Lossless vs 128kbps mp3 vs 320kbps mp3 blind test - Page 7

Quote:

Originally Posted by chewy4 

That would make sense if it does help save space, I wasn't really aware of this setting before though. 

There's a fair bit of confusion surrounding Joint Stereo, thats because there are 2 types of Joint Stereo. There's the newer version of Joint Stereo and there is the older version which is now known as "Intensity Stereo".

Intensity Stereo is lossy and LAME uses it for bitrates under 128kbps. But for bitrates at 96kbps and above, it uses the newer version of Joint Stereo, which contains the same lossless algorithm that lossless formats such as FLAC and APE use to help save space.

Full Stereo is pretty much obsolete, as there is no quality advantage of using it over Joint Stereo.

Also, it only saves space if you use it in VBR mode. You'll get the same sound quality that you would if you used Full Stereo, but the file sizes will be smaller.

If you use it in CBR mode however, the file size will be the same as it would if you used Full Stereo, but the quality will be greater.

Just for comparison, here is a track encoded at 128 stereo and 128 joint. (both created from lossless source)

Stereo

Joint Stereo

On both spectrograms, the frequencies cut off somewhere between 16khz to 16.5khz. But in the Joint Stereo spectrogram, it contains more frequency information within that limit.

You may have to enlarge the pictures to see the difference.

Quote:

Originally Posted by Sound Quest 

 

Intensity Stereo is lossy and LAME uses it for bitrates under 128kbps. But for bitrates at 128 and above, it uses the newer version of Joint Stereo, which contains the same lossless algorithm that lossless formats such as FLAC and APE use to help save space.

 

Full Stereo is pretty much obsolete, as there is no quality advantage of using it over Joint Stereo.

Source? Afaik this is wrong.

Intensity stereo is only used for extremely low bitrates that nobody uses for music anyway.

Joint stereo means that the encoder can switch between L/R and M/S (mid/side) stereo coding.

This is the default for <= 128 kbps and VBR quality > 4.

Higher bitrates / VBR quality uses normal stereo (L/R).

Joint stereo in M/S mode only works well if the signal has little stereo separation and in general if the encoder doesn't have to switch between the modes too often. Otherwise it can cause audible artifacts, i.e. worse quality.

And on spectrograms: don't use them to judge audio quality. Use ABC/HR for that.

Quote:

Originally Posted by xnor 

Source? Afaik this is wrong.

 

Intensity stereo is only used for extremely low bitrates that nobody uses for music anyway.

Joint stereo means that the encoder can switch between L/R and M/S (mid/side) stereo coding.

This is the default for <= 128 kbps and VBR quality > 4.

 

Higher bitrates / VBR quality uses normal stereo (L/R).

 

Joint stereo in M/S mode only works well if the signal has little stereo separation and in general if the encoder doesn't have to switch between the modes too often. Otherwise it can cause audible artifacts, i.e. worse quality.

 

And on spectrograms: don't use them to judge audio quality. Use ABC/HR for that.

I wasn't trying to use the spectrograms to prove my point, I was simply showing that using Joint Stereo allows the encoder to retain more frequency information below the cut off. The spectrogram clearly shows this.

Joint Stereo is lossless. Joint Stereo can never sound worse than full stereo, regardless of channel seperation. The encoder doesn't make any guesses when converting from Stereo to JS, it uses the same mathematical precision that lossless formats such as FLAC and APE use to save space.

The only time it isn't 100% precise is when you use a bitrate below 96kbps, which will activate Intensity Stereo, which is lossy.

Think of "Joint Stereo" vs "Full Stereo" like "FLAC" vs "WAV".

EDIT Sorry, mean't to say that intesity stereo is used at 96kbps and under.

Ahm, joint stereo in the lame encoder means that the encoder can switch between L/R and M/S coding.

When you say lossless, you mean that coding a L/R signal as M/S can be reversed, right? Because lossy compression is applied just the same as is with L/R coding.

When choosing the joint stereo option you do not force lame into M/S coding. It only switches to M/S coding when the stereo separation is low enough to save space because then most of the signal is in the M channel.

If you force L/R at low bitrates then the stereo image will be preserved at the cost of audible distortion.

If you force M/S at low bitrates then the stereo image there will be less audible distortion but the stereo image will collapse.

The output of lame should show you how many frames were L/R and how many M/S coded. Try the joint stereo option with different bitrates (128, 192, 320) or even VBR modes (5, 2, 0) and look at the number of M/S frames.

I've done this before, but i'm afraid I don't quite understand your point.

When you use the Joint Stereo Mode in LAME at bitrates 96 and above, it converts from stereo to joint stereo losslessly, before applying any lossy compression. LAME encodes each channel independently.regardless of whether you use L/R or M/S mode.

The encoder will take the lossless stereo file, then encode it to a lossless joint stereo file, then it will encode it to MP3. (Encoding the mid channel and the side channel independently).

The number of M/S frames in an mp3 file is meaningless. You could have a track thats 95% Stereo and only 5% Mono, but using Joint Stereo over Full Stereo would still make perfect sense. It just means that 95% of the frames will be in the side channel and 5% of the frames will be in the mid channel. This will either save you a small amount of space (if VBR is used), or will give the encoder extra bits to improve the quality (if CBR is used).

Let me ask you, if you think Joint Stereo is lossy, then what makes it different than Intensity Stereo? Why would there be 2 types of Joint Stereo, if they are both lossy?

I was not saying that M/S coding is lossy, read my post again, I was asking if you understand what lossless means in this case, which you seem to do.

Again, the encoder does not convert the whole file to M/S unless you force (option -m f) it to.

I did a simple test:

-V 5:

   kbps        LR    MS  %     long switch short %
  138.8       64.4  35.6        81.7   9.9   8.5

-V 2:

   kbps        LR    MS  %     long switch short %
  199.0       79.1  20.9        81.7   9.9   8.5

-V 0:

   kbps        LR    MS  %     long switch short %
  259.7       85.5  14.5        81.7   9.9   8.5

-b 320:

   kbps        LR    MS  %     long switch short %
  320.0       86.8  13.2        83.3   9.3   7.4

As the bitrate gets higher, lame chooses to convert less and less frames to M/S for reasons mentioned above (preserving stereo image is more important).

Btw, my lame version (3.99.5) always chose -m j as default. So the documentation seems to be outdated.

Ok, I see where you're coming from now. Apologies for the misunderstanding. I would still however recommend Joint Stereo mode at birates as low as 128kbps/-V5 CBR. Not forced, but simply set by default.

At higher bitrates 192kbps/-V2 and above, full stereo should be perfectly fine in most cases anyway.

I'm not sure I understand what the descriptions of sounding fuller or having more body actually mean when comparing a lossless file to a lossy version.  So, what, the upper-bass frequencies are louder?  For any that make such claims, would it be possible to ABX the lossless file and the 320 kbps file 15 times and post the results?  Being fuller and having more body seems applicable when eating too much pizza, but I don't see how it applies to sound quality with compressed audio unless, perhaps, the bitrate is exceptionally low?

http://www.head-fi.org/a/describing-sound-a-glossary