in CBR + Stereo. With r3mix out of the way, CBR having reasonable uses sometimes(i.e. old portable not supporting VBR), the stereo switch remains the biggest quality-hazard in LAME - and at the same time has one of the oldest myths associated to it.

- Lyx
I am arrogant and I can afford it because I deliver.

Joint Stereo Reverting to Stereo (Clarification)

Reply #10 –
Hi,

I want to thank everyone for all the replies.

However, I'm still confused as to why extreme uses so many more LR frames then MS frames. If an encode is to benefit from MS's efficiency, shouldn't the encode contain at least a majority of MS frames?

What constantly seems to happen with Extreme is the encode ends up having more than 90 percent LR frames. Here is just one recent example...

average: 275.8 kbps  LR: 9266 (95.40%)  MS: 447 (4.602%)

So, while it is technically a Joint Stereo file, the contents of the file is more like a Stereo encode. What there is of MS frames, I wondered does that 4.6 percent make that big of a difference for either quality or size? Maybe just a tiny, little bit?

It is Lame's extreme preset that is making the decision that nearly all frames, if given the choice between MS and LR, should be LR.

These questions of mine are meant for understanding. I'll stick with using the extreme preset as is, without modification.
Ronn

Joint Stereo Reverting to Stereo (Clarification)

Reply #11 –
I would asume that the extreme-preset - because of its intention - is very conservative and will pump up the bitrate & quality-related methods at the slightest possible difficulty. Think of it as a "sane" version of the "insane"-preset(*lol*). Those 5% can make a difference when at the same time the 320kbit limit is hit - because then you have a situation of limited bandwidth and in such a case M/S can pack more info into the available space of that frame.

So, one possible explanation could be that when the current bitrate isn't at 320kbit already, and there is the slightest uber-minor possibility that M/S may have difficulties, then lame would up the bitrate and use simple-stereo instead. I dont know if thats the case, so take this paragraph with caution - its just an ideo why the large amount of non-M/S-frames may be there. Someone more knowledgeable may be able to shed more light into this.

Also, there is a little confusion about the word "joint-stereo"...... some by it in reality mean the M/S stereo-mode, and some by it mean "that lame automatically switches between stereo-modes".

- Lyx
I am arrogant and I can afford it because I deliver.

Joint Stereo Reverting to Stereo (Clarification)

Reply #12 –
can i ask what sort of music is giving you 90% LR stereo frames?  remember it's all relative.

near-mono music (like portishead's self-titled album, and in fact a lot of trip-hop or sample-heavy music) will be 90% MS stereo at the same preset.

very wide stereo music, like a lot of '80s rock, or some electronic stuff (or live music, especially with lots of applause) will need more LR frames than MS frames.

LAME chooses which will be the most efficient for the frame it's currently looking at.

besides, when the bitrate limit is hit on a very difficult sample, an MS artefact will often be harder to notice than a LR artefact (depending on the listener's hearing).  i myself notice glitches that differ from 1 channel to the other before i'd notice a glitch that is more centered in the stereo field, such as one MS would give.

Joint Stereo Reverting to Stereo (Clarification)

Reply #13 –
Quote
I would asume that the extreme-preset - because of its intention - is very conservative and will pump up the bitrate & quality-related methods at the slightest possible difficulty. Think of it as a "sane" version of the "insane"-preset(*lol*). Those 5% can make a difference when at the same time the 320kbit limit is hit - because then you have a situation of limited bandwidth and in such a case M/S can pack more info into the available space of that frame.

So, one possible explanation could be that when the current bitrate isn't at 320kbit already, and there is the slightest uber-minor possibility that M/S may have difficulties, then lame would up the bitrate and use simple-stereo instead. I dont know if thats the case, so take this paragraph with caution - its just an ideo why the large amount of non-M/S-frames may be there. Someone more knowledgeable may be able to shed more light into this.

I probably should have included the entire histogram for that particular encode to show where the LR frames were being used. So I will do that now for clarity...

032 [  83] %*
128 [  18] %
160 [  43] %
192 [ 307] %%%%*
224 [1478] %%%%%%%%%%%%%%%%%%%%%*
256 [3337] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%**
320 [4447] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
average: 275.8 kbps  LR: 9266 (95.40%)  MS: 447 (4.602%)

As you can see, the LR frames were used in an area where there was still bitrate available. I agree that even though a small percentage of MS frames were used, it still an improvement over all frames being LR. While at the same time because of the preset, the tilt was towards more bitrate making the MS frames all appear below 320.

Quote
Also, there is a little confusion about the word "joint-stereo"...... some by it in reality mean the M/S stereo-mode, and some by it mean "that lame automatically switches between stereo-modes".

I hope I didn't confuse anyone on my use of MS. I was using it in terms of what's reported at the end of a Lame encode when using Joint Stereo mode via the preset.

Quote
can i ask what sort of music is giving you 90% LR stereo frames?  remember it's all relative.

The above was Two-Mix.

Quote
near-mono music (like portishead's self-titled album, and in fact a lot of trip-hop or sample-heavy music) will be 90% MS stereo at the same preset.

very wide stereo music, like a lot of '80s rock, or some electronic stuff (or live music, especially with lots of applause) will need more LR frames than MS frames.

Here's Stone Temple Pilots...

032 [ 101] %**
128 [  20] %
160 [  45] %*
192 [ 554] %%%%%%%%%%%%**
224 [1986] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*****
256 [2172] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%***
320 [2612] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*
average: 261.2 kbps  LR: 6939 (92.64%)  MS: 551 (7.356%)

And here's the 80's B52's...

032 [  110] **
128 [  28] %
160 [  155] %%
192 [ 2029] %%%%%%%%%%%%%%%%%%%%%%%%%%
224 [ 5342] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
256 [ 4621] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*
320 [ 3480] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
average: 248.3 kbps  LR: 15623 (99.10%)  MS: 142 (0.9007%)

While I haven't done a lot of encoding, the above is rather typical of the LR and MS frame breakdown using (I hope) similar content to what you have suggested.

However! Out of curiosity I just tried using Lame 3.90.3 using the same two tracks (as above and respectively) with quite different results...

032 [ 103] %**
128 [  24] %
160 [  35] %
192 [ 409] %%%%******
224 [1669] %%%%%%%%%%%%%%%%%%%%*********************
256 [2545] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%***************************
320 [2705] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%***********************
average: 264.5 kbps  LR: 4174 (55.73%)  MS: 3316 (44.27%)

And...

032 [  110] **
128 [  28] %
160 [  85] %*
192 [ 1549] %%%%%%%%%%%%%********
224 [ 4575] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*********
256 [ 5004] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%****
320 [ 4414] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%***
average: 256.0 kbps  LR: 13873 (88.00%)  MS: 1892 (12.00%)

I ran other tests and it seems that version 3.90.3 does encode with more (sometimes a lot more) MS frames. This did not translate into smaller files either. In all cases, version 3.90.3 created larger files using more MS frames and took up to 33 percent longer to encode with no ReplayGain tag, but it did have a higher average bitrate.

Quote
LAME chooses which will be the most efficient for the frame it's currently looking at.

besides, when the bitrate limit is hit on a very difficult sample, an MS artefact will often be harder to notice than a LR artefact (depending on the listener's hearing).  i myself notice glitches that differ from 1 channel to the other before i'd notice a glitch that is more centered in the stereo field, such as one MS would give.

That makes sense.
Ronn

Joint Stereo Reverting to Stereo (Clarification)

Reply #14 –
Quote
Quote
can i ask what sort of music is giving you 90% LR stereo frames?  remember it's all relative.

The above was Two-Mix.

For those of you not familiar with Two-Mix they are a JPop band.  Pretty complex music usually.  If you don't mind my asking, which song was that so I can compare results?

It's been my experience that the extreme preset almost always gives mostly LR stereo frames instead of MS.
Nero AAC 1.5.1.0: -q0.45

Joint Stereo Reverting to Stereo (Clarification)

Reply #15 –
Quote
Quote
Quote
can i ask what sort of music is giving you 90% LR stereo frames?  remember it's all relative.

The above was Two-Mix.

For those of you not familiar with Two-Mix they are a JPop band.  Pretty complex music usually.  If you don't mind my asking, which song was that so I can compare results?

Just to note, the histogram for Two-Mix was originally encoded with version 3.96. The track was...

Two-Mix - Fantastix - 02 - True Navigation

I particularly like the beginning with those very intense bells. 

Quote
It's been my experience that the extreme preset almost always gives mostly LR stereo frames instead of MS.

Yup, MS frames appear not to even be marginally important to the extreme preset for a many encodes. For example, the B52's track I mentioned earlier was...

B52s - Nude On The Moon (Anthology) - 1x03 - Rock Lobster

This encode had less than 1% MS frames for a near 7 minute track. Although, now I'm starting to wonder if I should have been using version 3.90.3 instead of 3.96 to increase the amount of MS frames per encode (when camparing the same source). More MS frames should in theory create a better encode because, as you originally pointed out, MS is a lossless procedure.
Ronn

Joint Stereo Reverting to Stereo (Clarification)

Reply #16 –
OK I just encoded the same track True Navigation in LAME 3.96.1 with presets standard, extreme, and insane.

Standard
32 [  87] %*
128 [  96] %*
160 [ 520] %%******
192 [2863] %%%%%%%%%%%%%%%%%%%%%%%%%*******************
224 [4365] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%**********************
256 [1437] %%%%%%%%%%%%%%********
320 [ 345] %%%%**
average: 216.6 kbps  LR: 5660 (58.27%)  MS: 4053 (41.73%)

Extreme
32 [  83] %*
128 [  17] %
160 [  44] %
192 [ 301] %%%%*
224 [1486] %%%%%%%%%%%%%%%%%%%%%*
256 [3324] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%**
320 [4458] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
average: 275.9 kbps  LR: 9267 (95.41%)  MS: 446 (4.592%)

Insane
average: 320.0 kbps  LR: 9629 (99.14%)  MS: 84 (0.8648%)

I guess LAME just uses fewer MS frames at higher bitrates.  I'd say on the insane sample the 84 frames that use MS probably had to in order to achieve the maximum quality.  Two-Mix music rarely has the same sounds coming from both channels so I'm not really surprised in the amount of LR frames personally even in preset standard.

I'd encode a sample with 3.90.3 for comparison but I am tired and don't have 3.90.3 ready to go for encoding.  Probably because I mainly use Vorbis 
Nero AAC 1.5.1.0: -q0.45

Joint Stereo Reverting to Stereo (Clarification)

Reply #17 –
Hi,

Sorry for the delay, I've been extreme(ly  ) busy.

Quote
OK I just encoded the same track True Navigation in LAME 3.96.1 with presets standard, extreme, and insane.

Here is True Navigation using three different versions of LAME (3903, 3960 and 3961). All encodes used the extreme preset:

LAME version 3.90.3[/color] MMX  (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), SIMD, SIMD2
Using polyphase lowpass  filter, transition band: 19383 Hz - 19916 Hz
Encoding TN.wav to TN(3903).mp3
Encoding as 44.1 kHz VBR(q=2) j-stereo MPEG-1 Layer III (ca. 7.3x) qval=2
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
  9710/9713  (100%)|    0:24/    0:24|    0:24/    0:24|   10.281x|    0:00
 32 [  84] %
128 [  22] %
160 [  57] %
192 [ 466] %%%***
224 [1144] %%%%%%********
256 [2305] %%%%%%%%%%%%%%%%***********
320 [5635] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%***********
average: 283.5 kbps   LR: 6720 (69.19%)   MS: 2993 (30.81%)

Writing LAME Tag...done

Encoded Size: 8,983,676 bytes

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
LAME version 3.96[/color] MMX  (http://www.mp3dev.org/)
CPU features: MMX (ASM used), SSE, SSE2
Using polyphase lowpass filter, transition band: 19383 Hz - 19916 Hz
Encoding TN.wav to TN(3960).mp3
Encoding as 44.1 kHz VBR(q=0) j-stereo MPEG-1 Layer III (ca. 5.7x) qval=3
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
  9710/9712  (100%)|    0:18/    0:18|    0:18/    0:18|   13.746x|    0:00
 32 [  83] %*
128 [  18] %
160 [  43] %
192 [ 307] %%%%*
224 [1478] %%%%%%%%%%%%%%%%%%%%%*
256 [3337] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%**
320 [4447] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
average: 275.8 kbps   LR: 9266 (95.40%)   MS: 447 (4.602%)

Writing LAME Tag...done
ReplayGain: -7.5dB

Encoded Size: 8,740,664 bytes

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
LAME version 3.96.1[/color] (http://lame.sourceforge.net/)
CPU features: MMX (ASM used), SSE, SSE2
Using polyphase lowpass filter, transition band: 19383 Hz - 19916 Hz
Encoding TN.wav to TN(3961).mp3
Encoding as 44.1 kHz VBR(q=0) j-stereo MPEG-1 Layer III (ca. 5.7x) qval=3
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
  9710/9712  (100%)|    0:18/    0:18|    0:19/    0:19|   13.574x|    0:00
 32 [  83] %*
128 [  17] %
160 [  44] %
192 [ 301] %%%%*
224 [1486] %%%%%%%%%%%%%%%%%%%%%*
256 [3324] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%**
320 [4458] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
average: 275.9 kbps   LR: 9267 (95.41%)   MS: 446 (4.592%)

Writing LAME Tag...done
ReplayGain: -7.5dB

Encoded Size: 8,743,490 bytes


Due to a command shift in version 3.94 beta and up, the qval values in all three versions is suppose to be the same. 

Quote
I guess LAME just uses fewer MS frames at higher bitrates.  I'd say on the insane sample the 84 frames that use MS probably had to in order to achieve the maximum quality.

Ah, but see that was my original concern when I used extreme and observed so few MS frames being encoded. It appeared, because of extreme's purpose, that quality was coming from the use of LR frames. Particularly in the B52 example were OVER 99 percent were LR frames. In the case of True Navigation, it seemed as if the 446 MS frames (out of 9710) were there because they could be reduced in size and not necessarily because the quality would come from making them MS frames. The same with the 84 MS frames you observed using the Insane preset.

The earlier B52's Rock Lobster example is basically a Stereo (LR) encode with a Joint-Stereo (MS) tag. Calling it Joint-Stereo is more of a formality.

Quote
Two-Mix music rarely has the same sounds coming from both channels so I'm not really surprised in the amount of LR frames personally even in preset standard.

Is that because of the discreet stereo that's present? To which more LR frames were used to preserve the stereo image?

BTW, 3903 used more MS frames. The VBR quality mode (3903=VBR(q=2) vs 396x=VBR(q=0)) used 30 percent MS frames, compared to just 4 percent MS frames for 396x. So out of curiosity I added -V0 to the 3903 extreme preset (to sort of balance it out):

LAME version 3.90.3[/color] MMX  (http://www.mp3dev.org/)
CPU features: i387, MMX (ASM used), SIMD, SIMD2
Using polyphase lowpass  filter, transition band: 19383 Hz - 19916 Hz
Encoding TN.wav to TN(3903-ev0).mp3
Encoding as 44.1 kHz VBR(q=0) j-stereo MPEG-1 Layer III (ca. 5.7x) qval=2
    Frame          |  CPU time/estim | REAL time/estim | play/CPU |    ETA
  9710/9713  (100%)|    0:32/    0:32|    0:32/    0:32|   7.7895x|    0:00
 32 [  84] %
128 [  15] %
160 [  13] %
192 [ 123] %*
224 [ 643] %%%***
256 [1453] %%%%%********
320 [7382] %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%**************
average: 299.4 kbps   LR: 6703 (69.01%)   MS: 3010 (30.99%)

Writing LAME Tag...done

Encoded Size: 9,489,288 bytes


The size certainly increased. When the stock extreme preset VBR(q=2) mode was used, the size was 8,983,676 bytes. Interestingly, the amount of MS frames basically remained identical:

Lame3903 --alt-preset extreme:
average: 283.5 kbps  LR: 6720 (69.19%)  MS: 2993 (30.81%)

Lame3903 --alt-preset extreme -V0:
average: 299.4 kbps  LR: 6703 (69.01%)  MS: 3010 (30.99%)


Could this mean that 3903 can find more frames which compress better then 396x or that 3903 didn't detect the complexity correctly and accidently allowed more MS frames to be encoded or (for lack of a better description) 396x gave up to quickly and just used LR to process the encode quicker?

Finally, what's up with the total source frames not equaling the total encoded frames? Not a big deal, but in all three versions of LAME the value of the last encoded frame was short and could not have reached 100 percent as indicated.     

I rambling to much, need sleep....
Ronn

Joint Stereo Reverting to Stereo (Clarification)

Reply #18 –
Quote
it seemed as if the 446 MS frames (out of 9710) were there because they could be reduced in size and not necessarily because the quality would come from making them MS frames.

Those two things are equivalent. Reducing the size means increasing the quality for a given size.

Quote
Finally, what's up with the total source frames not equaling the total encoded frames? Not a big deal, but in all three versions of LAME the value of the last encoded frame was short and could not have reached 100 percent as indicated.

This is only an infortunate display issue and is only impacting what you see, not what you hear.

Joint Stereo Reverting to Stereo (Clarification)

Reply #19 –
I also have some trouble understanding JS.

Lyx said:
Quote
Those 5% can make a difference when at the same time the 320kbit limit is hit - because then you have a situation of limited bandwidth and in such a case M/S can pack more info into the available space of that frame.


What do you mean by bandwidth? I thought MP3's threw out very high and low frequencies and then cut down the number of bits per second (compared to .wav's).

What is MS. L-R is easy to understand because of headphones and speakers are usually set up in this way
<==== Hydrogen Audio Bomb

Joint Stereo Reverting to Stereo (Clarification)

Reply #20 –
in this case, bandwidth refers to the maximum transfer rate (320kbps) rather than the lowpass cutoff, which is fixed in the LAME presets (different for each preset).

MS = Middle-Side.

middle = Left + Right (this is a straight mono mix, basically)
side = Left - Right

as far as speaker placement goes, it's hard to imagine, but bear with me:

it's similar to having one speaker going out in all directions (or for simplicity just facing forward), and the other speaker taken out of it's box, so there's no longer a baffle, and turned on it's side.  then place it exactly in the same place as the mono speaker.

this works better with microphones, and is actually quite a popular miking technique:  grab an omnidirectional mic (picks up equally from all directions), and a figure-eight mic (picks up everything in front or behind it).  then turn the figure-eight mic around 90 degrees, so it's sensitive areas are pointing left and right.  with a bit of clever wiring, you can get a regular stereo signal out of these 2 mics.

as far as audio coding goes, LR and MS are both lossless, but MS presents significant advantages in coding efficiency, because most stereo signals are very co-related between channels, and an MS coding takes advantage of this, giving the same quality for less bits (and therefore more quality for the same bits, which is useful at complex parts of the music when the 320kbps brick wall is hit... you need all the efficiency you can get).