Lame and noise?

2003-06-09 16:13:15

I downloaded a song from internet and in the end of the song i can hear some weard noise. It's some kinda rustling. I have no idea what program there is used for riping but EncSpot show for encoder Lame 3.88(beta) and 192kbps with full stereo. Does that encoder produce rustling or is it because unsecure ripping method? Song is Metal and i wonder that could it be some low bitrate artifact? I'm not sure, because i hawn't done any larger ABX testing with MP3 and encoding artifacts are not so clear to me.

Lame and noise?

Reply #1 – 2003-06-09 17:36:52

Too many unknowns to draw any valid conclusion.

Lame and noise?

Reply #2 – 2003-06-09 18:13:33

I heard some strange noise too and not only with mp3's.
And the problem I have is an onboard soundcard.
I build another computer and used a motherboard with an onboard soundcard.

Lame and noise?

Reply #3 – 2003-06-10 11:24:11

Askoff, I don't think anyone can identify the problem without more information, such as:

• Encspot's info about the file. Is it CBR 192kbps, what's the lowpass, etc. Using full stereo on lame and using 192 kbps CBR when --alt-preset standard is available sounds like it may have some dodgy commandline options being used, possibly to make the spectrogram or other graphs looks good at the expense of perceived sound quality. (Joint stereo is safe on Lame, but the dogma of using dual channel stereo developed from flaws in other encoders like Fraunhofer, or problems with intensity stereo coders)

• A clip (less than 30 sec) would also be useful, if you can post it on a website and provide a link. (You'd need to use mp3DirectCut, musiCutter or similar to take an excerpt of the mp3 without introducing a decode/encode cycle). A tool like mp3x might reveal a lot.

If you want to learn the basic artifacts and the language used to describe them there are some very obvious samples of most of them on ff123's Artifact Training page.

Lame and noise?

Reply #4 – 2003-06-10 11:50:54

Ok... So used software is foobar2000 with slow 64bit resamplin to 48kHz plus Advanced Limitter. Hardware is SB Live! useing S/PDIF out.

Yes it's CBR 192. I did't find any lowpass info with EncSpot, but with SoundForge Spectrum analysis i gues that is somewhere near 16kHz.
I'll cut the sample bit later. Now i have no time.

Thank's for trying to help me...

Lame and noise?

Reply #5 – 2003-06-10 17:28:02

So now i have the sample is ready to download.
Sample.mp3

Lame and noise?

Reply #6 – 2003-06-10 18:13:25

this is pointless imo. impossible to conclude anything without the source wav AND lame 3.88 is quite old. it can be a transcode too....

Lame and noise?

Reply #7 – 2003-06-10 19:09:37

There is no need to dig things more than you can. I was only wondering the problem and i don't require answers. If someone just know where source of the problem is.

Lame and noise?

Reply #8 – 2003-06-10 19:45:20

what do you mean? the "hiss" like sound at 9 seconds or so? it's not an encoding artifact.. it's most likely there from the source. there is also a hissing/distortion-like sound in the background throughout the sample you provided. to my ears, this is not an encoding artifact. sounds like possible internal clipping perhaps.. not sure

Lame and noise?

Reply #9 – 2003-06-11 13:35:44

Yup, there's some distortion-type sound (harshness) in the female voice at around 7-8 sec, which is probably deliberate, then the hiss at the end sound quite natural and not especially like an MP3 artifact. It's also about equal left and right channels, which in full stereo mode is less likely to point to an artifact. I'd say it's probably in the original sound.

The lowpass, if any, would've been about 19.0 to 19.5 kHz judging by the end before the hiss and a few other parts of the sound. (Ignore possible clipping at the start). This accords with the 19.5 kHz used in recent lame encodes at --preset cbr 192, though the lack of joint stereo is likely to risk some artifacts occasionally (but probably not all that loud). When I re-encoded (transcoded) the decoded sample using --preset cbr 192 in lame 3.90.3modified, 93.47% of samples encoded in MS mode because it's more efficient in safe joint stereo, so forcing LR mode for all 100% must lead to a considerable amount of wasted bits, leaving fewer bits to avoid artifacts. (Of course, any subtle artifacts that are then cannot be cured by reencoding the already damaged mp3, I just did this to measure the stereo nature of the track)

Going back to your original, the resynthesised waveform at around frame 380-382 in mp3x doesn't just look like uncorrelated noise either, it has more structure, tending to suggest it's in the original signal, then it ends in the last frame. The near silence following the end of the signal in the last frame is probably the mp3 padding with zeroes to the next frame boundary (not being a gapless format).

I'd have thought the hiss might be the start of the next track (bad cut point?) or an intended part of the sound, but probably not an artifact of the MP3 encoding.

If I had to guess, and knowing they had lame 3.88, I guess they used something very close to:

lame --alt-preset cbr 192 -m s

The -m s isn't safe, and will be worse quality (more artifact risk) than leaving cbr 192, but it's still a darn sight better than 128 kbps CBR. If you don't like the sound of the hiss at the end, you can remove it with the same tool you used to cut off the 9.9 second sample.mp3 or you can fade it or reduce its volume by 40-80 dB so it's essentially silent. (Just make a backup copy before committing the changes!)

Lame and noise?

Reply #10 – 2003-06-11 16:35:19

Thank you. Maybay i should buy original CD and try to rip it my self.

once i had same problem when i ripped and encoded with Easy CDDA Extractor 192 Joint stereo and burst mode using NEC DVD drive. I don't remember exactly when i did that, but it was years a go. (about 3 year)
Music style was pretty close to that in the sample i posted and i found it in several songs. Thats why i thought it could be some DAE qualtiy problem in CD drive.