Hello All!

this is my first post here.
normally i don't need to ask for help about lossless codecs.
but what i have here now makes me very curious.

i've got my hands on a norah jones album, 3 times.
2 times it was in ape format, but with different sizes.
and the third time it was the album completly in wav.

so i then also encoded the wav album to ape, and well.
this album is even smaller in size than the other two.

i really don't understand this. i've checked the filesize of
one song tooken of all 3 albums. they did differ abit, but
not much. maybe 300-500kb size differences. this doesn't
make the huge size different for a total album, does it?

and besides that, the first album i got, the songs were bigger
than the songs from the second album. and as i encoded the
wav version, the songs were even abit smaller.

can it be that i got my hands on one original wav>ape encoded
album and two mp3>wav>ape encoded albums? ..

Probably the APE files were encoded with different compression settings, or with different encoder versions. I know that foobar displays the compression in the "file info" box, and there may be other programs which can, as well.

My experience with FLAC (another lossless codec) is that WAV > MP3 > FLAC is generally smaller than WAV > FLAC, because MP3s throw out high-frequency info which can be hard to predict losslessly. So the other two versions of the file probably weren't encoded to MP3 before.

There is a good change it could go the other way too since an mp3 encoder adds noise that is difficult to encode.

edit: decode to wav and use EAC's track compare to see if there is a difference.

Differences can be occured by different rips from CD.
You can also check auCDtect or Tau Analyzer from True Audio Software

Well, from what I know about mp3, the noise added is only in terms of a difference between the original and encoded songs. That is, the energy at any particular frequency may be more or less than the original song. Assuming that more energy is harder to encode (which is a bit simplistic...), quantization noise will influence the lossless compression randomly.
Basically: the noise added by MP3 may either help or hurt lossless compression, since it may add or remove from any given frequency.

(This is all based on my view of MP3, which is by no means complete)

I did a quick test of 11 songs which I had in FLAC format on my hard drive. I did 3 compression settings:
1) WAV > FLAC -6
2) WAV > MP3 aps > WAV (Foobar strong dither) > FLAC -6
3) WAV > MP3 aps > WAV (no dither) > FLAC -6

(1) (2) (3)
849 836 817
755 763 718
766 778 757
675 707 659
748 768 735
409 517 385
502 538 484
625 666 616
520 586 503
629 667 621
788 796 776

Total size in bytes:
1) 479,525,618
2) 509,609,843
3) 466,929,053

Basically, without dither, MP3-encoded FLAC is consistantly smaller than normal FLAC. However, the dithering process adds noise back into the HF area, increasing the FLAC bitrate.

(Note that these songs are almost all classical, hence the low bitrates)

I suppose I forgot to mention dither settings in my previous post. Dithered MP3 may compress more or less than the original song, depending on the song's initial HF content. In my experience, though, they compress worse than the original song.

(Sorry for going completely off topic...)

Monkey should encode the same file using same version and compression type always in the same way.
I recommend you to do the wav compare test which JanS has mentioned.

very offtopic:
I'd like to know if Norah Jones albums are clipped and/or compressed. Since she's somewhere between pop and jazz, chances are that her CDs are (still) relatively well-mastered.

Values from my mpc xtreme encodes:

Norah Jones - Come Away With Me
replaygain_album_gain = -5.84 dB
replaygain_album_peak = 1.055075

Norah Jones - Feels Like Home
replaygain_album_gain = -8.08 dB
replaygain_album_peak = 1.075975

I cant be bothered to scan, but there are very few visable possible clipps in "come away with me". I think its safe to say the mastering is above avarage.

Replaygain of at least -4dB is necessary for an unclipressed record. These values are too low. -5.84 is relatively good, though.

Sorry, but I don't get it. Why on earth would anybody do that? When the audio's been lossy at one time in the chain, that's the max quality it's ever going to be even if it would be later transcoded to FLAC or WAV (but bigger in size).

Or is it that you just wanted to experiment with the compression?

The (possible) decoding process from mp3 (or lossy) to wav has already happened without him having any influence on that. He wants to avoid transcoding (additional loss), of course.

Heh, you obviously never downloaded anything from emule

I certainly didn't want to do it, but I had no say in the matter. One of my friends gave me a CD that she burned from somewhere (don't ask, don't tell ). I listened to it, and it sounded compressed with smeared transients and all that. I looked at a frequency analysis of it, and sure enough the CD was made from MP3s. So, not wanting to transcode, I encoded it to FLAC for storage on my computer.

That made me curious as to the effects of MP3 encoding on FLAC encoding, so I tested it out on a few of my CDs and found that, without re-dithering, WAV > MP3 > FLAC are smaller than WAV > FLAC.