Well, here's the thing: If the player doesn't support ReplayGain (most don't if they can't be Rockboxed), then you'll need to transcode from a lossless source to get your tunes at the same volume level. At that point, you may as well go for a high bitrate Ogg or AAC to save a little space. Chances are you won't be able to hear the difference 99% of the time. Ogg sounds pretty good at -q8 and beyond, and I bet that with any portable listening setup you'd be hard pressed to catch any differences between that and a FLAC file in a genuine ABX test. For me, the differences between -q7 and -q8 are subtle at best, and only when I'm really paying attention. Beyond that, I doubt I'd be able to tell much difference without some serious desk-bound hardware.

I'm so done with MP3, though. I hate what it does to cymbals, and I can hear the difference between that and Ogg -q7 quite clearly on (good) portable hardware.

Quote:



I've never seen a single (controlled and objective) listening test reveal any statistically meaningful difference between these formats at higher (192 and above) bitrates. Given that MP3 is actually a very good choice due to its universal support on virtually every player and platform available. Switch to something like Ogg and you give up that important feature for nothing in return.

I have done blind testing between LAME v3.97 -V0 --vbr-new against Ogg -q7, and I can tell the difference. As I've said already in this thread, cymbals show off the clearest difference. They have more sparkle and life in the Ogg file, which is considerably smaller. Going from -q7 to -q8 provides a subtle improvement of soundstage depth while still being smaller than the VBR 0 MP3. Granted, you need decent hardware to be able to detect these subtleties, but it's quite possible with even just the Future Sonics Atrio and my Cowon A2.

First of all, we shouldn't give a certain number to VBR mp3. Just call it V0, V2, etc.
Secondly, 320 CBR is theoretically better than V0 but that's only if it's encoded by good programs like Lame.
I'll take Lame-encoded V2 over crap-encoded(FhG, etc) 320CBR any day.

Quote:



You must be kidding. The FhG codec contained in WMP10 or 11 is better than the newest LAME. Taken both 320kb/s, CBR LAME 3.97 is blurry, with worse soundstage and imaging. I don't know what kind of FhG codec you used to obtain such results. In general my ears say (comparing also to the original/lossless): FhG 320 > LAME 320 > FhG 256 > LAME -V0 > FhG 256, and so on.

And I agree that the ogg Vorbis -q8 is yet out of the mp3 format quality range. -q7 is slightly worse than mp3 in terms of midrange reproduction but that's all. -q8 overpowers the mp3 in terms of sondstage, clarity and bandwidth, all of that at ~256kb/s bitrate.

Lame is full-stereo while FhG is joint-stereo.
When I listen to FhG encoded 320CBR, it's not as good as Lame-encoded 192,256,320,V2,V0. If you compare them in Adobe Audition, you will see the structure and the structure of Lame-encoded file is a lot more solid. For that reason, when I download albums in 320CBR format from Waffles/What.CD, I always download one file first to make sure it's not encoded by FhG.

Quote:



LAME is whatever you turn on - Joint Stereo, Stereo, Force Joint Stereo, Dual Channel, and any of them gives worse soundstage comapred to the FhG Joint Stereo. Another thing is that in case of VBR you just increase the bitrate in the Stereo mode while preserving the same (low) midrange quality, but in the CBR mode you lose more bits for encoding channels and then less bits remain for other part of encoding, which causes higher THD level, and I hear it. I don't care about any charts, graphs or statistics, mp3 is a perception based lossy format so aural impressions matter more than similarity in graphs and numbers. FhG wins in these terms.

how certain are you that you can hear the difference between 320k vs original let alone 320k of two different encoders? i am quite skeptical but i'd like to find out for myself. i couldnt get fastencc.exe to work with eac or foobar. i use foobar to encode everything now, it's a great frontend. but i'm having a problem with the commandlines for fhg encoder. for lame i use full bandwidth(no filters) mode with no noise shaping in full stereo mode. i'd like to find out which retains high freq info the best. i want to shy away from joint stereo in fear of losing stereo separation. the original is dual channel so i can't help but think that combining info into a mid channel would cause all kinds of collapsing in imaging.

fastencc.exe has nowhere ever been released free, the version you are using is either the 10s demo encoder or the hacked one with corrupted channel encoding. Use the WMP10/11 or dBpowerAmp Reference or Adobe Audition to have a reliable and up to date FhG encoder. Regarding joint stereo, it's lossless itself and widely used in both lossless and lossy codecs. All it does it creates signals M = L+R and S = L-R. It's fully reversible because L = (M+S)/2 and R = (M-S)/2. Pure maths, no loss but the advantage is that the S contains only the differences between two channels, say no vocal, no drums and all the stuff from the middle. This leads to space savings because instead of coding two times the same information once for L and another time for R, you have it all in M, and then the small reminder in S. So, when the algorythm is correctly tuned (FhG seems to have done it better) there is no loss because of joint stereo encoding.
Regarding the distinction between codec I hear it as I desrcibed. I'm also not afraid to perform a blind test using good headphones/source.