With MPC and Ogg, I just wanted to compare file sizes of the higher quality encodings, but I couldn't really consider using either one because of compatibility. But in terms of quality/size, I am *very* impressed with what I've heard in those formats. Granted, my PC can play virtually anything, but here's my criteria...
My primary encodings all have to be able to play on my PC *and* in my car when I copy them to the Phatbox. I don't have the space to keep WAV's to encode from, and I'm not going to have time to re-rip and encode into a different format every time I want to change the music in my car. High-bitrate to lower-bitrate lossy transcoding is an option for my portable MP3 player, but *not* for my car. The Phatbox only speaks MP3, WAV, FLAC, and WMA. I'd love to keep all of my music in MPC format and regain some 10% or more of my disk space, but alas, the Phatbox is blind to MPC.
The same goes for Ogg. I even sent PhatNoise customer support an e-mail asking whether they had plans to support the Ogg Vorbis encoding format. They said "no plans at this time".
And I was very un-impressed by what I heard from WMA9. I don't know if I used a bad compile or something, or if there was a bug in its psychoacoustic model, but it completely failed my "Kalifornia" test even at 373kbps with a file size 900kB larger than the Lame --api one! Scored 20 out of 20 in WinABX...*easily*! (I use Fatboy Slim's "Kalifornia" as one of my several baseline tracks for testing, since it's typical of my listening taste. If it can't get close on "Kali", then I've got no use for it.)
So, dropping WMA narrowed everything down to WAV (too big), FLAC (maybe too big), and MP3. Did some math, figured I could have about 50 albums on 20GB in FLAC, 150-160 in --alt-preset insane, 180 or so in --alt-preset extreme and just over 200 in --alt-preset standard. Gotta have more than 50..so FLAC's out. Now it was down to which Lame quality setting for MP3. Transparency was of the highest importance since I could live with as few as 150 encoded albums just fine, if that's what it took.
That's when the ABXing started. As for my results, I think most people could hear much of what I did for themselves...
> The Beatles - Rubber Soul - Norwegian Wood (This Bird Has Flown). Listen to the passage between the timemarks from 5.0 to 7.8 seconds. It's the very hard guitar strum by John that's so fierce that it's a little out of key. The WAV version (to my ears) returns to key in the split second as it fades, and in the --aps version the effect lingers a little longer...very faint, but I was able to detect it on almost every try. WinABX gave me a <0.1% probability of guessing.
> Fatboy Slim - You've Come A Long Way, Baby - Kalifornia. From what I've read, one of the hardest popular songs to encode, mixed by "King Codec Killer" himself.

Anywhere you choose among the first 14 seconds of the song...'nuff said. The --aps-encoded file I tested had a faint crackling noise under the synthesized voice in this passage. Scored 100% in WinABX...20 out of 20. But it sounded *much better* than the 192 kbps FhG version I had been listening to before. It was harder to detect in the --ape file...scored <1.8% probability of guessing. --api was transparent to me, and was the first MP3 that was ever able to reproduce this song this well (to my ears, anyway).
> The Beatles - Sgt. Peppers - A Day in the Life. The peak amplitude note almost at the end of the song (can't remember the timemark), where all the instruments are building up to it. Right after the note, there is an echo prevalent even in the WAV. Compared to --aps, I could just barely detect less lingering tendency. --ape and --api were transparent to me. Probability of guessing when I tested --aps against the original WAV was <0.9%.
> Nine Inch Nails - Deep. (Track 2 from the soundtrack for Lara Croft: Tombraider.) Timemark from 14.3 seconds to 17.0 seconds. It's very faint, but the original WAV has a little more "liveness" than --aps. --ape and --api are transparent to me. My probability of guessing for the --aps one was <0.1%.
I haven't ripped either of my "On the Floor at the Boutique" albums yet (Lo Fidelity Allstars and Fatboy Slim). I'll test more tracks from those albums, as well as Pretty Hate Machine, Downward Spiral, Fat of the Land (Prodigy) and some classical as well.
QUOTE
Another advantage of Musepack and Ogg Vorbis is gapless playback and integrated ReplayGain support. MP3 is not gapless between tracks (live and DJ mix albums).
As for gapless playback...it would be nice, but it's not important enough to me yet. Integrated replaygain is not something I really need either as my method is to have EAC run all of my ripping and --api encoding, then as each album is finished I dump it into MP3Gain...a manual step, but nothing for me to complain about. (And, of course, I never gain test tracks.)
QUOTE
It's clear you're being careful and objective in ABXing, and you seem to have done well at picking out artifacts if you can ABX lame --aps from the original occasionally (usually pretty difficult except on killer samples).
It's very hard for me...I must have gone through 10 songs (that I thought would be likely to artifact) for each *1* that I could ABX aps from WAV on. And I'm definitely not aps-bashing...like I've said before, it's far better than my previously used MP3 codecs at equivalent bitrates to my ears. I can say that it's not quite as hard to find complex, high-transient, special-effect-ridden techno/electronica songs that will "break" a codec vs. many other music types...except, strangely enough, The Beatles, some of which I discovered (almost by accident) I could ABX. Then again, no one ever said that there is any such thing as an infallible MP3 configuration.
QUOTE
While on the subject, you might even want to consider Wavpack lossy mode (a lossless-type compression algorithm with a certain amount of allowed noise to lower bitrate), which some people support as a medium-sized format suitable as a source for transcoding (particularly to MiniDisc).
Again, the compatibility thing...2/3 of my devices won't like .wv files. Then again, if I had 80 GB or more, I would probably go FLAC or WavPack lossless...compatibility requirements would point to the former.
As for --api, I don't hear many people considering/discussing it in most HA threads where it might otherwise be discussed. I'm starting to think that --api gets short-shrifted because it's not as compact as --aps or --ape, but lossy and therefore not comparable to FLAC, APE, etc. My contention is otherwise...if your space requirements demand lossy, your lossy compability requirements say "MP3", but you want the *best* possible quality (and if you listen to music in which you can tell the difference), then I think that a 22% compressed file size is a very reasonable compromise for the highest MP3 sound quality possible (by ISO standards, anyway)...