LAME problem samples - samples, list of problematic samples Options

[Gabriel]

This topic should be a centralized place to post problematic samples for LAME.

You should post here:
*sample link
*short description of the problem
*LAME settings regarding the problem
*LAME version used

This thread is only to collect samples. Discussion should occur in this thread.

This post has been edited by Gabriel: Dec 9 2005, 10:13

[Wombat]

This sample adds distortion that sound like sandpaper. Mostly noticeable on the first "e" when she sings become.
This is for lame 3.90.3 aps but more problematic on 3.97b -V2. Adding --vbr-new worsens it even more.

Birds.flac

This post has been edited by Wombat: Dec 1 2005, 22:20

[Wombat]

Here are rythmic plops added mostly on second 2-3 and louder at second 9-10.
Don´t know if plops descripe this sound correctly. Another impression would be some "pumping" noise is added. Once spotted, it is really loud.
This happens with 3.97b2 and 3.98a2 -V2 --vbr-new

It is velvet.wav and is offered at Sourceforges Lame Gpsycho test sample page

[halb27]

This is a sample I found in a thread but I'm sorry I don't remember it and was not able to find it again:
Trumpet sample

It is seriously distorted when encoded with Lame 3.97's and 3.96's VBR modes.
It is a bit less distorted when encoded with 3.90.3 VBR modes.
It is also distorted with 3.97b1/b2 cbr 320.

There is some listening experience (by StephanV, Wombat, myself [hope I didn't forget somebody]) in:
Using 'insane' settings (cbr 320)
The mother of all problematic samples

This post has been edited by halb27: Dec 2 2005, 13:10

[Alex B]

A problem sample (herding_calls):
http://www.hydrogenaudio.org/forums/index....showtopic=37002

The previous discussion and test reports are in this thread:
http://www.hydrogenaudio.org/forums/index....showtopic=37003

[Wombat]

Similar problem like Birds but should be even easier to hear when using 3.97 -V2 --vbr-new 3.97 -V2 is not much better. 3.96 does best again.
The whole sample suffers by added noise. This happens on several places on this album but to a less degree.

Deploration

This post has been edited by Wombat: Dec 7 2005, 01:58

[kkumul]

How can I hear the noise?

http://my.netian.com/~lsj17/Birds.mp3

I used 3.97 b2 with -V2 --vbr-new options.
but,
I couldn't find any noise...

Please try my link to hear...
Can you guys find any noise from the link?

or

Am I just one person who can not find the noise?

This post has been edited by kkumul: Dec 7 2005, 04:53

[ajcool123]

I couldn't hear anything either. Sounded fine to me.

[Wombat]

Well, if you don´t hear it, it is fine. Some people are more sensitive to different artifacts.
I don´t need any effort to hear this while i have problems hearing pretty obvious pre-echo.
Any further discussion, like the 2 posts above belong here:
Lame problem samples discussion

[level]

I found in the upload section a sample that produces a strange and very annoying artifact. This is something as phasing with modulation and dropout.

I tested in ABX with Lame 3.97b2 -V2 and Lame 3.97b2 -V2 --vbr-new; both seem to be in the same situation.

This is the sample:

aps_killer_sample

I will test with other versions of Lame. In this moment I am very busy.

This post has been edited by level: Dec 8 2005, 16:56

[Shade[ST]]

I found in the upload section a sample that produces a strange and very annoying artifact. This is something as phasing with modulation and dropout.

I tested in ABX with Lame 3.97b2 -V2 and Lame 3.97b2 -V2 --vbr-new; both seem to be in the same situation.

This is the sample:

aps_killer_sample

I will test with other versions of Lame. In this moment I am very busy.

Interesting sample. I'm actually able to abx it fairly easily at -V 5, which I've not been able to do with other samples. Cool!

[Wombat]

This piece of song is about 2 minutes long and adds noise often. As meanwhile common 3.96 -V2 --vbr-new is pretty clean! Here a small clip:

Moon


For the people above that can´t hear a problem with Birds, please try this one or Deploration and give some feedback please in the Lame problem samples discussion thread.

Edit: the problem occurs with 3.97b2 -V2 --vbr-new, of course

This post has been edited by Wombat: Dec 9 2005, 09:45

[kkumul]

http://my.netian.com/~lsj17/Deploration.mp3
http://my.netian.com/~lsj17/Moon_short.mp3

Can I get some result files?
I mean mp3 files with the noise.

Please just try my links.
I couldn't find any noise yet.

This post has been edited by kkumul: Dec 9 2005, 04:59

[Wombat]

http://my.netian.com/~lsj17/Deploration.mp3
http://my.netian.com/~lsj17/Moon_short.mp3

Can I get some result files? 
I mean mp3 files with the noise.

Please just try my links.
I couldn't find any noise yet.

If these samples are 397b -V2 --vbr-new and you can´t hear something in Moon_short i doubt you ever will. Moon_short was only downloaded 5 times till now. Seems like mp3 isn´t much interesting these days here.

[kkumul]

So..
Do you hear the noise from my links?
If you do, I guess my ears aren't good.
Right?

I have a question.
Is there sample original?

I mean,
Am I supposed to find the noise from your file?
or
Do I have to encode your file to hear the noise?

I encoded those files from what you uploaded, then I compared that with your file.
Is something wrong?

This post has been edited by kkumul: Dec 12 2005, 14:55

[Gecko]

Sample: angelic
A filtered sawtooth with heavy reverb and phasing.

Problem:
It sounds like someone is crumpling the styrofoam of a microphone.

Settings:
lame -qX angelic.wav

Versions:
3.90.3: Does not have this problem with either q0 or q2 (no presets used)
3.96.1: Shows a certain amount of fluttering which increases with each q level (tested 0 through 3)
3.97b2: q0 shows extreme amounts of the crumpling noise. I think everyone will be able to hear this. q1 through q3 have the same problem but to a lesser extent.

In two of the 3.97 alpha testing threads you will find some abx tests at higher settings.
http://www.hydrogenaudio.org/forums/index....ndpost&p=276433
http://www.hydrogenaudio.org/forums/index....ndpost&p=279873

edit: Easily ABXable with 3.97b2 -V2 --vbr-new: 15/16 (was clicking too fast)
edit2: removed a "don't" where it shouldn't have been.

This post has been edited by Gecko: Apr 16 2006, 14:02

[drezon]

Sample: thunder

The comment in that post is for iTunes AAC, so never mind.

After severe iTunes problems I tested LAME and was able to ABX 397b2 -V2 --vbr-new with 10/10. -V0 --vbr-new was transparent to me with 6/10.

I'll try to describe what I'm hearing: I focused on the synth sound starting shortly after 20s. Especially that second synth which is positioned to the right sounded smoothed or smeared to me.

[Alex B]

I forgot to add this link here:

ftp://ftp.tnt.uni-hannover.de/pub/MPEG/au...am/harp40_1.wav

harp40_1 is an old problem sample. I can easily ABX it. (LAME 3.97b2 @ -b 160 and -V3 --vbr-new). More in the discussion thread.

[jmartis]

Hi, btw my 1st post here
I know synthetic samples like this are not going to occur in normal music, but anyways here it goes.
It's a modified sawtooth wave with harmonics and stereo sorround. It produces a NASTY hissing @128kbps cbr w/ Lame 3.97b2. The hiss is about the same loudness as the sample itself It goes away a bit with 3.98a3, but is still here. Also with -V 2 --vbr-new the bitrate inflates to 270kbps (not normal)
surround.flac

[jmartis]

Here's another sample, this time from real music
It is great for pre-echo testing, 3.97b2 --preset standard has problems with it. Also in CBR mode there is some bubbling artifact accompanying the bass.

scooter.flac

edit: WinAbx (this is with 3.97b2 --preset standard)

CODE

13:29:15    1/1  p=50.0%
13:29:52    2/2  p=25.0%
13:30:21    3/3  p=12.5%
13:30:29    4/4  p=6.2%
13:30:36    5/5  p=3.1%
13:30:54    6/6  p=1.6%
13:31:02    7/7  p=0.8%

edit again- the bubbling or gurgling is gone when *not* using high quality ( -h) option

This post has been edited by jmartis: Apr 16 2006, 14:03

[Gabriel]

Artificial sample triggering problems:
http://www.hydrogenaudio.org/forums/index....showtopic=44228

[jmartis]

looped part from "scooter" sample. high-frequency dropouts in latest Lames (cbr 128k and 160k). Produces artifacts that sounds like short "water splash" or "blip"

scooterlooped.flac

(see this thread: http://www.hydrogenaudio.org/forums/index....showtopic=45049)

J.M.

edit: added blip

This post has been edited by jmartis: May 31 2006, 16:45

[guruboolez]

Maybe a similar issue (frequency dropout), ABX = 11/12.

The sample is No. E50 from my catalog.

additional information:
http://www.hydrogenaudio.org/forums/index....ndpost&p=396742

This post has been edited by guruboolez: May 28 2006, 10:42

[jmartis]

Another sample: scooter2.flac

Problem: Ringing with bitrates up to 320k. (--preset insane)

Late edit: happens with -q 0, -q 1 and q 2

Spectrograms, that really don't say anything: original sample, Lame --preset insane sample.

WinABX (with Lame3.98a4 --preset insane -h) :

CODE

WinABX v0.42 test report
05/31/2006 16:30:19

A file: J:\scooter2.wav
B file: J:\scooter2lame.wav

Start position 00:00.0, end position 00:11.3
16:30:44    1/1  p=50.0%
16:30:50    2/2  p=25.0%
16:30:56    3/3  p=12.5%
16:31:02    4/4  p=6.2%
16:31:07    5/5  p=3.1%
16:31:13    6/6  p=1.6%
16:31:18    7/7  p=0.8%
16:31:25    8/8  p=0.4%
16:31:33    9/9  p=0.2%
16:31:42  10/10  p< 0.1%

though I can hear it only in my left ear which is better at high frequencies.

J.M.

This post has been edited by jmartis: Sep 4 2006, 11:05

[halb27]

Another sample: scooter2.flac

Problem: Ringing with bitrates up to 320k. (--preset insane)

Do you mind trying 3.90.3 --alt-preset 270 (or alt-preset insane)?

Edited: Sorry, should go to the discussion. Can one of the moderators move it there pleease?

This post has been edited by halb27: May 31 2006, 22:38