Or, to be more specific: Can I call a certain encoding non-transparent if I can only hear the difference when I adjust the frequencies to my liking on the EQ?

The background to my question is this: I have a song (by Conjure One - "Sleep") which is dead easy, even at -b 320 (using Lame 3.97b), to tell from the original wave when I play them in winamp where I have the bass and treble turned up. (The lowest frequency at +12 and the three highest at about +6 - it's the higher freqs that give the mp3 away.) However, when I decompress the files and try to ABX them in WinABX (where there is no EQ, obviously) it is very difficult (impossible?) to tell the difference.

I would like to argue that something that doesn't sound like the original in my listening environment, and with my frequency tweaks, isn't transparent. Is that reasoning reasonable, so to speak?

Winamp uses different EQ's for .wav and .mp3.

And to answer the main question: No.

Well, almost no. To be on the safe side, dont use any EQing at all. However, a bit of EQing... if its done in a natural subtle scale.... should be ok. The point which i'm trying to make is: lossy encoders are meant to throw away stuff which you cannot hear/allow artifacts in cases where they cannot be heard. But if you artificially amplify certain frequencies to levels which simply would be ridiculous for normal listening, then its okay if the encoder messes up...... because its job is to make the file sound okay for normal listening, not for cases where you intentionally screw the signal.

Setting bass at +12 dB and treble at +6 dB will introduce a hell of a clipping. This clipping can be different with the original and with the mp3.

Ok. I didn't know. What about ogg?



True. I forgot to mention that I've set the overall amp to -12.



But, if the difference will be clearly audible while I actually listen to the music... Hmm.

Ok. But if winamp uses different EQs for mp3 and wave, that might explain the problem. I thought the difference was too clear to be true :-) considering that people usually talk about mp3 transparency at ~200 kbps.

Yes, I see your point. But the settings I have are not there for the sole purpose of differentiating between mp3s and waves. I chose them to fit my listening, and I don't think they are rediculously turned up. (Apart from, perhaps, setting the bass at +12 :-) but since that wasn't really what made the difference I can't see that it matters.)

Ah, okay. Whatever, the problem simply is that winamp has got a reputation for being unreliable in terms of ABXing..... especially its EQ. My proposal would be to do a test-installation of foobar2000, just for checking if the problem disappears then(while using EQ) - simply to make sure that its not the audioplayer or its EQ which is to blame. If yes, then blame winamp's EQ and either live with it, or look for another audioplayer.

Pardon my ignorance, but is there even an ABX plugin for Winamp? It could be kinda difficult without it...

If i understood him right, he decoded the output to wav, and then abxed it...... so if winamp does messup anywhere in the signal-pipeline, then it would also be in the WAVs.

You are right, I should read properly.

Well, the Winamp EQ is known to suck. What happens when you listen to the wav decoded from the mp3?

That's a good question. I was just thinking about that.

I've got foobar allready, but can't seem to find the EQ. I'll have to hunt around a little.

Oh - and I hadn't actually ABX-ed anything - the difference I heard was very clear. (And now I'm not being one of those trolls who say's "I can hear the difference between lossless and the original, blah blah blah" :-) The difference in this case truly is clear.

Here they are so you can listen for yourselves:
Wave (13 MB)
mp3 (3 MB)

You only need to listen to the very first snare hit (first few secs of the track) in winamp, with the treble upped a bit, to hear the difference.

As already mentioned Winamp uses different equalizers for MP3 and WAV, which explains the differences you heard.

Yeah. I'll confirm that tonight after decompressing the 320 kbps mp3 using EAC, and comparing the original wave to the trancoded one using winamp.

Off-topic: this somewhat reminds me of something I read a while back about a Sony player which would EQ down MP3 to make it sound inferior; so that people would like ATRAC (sorry, A-who now? ). Now that's just nasty!

Correct, you're talking about the latest generation of Minidisc players. Unbelievable action from Sony.

That just reinforces my feeling that you should never use Sony gear for anything digital audio-related.

Ok. I've now transcoded the 320 kbps mp3 back to wave (using winamp disc-writer plugin, with EQ off!) and the resulting wave is, just like you guys predicted, impossible to tell from the original (played in winamp).

That means that winamp is constantly degrading peoples mp3 music, if the EQ is turned on. The mp3s are played no were near as crisp as the corresponding (decompressed) wave-file. Comparing the two, the 320kbps mp3 sounds like it has been encoded by some n00b at 64 or 96 kbps...

I'd say this is yet another reason to use Ogg, since (it seems) winamp uses the same EQ for oggs as for waves. At least for most avergely-skilled people like myself who get scared by the horrendous (or wonderful? :-) ) amount of prefs in Foobar...



ATRAC never was even remotely popular... At least that's my impression.

So Winamp, by default, (badly) EQs MP3 files? But why?

Because doing the equalization directly in the MP3 decoding is faster.

Doesn't the Winamp Musepack plugin do something similar (but better)? I'd check, but I have'nt used Winamp in ages.

In this case, mp3 vs.wav I agree with you that technical issues can arise, but what about mp3 vs. mp3 and some room EQ which you use to correct the interaction between your speakers and your room? I use very low levels of correction ( max 2 dB in some bands) and really like the sound this way...

I've now investigated what it takes to hear the difference between waves and mp3s in winamp, and it seems that just switching the EQ on (with all bands at 0) is not enough. You need to pull the top two bands up a few notches to hear the difference.

All the same, very annoying.

Is the audio clipping?

No. What happens is that if you turn the high freqs up a few notches it doesn't have the same effect on mp3s as on waves. The mp3s sound much more muffled in comparison. The effect is similar to the mushy sound you get for cymbals and crashes if you over-compress an mp3.

Several clues to this thread:

A) Winamp General EQ (which is not the same of the old Winamp's, less than 2.90 or around that time), acts over the decompressed signal, while MP3 EQ works directly on the MP3 data before decoding, ( as Garf said).

A+) The difference in sound is mostly because of the frequency bands to which the sliders are mapped. In MP3 mode, they are not linear ( 4 top ones being 6, 12,14,16), while in General mode, they are ( being 5, 10,16,20 or something like that, can't remember now)

B) MP3 EQ can be disabled ( so make it use the General EQ ) by going to options-Preferences, Pluins - input, select Nullsoft MPEG decoder, Configure, and in the decoder tab, unselect "Fast Layer 3 EQ".

C) Like being said here before also, Musepack decoder has an option to use the General EQ or an internal one. I don't have information of how this internal one works (if it modifies the data before decoding, or after).

D) Winamp EQ, at the "Flat" position, is auto-disabled, so you have to move any slider up or down by any amount even if it is in the "On" position to listen to it. ( I developed a plugin once, when the General EQ was bad).

There is no universal transparency (that I know of) with lossy encoders.

Therefor ALL claims, if you want to be accurate, must be followed by conditions under which you make them.

For example:

XXYZ is not transparent for me on my setup using my normal EQ setup.

Of course, it is easier to hear some artifacts with really unnatural EQ setup, that nobody in their sane mind would ever use.

This kind of EQ manipulation is easily seen as purposeful in order to cheat in tests.

However, if you use your normal EQ settings, then you can fairly say that X is not transparent within that application (eq settings) for you.

To be more scientific: it all depends on your test hypothesis.

What are you trying to prove (or disprove)?

OK. That's interesting, and kind of what I was thinking in the back of my mind that it must be.



Great! Thanks. I didn't even think of looking for that option.



Ok. I kind of suspected that too.

You are completely correct. I observed the same phenomenom for higher frequency music, such as cymbal and hi-hat hits. The WAV sounds much brighter than the MP3 w/ EQ on.