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: Lame's Stereo VS Joint Stereo (Read 10075 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Lame's Stereo VS Joint Stereo

I read about joint-stereo and I found that it saves space without any noticeable quality difference.
I tried ripping a single track with Eac and Lame 3.93.1 using the VBR method and joint stereo (-m j -q 0 -v -V 4 -b 80).
Then I ripped the same track again using stereo (-m s -q 0 -v -V 4 -b 80). Though the track size with joint stereo should be smaller it didn’t: the joint stereo mp3 was 4,88 mb while with stereo it was 4,61 mb!!!
Then I tried again with a higher VBR quality setting (-V 2). This time the gap was smaller but still it doesn’t make sense. Has or can anybody else try that to see the results? Is it that I haven’t understand something?

Lame's Stereo VS Joint Stereo

Reply #1 –
Sorry for posting the topic 3 times...

Lame's Stereo VS Joint Stereo

Reply #2 –
Since you are using VBR with the same quality level your assumption (joint-stereo files to be smaller) makes sense. I would have expected the same behaviour. - An interesting find out that we're both wrong.

I'm pretty sure (as well as others of HA) you'll get more quality when using joint-stereo for CBR files, though.
It certainly depends on the encoder implementation but it just makes sense. The encoder can choose between L/R and M/S. - Whichever gives a better noise-mask-ratio is the better encoding method for the current frame.

If interchannel correlations are present entropy can be concentrated mostly in M (in-phase correlations) and somtimes S (out-of-phase correlations) and M/S will be a more optimal representation.

IMHO, size(joint-stereo VBR file) > size(stereo VBR file) (at the same quality level) may imply inconsistencies in the VBR routines / psychoacoustic model. I'm not an lame developer, so I might be wrong.

bye,
Sebi

Lame's Stereo VS Joint Stereo

Reply #3 –
Quote
I'm pretty sure (as well as others of HA) you'll get more quality when using joint-stereo for CBR files, though.
[a href="index.php?act=findpost&pid=226738"][{POST_SNAPBACK}][/a]


Using Joint Stereo won't affect quality at all. It is only a matter of saving disk space.

Lame's Stereo VS Joint Stereo

Reply #4 –
Using CBR, JS can't save space (filesize is linked to the targeted bitrate, and not on channel mode), but only have an impact (generally positive) on quality.

Lame's Stereo VS Joint Stereo

Reply #5 –
stereo r old fashion's in 80's

Lame's Stereo VS Joint Stereo

Reply #6 –
Quote
Using CBR, JS can't save space (filesize is linked to the targeted bitrate, and not on channel mode), but only have an impact (generally positive) on quality.
[a href="index.php?act=findpost&pid=226743"][{POST_SNAPBACK}][/a]


OK, thanks. Never thought that JS would improve encoding quality.

Lame's Stereo VS Joint Stereo

Reply #7 –
I may be wrong but I believe that jstereo improves quality by allowing the bits which would normally be used for the stereo image to be used for the sound instead. Or something like that.

Lame's Stereo VS Joint Stereo

Reply #8 –
I confirm JS greatly increase the quality of the encoding.
The centre of the stereo image is encoded once, rather than two times. Therefore you gain space for quality.
Another benefit is the precision of the centre. Sounds in the centre of the stereo image (it's particularly noticeable with the voice) are perfectly reproduced in the centre. It's not necessary the case with stereo encoding. For me it's the easier way to hear the difference between JS and S encoding.

Lame's Stereo VS Joint Stereo

Reply #9 –
Quote
I tried ripping a single track with Eac and Lame 3.93.1 using the VBR method and joint stereo (-m j -q 0 -v -V 4 -b 80).
Then I ripped the same track again using stereo (-m s -q 0 -v -V 4 -b 80). Though the track size with joint stereo should be smaller it didn’t: the joint stereo mp3 was 4,88 mb while with stereo it was 4,61 mb!!!


I think V4 always uses joint stereo, so perhaps its overrideing the "-m s"  in your second commandline?  Maybe you could check the output file and see if its actually Full Stereo?

Lame's Stereo VS Joint Stereo

Reply #10 –
Quote
I think V4 always uses joint stereo, so perhaps its overrideing the "-m s"  in your second commandline?  Maybe you could check the output file and see if its actually Full Stereo?
[a href="index.php?act=findpost&pid=226811"][{POST_SNAPBACK}][/a]


With Lame, joint stereo is the default mode for VBR encoding with -V > 4, and stereo for -V <=4. You can override all of these defaults by specifing the mode on the command line.

Anyway I checked the files and the one is stereo and the other joint stereo.

I found some more things about joint stereo but I can't understand what is the mid and side stereo (mentioned by SebastianG too).

Lame's Stereo VS Joint Stereo

Reply #12 –
Quote
With Lame, joint stereo is the default mode for VBR encoding with -V > 4, and stereo for -V <=4.

It would be surprising. From where comes this information? and for which lame encoder?

I've just tried to encode a track with -V 0 (aka --preset extreme) and -V 6 (~115 kbps) and both are in joint stereo mode (using stereo frames when needed). I'm using lame 3.97a3

Lame's Stereo VS Joint Stereo

Reply #13 –
Quote
It would be surprising. From where comes this information?

I think it is EXTREMELY old bug in Lame documentation.
Need to download full Lame packae and check...
EDIT:
To eliminate misunderstanding about stereo modes used in layer 3.
There are 4 modes:
Dual channel (understandable)
Stereo (bits per channel can vary)
Joint Stereo (use mid-side coding technique, losssless by idea, but it can hurt, if switch algorhytm between stereo and joint stereo is bad)
Intensity Stereo (use mid-side coding technique, lossy by defalut)
EDIT2: concept of Joint Stereo a bit

Lame's Stereo VS Joint Stereo

Reply #14 –
Quote
It would be surprising. From where comes this information? and for which lame encoder?


I use Lame 3.93.1 which, I think, is the latest stable release. I read it from the documentation which comes with it. It's the file named USAGE (the file has no extension).

Lame's Stereo VS Joint Stereo

Reply #15 –
Quote
I use Lame 3.93.1 which, I think, is the latest stable release. I read it from the documentation which comes with it. It's the file named USAGE (the file has no extension).
[a href="index.php?act=findpost&pid=226847"][{POST_SNAPBACK}][/a]

Thanks for the answer.
You're wrong about the latest stable release: it's 3.96, and not 3.93.1, which is pretty old (20 months). Latest release have a different VBR scale, including now the previous --alt-preset VBR (medium, std, extr...). In other words, -V x were tweaked a lot.

Lame's Stereo VS Joint Stereo

Reply #16 –
what about the behavior of the current LAME version ? I currently can't test it.

crystal clear: maybe you want to verify the current LAME version (whether the size of a joint-stereo VBR file will still be greater than the size of the simple stereo VBR ile at the same quality level)

If it's still the case then it's likely that the current L/R <-> M/S switch strategy and scalefactor selection strategy needs to get improved.

bye,
Sebi