Problems With Joint Stereo (LAME), high tunes distortion, LAME 3.93 Options

[NumLOCK]

Sonny: Have you tryed only "--alt-preset standard" setting wihtout any other settings of yours? How does it sounds against your own settings?

Yes he has, you should read his initial post:

Here what happen to me practically :
Sample encoded in 320kbps and TS (archive quality, 1.77Mo). Very close to the WAV sample.
Sample encoded with the --alt-preset standard, in JS (195kbps, 1.07Mo). You can hear the distorsion of high tunes even with a crappy sound board like mine !
Sample encoded with the --alt-preset standard -V 3, in JS (176kbps, 0.98Mo). In this file, I’ve lowered a bit the quality of the encoding (standard is ‘-V 2’). The overall quality should be good but the distorsion is now HUGE. 
[...]

[Sonny]

The only reason I asked is because Vorbis -q 6 uses lossless stereo coupling.  I want to make sure that this isn't in your head.

~~~~~~

[tigre]

I'd like to summarise:

There are 3 possibilities


1. The distortions are caused by equipment.

The main part of cymbal-like sounds is in high frequencies. These frequencies are reproduced by lame somewhat inaccurate = distorted, especially if there are other sounds that would mask the distortions. If equipment amplifies these high frequencies for some reason, these distortions can become clearly audible. Another equipment-related possibility would be aliasing: 16-17.5 kHz sounds would be 'mirrored' to some lower frequencies where they are audible much better.

To test it you can do this:
Use the sample from this thread. It contains low volume dialing sounds and high volume 19-21kHz sweeps that should be inaudible (try at low volume first, this might damage your equipment/hearing). If you hear something else than the dialing sounds, your playback chain needs some modification (audio player with decent resampling to 48kHz etc.)
Or you could decompress the mp3 files to .wav, burn them to audio CD-R and listen on some standalone equipment / discman etc. Problems still there? ... ->

2. Your hearing is exceptional in high frequency range. There are some "how high can you hear" threads with samples for testing, but you need some good equipment/settings (test see 1.), otherwise you'll hear aliasing and think you hear 22kHz sounds.

3. At least some of your findings could be based on immagination (don't take this personal, please - it's just the only possibility left from a scientific point of view). Unless you are sure that your playback chain isn't responsible for the problems, this question is uninteresting anyway.

[Sonny]

ok, i think i've finally found out what goes wrong. i thought it was my equipement, particularly my sound board. i was right... and wrong.
i've made this simple thing : i usually pass througt my amp (pioneer xd-z54t, good old stuff.... well i thought), so i've plugged in my headphones directly to my computer and listen for the 284 times to the JS/TS samples... and omg !! i could picked up the JS sample as I did the previous 283 times!
i did another test : i've uploaded 2 samples on my mp3 player (creative zen nx w/ new headphones, standard ones are big crap) and made the test : again, i couldn't picked up the JS sample.

so the distortion is created by my amp or the cable i use.
but how come a cable or an amp create distortion of high tunes on a JS sample and do not with TS ?! any idea ?
i'll make a final test on a friend's computer as i said before, just to be sure.

anyway, sorry for disturbing and big thanks for helping

ps: IT WAS NOT MY IMAGINATION !!!

[2Bdecided]

Thanks for letting us know. Glad you found the cause! See - JS isn't so bad after all!

Maybe your amp has some "enhancer" or "wide stereo" or "EQ contouring" or "loudness" or "virtual surround" switched in by default?

It's well known (and mentioned many times on this board) that certain DSP (including extreme EQ and stereo processing) is incompatible with psychoacoustics. This is because psychoacoustic coding uses a rough model of human hearing, but what is eventually sent to your ears (after extreme DSP) may be nothing like what the model in the coder "heard" before the DSP.

The ideal psychacoustic codec would only store what you could hear, so any post-processing would cause problems. Most codecs leave some headroom (i.e. parts of the sound that they think you probably can't hear are still left intact) because they're not good enough to be 100% certain of what is audible and what is not. The more cautious a codec is, the more "excess" it will encode (and the more bits it will use). Forcing Simple stereo will prevent a codec from removing redundency (statistical or psychoacoustic) between the two channels, but this can be a bad thing in some circumstances (See all the JS threads in the FAQ).

If you are going to post-process the audio, or apply extreme DSP, the best solution is to use lossless, or near-lossless encoding. Failing that, tune an encoder to give some leeway in whatever domain you intend to post process: reduce spectral masking if you're going to use very heavy EQ, and reduce stereo masking is you're going to process the channel difference. Better still, do whatever you have to do before encoding, and listen to the encoded version without any DSP.

(I realise you didn't know you were using DSP, but this may be useful information if you decide to want to in the future, or are thinking that maybe lossy is not too useful with this limitation. FWIW Most of the time, most DSP isn't a problem.)

Thanks again for coming back to say you'd found the problem. You'd be surprised how many people don't!

Cheers,
David.

This post has been edited by 2Bdecided: Dec 9 2003, 14:21

[n68]

gday..

how come a cable or an amp create distortion ?

1. output gain. (depends on what soundcard you have.)
the cable itself.. is not likely to create such spesific problems..
even it is bad Q.

(btw. my answer on JS. still have a point.
the nature of channel coupling have that effect..
depending on the input signals.)

[Hanky]

This turned out to be a quite informative thread after all. One thing that has to be checked still is what tigre found out about the high frequency dropouts in --alt-preset CBR 128. Please note that he did not hear dropouts, just saw them in graphical analysis, which we all know tells little about audible artifacts. Since my HF hearing is tragically bad, I can't do the tests myself. Could someone with decent hearing capabilities please (re)do some tests with the 128kbps preset. Does he/(she??) hear 'ringing' effects ? If yes, we should probably lower the default --lowpass frequency for that bitrate range. If no, everyting is still ok and we don't need to worry about it.

[2Bdecided]

We should put these threads in the FAQ.

Under a section "people who thought they heard a problem with --alt-preset standard"...

Sort them by the final result, i.e.
"new problem sample found"
"other reason found for apparent problem"
"original poster vanished"

The relative numbers of each would give some kind of indication to new members of the likelihood that there really is a problem. And, of course, anyone who bothered to read them would probably find the answer to their problem.

Cheers,
David.

[tigre]

2bdecided, I don't think it's a problem related to stereo.

Using 'true' stereo wastes that much bits on the sample that there's almost nothing left for > 16kHz information, while with joint stereo much more information 16 - 17.5kHz is stored (see the attatched image - 1st CEP sreenshot is 'true' stereo, 2nd is joint stereo). I think the reason is aliasing, probably caused by the amp. If this causes the 16-17.5kHz information to become mirrored to e.g. 4.5 - 6kHz, it's obvious why joint stereo sounds worse.

Attached thumbnail(s)

 

[Pio2001]

Drop outs ? They seem like "noise-gate" filtered frequecies, or rather in this case "ATH filtered" frequencies... When it is too weak, it is left out, creating holes in the graph.

Sonny, is your amplifier part of a Home Theatre system ? Letting it decode stereo as Dolby Pro-Logic can lead to such problems.

[Sonny]

no it's not, it's just a 'normal' amp.
i'm still searching a way to avoid those distortions using dsp after the encoding process (whithout changing my amp!)

[KikeG]

What operating system, player, and output plugin/type are you using? I say that because for example, directsound output plugin under Win9X causes horrible distortions and makes audible things that with good quality audio are not audible, I've verified this myself.