Still not impressed with ogg

[edg]

So I just ripped a CD to my hard drive. (Desert Roses and Arabian Rhythms, if you're curious.) Being the inquisitive sort, I decided I'd rip the CD in Ogg format, and then rip four songs (of the 13 on the CD) in MP3 format, to compare file sizes and sound quality.

I saved an average of 3% in file size. (In particular, track 1 Ogg was 86.9% the size of the MP3; track 2 Ogg was 93.5% the size of the MP3; track 3 Ogg was 103.9% the size of the MP3; and track 4 Ogg was 104.2% the size of the MP3.) I can hear no perceivable difference in quality. I have noticed, however, that Ogg fudges its bitrates - I note that tracks 1 through 4 have average bitrates, in order, of 166 kbps, 179k, 199k, and 200k, regardless of the fact that I told all of them to rip at 192k (and, in fact, the track information lists a "nominal bitrate" of 192k.)

What's more, when I ripped an MP3 of the second track at 179k, I still get no discernible loss in quality, and the file size is within 20k of the Ogg file.

Thus ends my venture into the Ogg format.

Comments

[cherrypep.livejournal.com]

Ah. Err. OK, well, I'm going to have to go into pedantic bastard mood here.

Could you post the command line you used to rip the Oggs?

First things first: if Ogg had been obeying your command to rip at 192k (constant bitrate), the oggs would have been the same length as the MP3s, plus or minus a little due to differences in file format header overhead. Ripping at 192k *means*, "Retain exactly 192 kb per second of information", no matter what the codec. By default, oggenc uses variable bitrate, meaning that your command to rip at 192k was interpreted to mean, "average around and about 192k".

In fact, you have discovered one of the interesting features of Ogg, which is that it does indeed fudge its bitrate unless you command it *very* definitely not to do so -- and why not? It makes perfect sense to throw away useless information; if a signal is simple enough to describe adequately well in less than 192k, then why waste the space? Equally, if another signal is very complex and does not submit happily to lossy compression, then I say go for it, keep the extra information... and so do the authors of Ogg.

Secondly, "bitrate über alles" is not quite the way it goes, when trying to ascertain the quality of lossy compression. In fact, it is hard to judge the quality of compression formats in a rational way, since everyone has their little preferences in terms of sound. One can use various methods of physically checking the similarity of the final ogg or mp3 to its original wav, but it isn't all that revealing in terms of the actual perceivable quality of the various rips (trust me, I did this at university). Instead, one generally ends up doing listening tests, such as this one (http://www.cdfreaks.com/news/9795), and as it so happens, Ogg Vorbis generally comes out at or near the top. Such tests are also not generally conducted using cheap audio equipment, by the way... as it turns out, much of what one hears is the result of one's own setup. So-called scientific testing has to attempt to overcome these issues.

A suggestion: you won't find much difference in Mp3s and Oggs at high bitrate. If you want to hear the actual differences in capabilities of the two formats, you must check at the low-bitrate end of the spectrum. Rip some tracks in Mp3 and Ogg at 128, 96 or lower, and all will become clear.

[edg.livejournal.com]

I actually don't know the command used to rip the Oggs; it was done by a third-party program (Easy CD-DA Extractor, which is so amazingly good at what it does that I encourage anyone using Windows to try out a copy), which doesn't give me the command-line equivalent.

I'll take you at your word about the rest of it. My own listening tests gave me little enough difference at a similar filesize that I have no compunctions about using MP3 to store other people's music (especially given that Sony's software interface for my MiniDisc player doesn't appear to understand anything but CBR MP3s and WAVs). If I (someday!) have music of my own that I want to share, I'll start using Ogg as a distribution format. :)

Apropos of nothing, is etonkin@free.fr still a good address to reach you?

[cherrypep.livejournal.com]

Well at high bitrates I think it is all pretty darn academic really. To be absolutely honest with you I don't waste any sleep on whether my music collection is stored in AC3, Ogg or Mp3, apart from the small point that my mp3 cd player won't play oggs. Low bitrates, ok, but I don't digitise all that many audiobooks anyway. I save my sleep wastage for more important things, such as the relative availability of purple sou'westers and apricot stilton. So when I said I was being anal I meant it, it was literally the subject of a Masters' thesis in which I played some part a couple of years ago and I was just regurgitating stuff I learnt back then :-)

etonkin still works, but is a bit of a spam trap so unlucky emails may not be noticed.