Hi,

This is my first post, though I've been reading the boards for a few weeks. I've learned a lot from many of you, and I'm very thankful for that.

Here's a phenomenon that I found only yesterday that probably isn't new at all to many of you, and hopefully someone could explain it to me. I've tried to search the archives to no gain, and the fact that "EQ" is too short for a valid keyword in search didn't help.

I tried for the first time to encode MPC and OGG files. (I'm not very interested in AAC due to the heavy patent concerns.) I ripped around 10 tracks of various classical music (that's what I listen the most) -- symphony, opera, piano ... etc. I then encode each of them into several MP3, MPC, and OGG files.

For MP3 I used LAME 3.91 with APS, APX, API (insane). For MPC I used mppenc 0.90s with standard, xtreme, and insane. For Ogg I used Oggenc RC3 with -q 5 to -q 10. All on WinXP.

I listened to them in Winamp 2.78 (with a headphone), and I don't need an ABX to concede they're all transparent to me. Let's say they all sound like 'A'. It's hardly surprising considering that in general classical music isn't hard to encode, and my audio equipment is way below average. And I'm not trained to hear compression artifacts.

Now the interesting part. I turned on the EQ (I'd checked to use musepack plugin 0.90k's EQ, not Winamp's native EQ), and made some drastic adjustment -- I adjusted 12k, 14k, and 16k (the three on the right) to +20db, while leaving others alone.

The orginal .wav now sounds like B, which is like, well, A with heavily emphasied high frequency. Those MP3s (regardless of presets used) sound like C. C is also like A with heavily emphasied high frequency, but is obviously different from B. Then the MPCs. They all sound like D, which is again different from B and C. IOW, I could now easily tell each group apart. All MP3s sound alike, and all MPCs sound alike, and they're all different. The difference is very obvious. (I'll do ABX if necessary, but I think it's so obvious, and I guess it's probably a known phenomenon to most of you.)

The exception: Ogg. All OGGs sound alike (from q5 to q10), but they sound like B (the original wave under the same EQ setting)!! I'll have to do ABX and listen more carefully to see if I really can't tell them from the orginal, but I'm almost sure.

Thereafter I tried other EQ settings (emphasizing base, e.g., or 1k band), and the phenomenon remained the same. And it applies to all the tracks (about 10) I tried.

From past reading I've learned that in terms of sound quality at high bitrates, MPC has no peer. Ogg is second (not counting AAC), and then MP3, when comparing files of roughly the same bitrates. And my impression is that basically the conclusion is drawn based on how well they encode "hard" samples -- samples identified to be hard to encode.

I've nothing to counter that. But from my experiment, it seems Ogg is best in staying faithful to the orginal, even when EQ manipulation is applied during playback.

Am I right about that?

Welcome aboard!

Winamp's own EQ messes up the whole frequency range pretty bad. Don't use it if you don't have to! That's probably why mp3s sound different. Check this thread: http://www.hydrogenaudio.org/forums/showth...s=&threadid=913

In theorie mpc --insane should be very EQable since it preserves the whole spectrum (note: this is "insane" from a psychoaccoustic point of view). I don't know about ogg. Maybe it has it's own built in EQ which doeas better than the Winamp generic one?
Dude! You got me seriously confused! :confused:

I think what he's saying is that while he can't tell apart the various codecs normally, when he pumps up the high frequencies in their respective EQs, he can now tell them apart fairly easily, both from each other and from the original WAV, with the exception that Ogg sounds like the original wav.

Possible explanations for this:
1) Ogg transparently encodes high frequencies (at least on this type of music), while MPC and MP3 do not.
2) the Ogg EQ boosts frequencies in the same way as the WAV EQ, while the MP3 and MPC EQs behave differently.

Of course without more information it's impossible to conclude what this means or what should be done about it. It could be that the MP3 and MPC equalizers are broken. Or it could be that the WAV and Ogg equalizers are broken, but happen to be broken in the same way. Or some other possibility.

This has no informative value about better or worse preservation of the original sound in different codecs. Listening tests showed that indeed MPC stays most faithful to the original, especially so with --xtreme or --insane... it's very unlikely that Ogg at q5 (!) and up will sound better on difficult samples. As for your test with "easy" classical music: On easy-to-encode samples, there shouldn't be any audible difference at all among those codecs, not even with LAME. They're all good enough to mostly give flawless results with easy samples. This all suggests that it's an EQ phenomenon.

Hi,

Thanks for all your replies.



I knew about this problem and that's why I said I was using musepack plugin's EQ, not Winamp's native EQ.



I don't think OGG plugin has its own EQ. At least I didn't see one.



Sorry if I didn't make myself clear enough (pardon my poor English, which isn't my first language). Let me clarify, then.

For each test sample, I had ...

1 sample.wav
3 MP3 files, encoded with --alt-preset standard, extreme, and insane respectively.
3 MPC files, encoded with --standard, xtreme, and insane respectively.
6 OGG files, encoded with -q 5 to -q 10 respectively.

That's a total of 13 files per test sample.

Now, when EQ was off, all 13 of them sounded the same -- let's call it A.

With EQ on, the WAV and 6 OGGs sounded like B (the EQ was adjusted so that the difference between B and A was significant). The 3 MP3s sounded like C, and the 3 MPCs sounded like D.

The fact that I couldn't tell "standard" from "insane" (or q5 from q10) was expected because, as I said, my hardware was poor, the original WAV was not hard to encode, and I wasn't well trained to hear artifacts.

What surprised me was that, with EQ on, I could easily tell MP3 from MPC, and tell both of them from the orginal WAV. But I couldn't tell OGG from WAV.



Mostly right, except that I tried various different EQ settings, not just pumping up the high frequencies. (I also tried to pump up low frequencies, or middle frequencies -- respectively, not all at once.)



Since the phenomenon was universal with all the EQ settings I tried, this shouldn't be the case.



I thought Ogg Vorbis plugin doesn't have its own EQ and I was using musepack's EQ, but since both you and Gecko mentioned this, I'm not sure anymore. I checked just now, and if there's an OGG EQ, I couldn't find it.



I can provide more information if I know what I should provide. I could provide samples as well if necessary, but since it happened to all the 10 or so test samples from different CDs, I thought it's obvious and anyone could duplicate it with their own CDs. But judging from the 3 replies so far, it seems none of you have encountered this before. This makes me wonder if I did something wrong. Any ideas?

Is it true that on your system, you couldn't tell a MPC (insane, e..g.) from the orginal WAV, even with the EQ on and, say, the bass band emphasized by 20db?



I never argued against that, did I? I wasn't (am not) trying to suggest Ogg at q5 would be better than MPC --insane (or Ogg at q10 for that matter). Especially on difficult samples. As I said, I've read about that and I've no reason to challenge that conclusion. (My audio equipment won't allow me to anyway.)

I was (am) merely saying that, how surprised I was to find that when EQ was on and adjusted heavily, it's so easy to tell MPC (even at --insane) apart from the orginal WAV, even on my poor equipment, with "easy clasical music", and to untrained ears, all that. Same with MP3, but not with OGG.



I'm not sure I understand you here. If you mean this happens only when EQ is on (and not when EQ is off), then yes, that's exactly what I found.

If you mean this is related to a buggy EQ, and shouldn't happen with a good EQ, then could you point me to a better EQ (for Winamp) so I could test it out? (Too bad my handheld CD radio doesn't have an EQ, otherwise I would burn them to CD and try it there.) And since I was using Musepack's EQ, it's interesting if it's indeed its fault that MPC was easier to pick out than Ogg.

If you mean this is not an issue since it happens only when EQ is on, then could you explain why? IMHO, we want an encoded file to sound as close to the orginal as possible in any circumstances, including when EQ is used. Don't we?

tw101: try testing all codecs with same EQ, otherwise the results are pretty pointless. You could try Naoki Shibata's superEQ. I think it's available here:
http://shibatch.sourceforge.net/

Agreed. But adjusting single frequency regions by +20 dB is umm... daring, to say the least. That's simply not the field that lossy codecs are optimized for. If you want a codec to sound best with partial +20 dB EQ boost, well, that's not too hard of a task. But it's an unrealistic test.

Remember, the dB (Decibel) scale is logarithmic. 20 dB more means that the acoustic pressure is 10 times increased. No user in the world will set his EQ like that.




You should try this one: http://prdownloads.sourceforge.net/shibatc...perequ-0.01.zip

Mind you, it doesn't boost any frequencies. You have to attenuate the opposite frequencies of those you want to amplify.

It's not a given that this EQ will provide more consistent results. It could show the same behavior. But, as i said before, i don't think this has a valuable meaning.

This is a quick reply to JohnV and CiTay before I leave:

Thank you so much for pointing me to a good EQ. I haven't got enough time to test it thoroughly yet, and it took me a while to figure out how to use it. But a quick listen showed that they all sounded the same again. (I.e., using my original analogy, they all sounded like B when EQ is on.)

Mystery solved. Thanks a bunch. You're really knowledgeable and wonderful people.