alt-preset standard problem samples, with Lame 3.93.1 and before. Options

[JohnV]

There are lots of cases when alt-presets or any mp3 are not transparent, but I gather few "easy" but distorted cases which should be used to further tweak and check alt-preset standard quality.

I'll put more samples here in the future so that those are easier to find and can be found from the same place.
My intention is not to list clips with very hard attacks, other known extreme clips or clips which cause great trouble for every codec.
More like "easier" cases which in my opinion should be handled better by alt-preset standard.

http://static.hydrogenaudio.org/extra/samp...es/sophia2.flac
Clip found by Wombat. Introduces added noise from 1.2-2sec and 4-6sec.

http://static.hydrogenaudio.org/extra/samp.../vangelis1.flac
Although soft attacks, still introduces pre-echo hiss (most noticeable in the beginning).

http://static.hydrogenaudio.org/extra/samp...dogwhistle.flac
Dogwhistle is chirping.

http://static.hydrogenaudio.org/extra/samp.../liebestod.flac
Audible distortion in the background.

http://static.hydrogenaudio.org/extra/samp...ples/birds.flac
In all 4 vbr presets distortion is added, mostly with the "e" when she sings "become".

http://static.hydrogenaudio.org/extra/samp...mples/erhu.flac
Very bad distortion at about 3.5 sec.

This post has been edited by JohnV: Dec 5 2002, 09:22

[Wombat]

I have to add another sample from the kind of Sophia.

It produces the same sandpaper sound with standard, but
more surprising - extreme generates even audible plops.

The file is here:

http://www.halke.net/files/Birds.flac

Wombat

Edit:
In all 4 vbr presets distortion is added, mostly with the "e" when she sings become.

aps adds a bit distortion
ape surprisingly adds small plops (or how should i call it?) but less distortion

apfs + apfe are adding both the most distortion
_______
Edit: Just switched to flac, also in Liebestod it is very easy to hear the same prob
http://www.halke.net/files/Liebestod.flac

[Dibrom]

Just so everyone knows, I'm not going to have much time to look at these clips for awhile so modifications to alt-preset to help with these (if possible) may be a ways off. It is helpful to list the clips in this manner though so that they are easier to find in the future..

[Wombat]

Hi dibrom!

I strongly believe it has something to do with nssafejoint.

Lets hope you find some time soon.

Wombat

[Wombat]

Damn! The solution was that easy, i had to play around with that
Liebestod sample to try a switch we all know for ages.

--ns-bass -8 DOES the trick!! The distortions are nearly gone!

Not only with that, sophia and Birds are cured also.

Does ns-bass affect anything else with alt standard deep inside the code?

Wombat

[ff123]

I verified Wombat's Liebestod problem sample with lame 3.91 --alt-preset standard. It's got low frequency plops, just as wombat says (ABX 10/10) which go away when --ns-bass -8 is added.

ff123

[Gabriel]

I'm starting to believe that the new ATH I introduced have a too low sensitivity in the low freq area.
This needs to be confirmed, but if confirmed the cure will be easy.

[Dibrom]

Originally posted by Wombat
Damn! The solution was that easy, i had to play around with that
Liebestod sample to try a switch we all know for ages.

--ns-bass -8 DOES the trick!! The distortions are nearly gone!

Not only with that, sophia and Birds are cured also.

Does ns-bass affect anything else with alt standard deep inside the code?

Wombat

Hrmm, interesting. --ns-bass -8 shouldn't really cause any problems though, other than causing the bitrate to increase. I'll try to take a look at this sometime this week, maybe it would be possible to implement a fix without having to resort to using --ns-bass all the time on these samples..

[Dibrom]

Originally posted by Gabriel
I'm starting to believe that the new ATH I introduced have a too low sensitivity in the low freq area.
This needs to be confirmed, but if confirmed the cure will be easy.

This could be possible, though I'd be a bit surprised if it was solely related to the ATH. Though, if it is, it should be easy to test just by using --athlower..

[Dibrom]

I just spent some time listening to the Birds and Liebstod samples, and it seems to me that -Z makes more of a difference, especially on Liebstod, than --ns-bass -8 does. This would basically make it another issue with noise shaping 2 then... It could probably be fixed fairly easily..

In all honestly... I think it might be better to look at nspsytune2 also, instead of just trying to hack in another tweak to the current psymodel.. I think at least that JohnV said that an early version of nspsytune2 sounded better on the Birds sample (though it also increased the bitrate a lot).

[JohnV]

Originally posted by Dibrom
think at least that JohnV said that an early version of nspsytune2 sounded better on the Birds sample (though it also increased the bitrate a lot).

I think it was Sophia2, dont remember if i tested birds, but anyway..

[Wombat]

Hi all!

Thanks for testing this ff! No one seemed to be interested otherwise!

Dibrom you´re right! -Z makes it even better than only nsbass -8 but
also bigger.

Is there an easy way to test around with nspsy2? The last time i did i
had to do everything in two passes. Also it doesn´t make much sense to test
nspsy improvements with standard switches. The produced nspsy2 files should
at least implement the more tuned short block allocation from the alt presets
to have a fair compare, or does nspsy2 implement this?


regards

Wombat

[JohnV]

Wombat,
in --aps, putting -Z uses noiseshaping-type 1 always. But it's a bit wrong to say that --aps uses (only) ns-type 2, because it uses both 1 and 2. When you use -Z, it uses 1 only.

[outscape]

one track which is essential to listen to when tweaking the --alt-preset modes is "adiemus - adiemus". killer sample for pre and post echo, even when using --ap extreme

[JohnV]

Updated with erhu -sample.
http://www.hydrogenaudio.org/forums/index....=ST&f=16&t=4687