I've been transcoding some of my flacs to aac in anticipation of getting a mac soon and as I listened to them I started to feel that they sounded muffled. I decided to sit down and try and do some real abx testing for once. To my great surprise I was able to abx quite a few metal tracks on iTunes 192k encodes. Mostly a clarity in the high end thing. I realize that metal is fairly difficult to encode but I thought I'd be safe at 192.

Both ogg at -q 6 and lame -aps go above 210kbits for this material which also suggests to me that making this stuff transparent at 192kbits isn't easy.

Which version of iTunes were you using?

4.6.0.15 on windows.

you are not the first one!

Look at this

http://www.hydrogenaudio.org/forums/index....showtopic=24462

So maybe kuniklo is experiencing the same decoder issue that Cygnus X1 mentioned in that thread...

Hmm.. you don't happen to have the sound enhancer turned on by any chance? Try to play the file directly in QT and see if you still get the artifact.

Hmm. I decoded the m4a file to wav with foobar2000 before abx testing. The decoder in foobar shouldn't exhibit this itunes bug, right?

Yup, foobar2000 uses faad2, no such bug.




IIRC, member "guruboolez" can even distinguish compressed vs original for normal music at bitrates around 192 kbps.

congratulations.......
you have good hearing

I don't think I'd be able to spot the things I'm hearing in less difficult music. Maybe I'll give it a try.

I guess the good news is that all those punk rock shows in the 80s didn't completely destroy my hearing.

The bad news is that I'm back to tired old mp3 for now. Or maybe I'll try to figure out a way to get mpc to play in VLC on the mac.

You normaly abx a particular trouble passage for a second or so but not much more - esp at 192k. Do these artifacts persist throughout the song ?

Also how many trials did you set?, are you getting instant good scores like 7/7 or 16/16 ?

Try abx with replaygain on the files and also without replaygain - maybe even with the winabx program.

I don't doubt you but from my experience mistakes with abxing can happen that are setup related.

I chose passages that I knew would be difficult to encode but are also fairly typical of a lot of the music I listen to. For some of them I got pretty solid 15/16 results. Others were more in the 12/16 range but that's also partially due to fatigue. If I really focused and listened back and forth several times I did better.

I focused on high-frequency transients like cymbals, vocals & guitar. They just sould a little flat. Nothing horrible, but since it's exactly the clarity of those kinds of sounds I like it's bothersome enough to keep from using aac at that bitrate.

What are the listening material? earbuds? headphone? loud-speakers?
I could ABX some tracks at 192 kbps, but only with a good headphone...

The material is recent metal recordings - Soilent Green and Borknagar specifically.

I'm listening on decent headphones, Sennheiser HD 580s and a better than average interface, a M-Audio Firewire Audiophile 1814. I wouldn't be surprised if I couldn't hear what I'm hearing on cheaper gear but this is how I generally listen to music.

Perhaps the 16 kHz lowpass is causing this flatness? I've never checked the lowpass at bitrates higher than 128 kbps from iTunes before so I'm not sure.

Verified with latest iTunes/QT [6.5.01]:
192 kbps = lowpass at 19,5 Khz with a slight desertic frequency band at 17...17,5 Khz.

It's maybe something else. I remember that long times ago I was able to ABX QT AAC at 320 kbps (with a very high lowpass value) on harpsichord: sound was slightly dull, synthetic. It was probably something else than lowpass.

That sounds like the same effect I'm hearing. It's subtle, but noticeable.

I've also been able to abx vorbis 1.1 -q6 on some, but not all of the same material. Surprisingly, I haven't been able to abx lame -aps on any of this stuff yet but I haven't spent as much time on it.

Surprisingly? Why "surprisingly"?

becoz both ogg vorbis n aac should be more efficient than mp3?

LAME is better, so its a surprise
EDIT: official Vorbis 1.1?
ummh, you should have tried Megamix2 or GT3 instead of the official one!
both megamix2 or gt3 should offer much better quality.

Well, suprising to me anyway. I expected that new newer codecs would do better than lame.

Try with samples like castenets: modern AAC encoders are sharper than any previous, current of future mp3 encoding at the same or even higher bitrate.
Lame is good, and could compete in several case with modern audio format. But not all...