Dear Gabriel, Halb, and all intelligent speakers from here,

I'm not sure why you decided you have only ONE sample!

There are lots of samples!

I heard this Trumpet sample and it sounds awfully!

This means that all trumpet-like sounds are potentially screwed in the latest version of Lame.

This means any music with rich frequency timbres is potentially crewed.

This means any electronic music, like trance or techno, with lots of trumpet-like sounds is potentially screwed.

Isn't it?

I think it should be easy to find any number of samples with such screwed distorted sounding.

I'd welcome your comments on this. Thank you.

Find them.

Then you should have no trouble posting lots of samples + ABX test results!



See the rules. If you're going to make claims, you should post ABX test scores if you want other people to look at something. You should also post that sample. Saying "a sample exists" isn't helpful. You need to actually prove that it exists and then provide it to everyone else so it can be studied.

In this case you would at very least first need to be able to ABX against the wav file using 3.97b -API. Then you should do some sort of blind comparison (like an ABCHR test) to show that the 3.90.3 version does sound better. Or at least you should be able to point out how 3.90.3 is better if its so striking that a test isn't required.

I know the rules, please don't try to teach me.

Everybody, please see the point. Trumpet sample was easily ABX'ed by quite a number of people in this thread. If you can extrapolate and take into account the nature of the trumphet signals (a lot of harmonics in the high frequency band), you can be pretty sure a lot of signals and samples similar to Trumpet sample potentially can be found.

Have a little of imagination and common sence. Don't try to blindly ask for samples and proof - we're not in the court. Just think a little.

Addendum: I checked the Lame 3.96 MP3 320 encoded Trumpet sample with EncSpot - it indicates that the file has a lot of short blocks - about 45%! I have never seen such a ratio - a typical ratio in my 100 GB MP3 256/320 collection is 3-10% for short blocks and 90-97 for long blocks.

I made available 150 samples of classical music (and 60 of solo instruments).
Try to find another artefact like this. Shouldn't be to hard, isn't it?

http://guruboolez.free.fr/
ftp://ftp2.foobar2000.net/foobar/

Diagnosed here.

Sorry, I don't agree here. The artifact is noticeable at -b 320, but not something I'd pay attention to if I didn't have the original to compare. Could you please explain what is awful about it?


Then do so, instead of letting others do the work for you. No developer can improve a codec on imaginary samples. Just think a little.

If it's that easy to find more samples, why do you think people here haven't found any more? Please go ahead and find more for us. It would be most welcome.

It should be clear from this discussion that you can't draw such conclusions you did from only one sample.

> Sorry, I don't agree here. The artifact is noticeable at -b 320, but
> not something I'd pay attention to if I didn't have the original to
> compare. Could you please explain what is awful about it?

Perhaps someone should have trained skills to hear and feel the music.

The MP3 sample is dead for me, it's void of life, devoid of
any meaning.

While the original sample sounds like real live music, the MP3 sample
is simply annoying signal.

The original sample is full of life, it feels like the real trumpet
behind me, it resonates deeply with me, gives me a kind of emotion and
mood and leaves me in that mood for some time after the music is over.

The MP3 sample is dead - it just does not resonate with me. It sounds
like a distorted, unnatural musical instrument to me. It sounds like nothing.

Maybe you guys don't have enough trained ears or good enough equipment
to hear what MP3 makes to a properly mastered CD music.

For those who can read Russian language, here is a blind test by a
third-party person:

http://klem.ru/forum/index.php?act=sht&s=&postid=33916

He can ABX almost any MP3 file in any bitrate (up to 320) on Chinese $65
speakers with chip $10 sound card.

I tell you MP3 320 music is dead on any real Hi-Fi (>$5K) or Hi-End (>$15K)
audio system. This is the main problem with MP3 music.

It is dead on $75 systems for some trained ears as well, who can
really feel and understand the music.

> Then do so, instead of letting others do the work for you. No
> developer can improve a codec on imaginary samples. Just think a
> little.

I think developers should be concerned anyway.

> It should be clear from this discussion that you can't draw such conclusions you > did from only one sample.

For me all this MP3 junk was just a waste of time. Since 1999 I've collected a huge 100 GB MP3 collection. 80% of the files are Lame MP3 320, and am now going to throw it all, with my new audio upgrade.

I'm also going to listen more to vynil sound, because it's deeper, warmer, and has much greater dynamics.

Deeper and warmer - because it keeps ultra-sonic frequences up to 40-60 KHz.
Properly mastered DVD-A's and SACD's with proper DAC's is also OK for warm sound.