Kuro5hin.org: technology and culture, from the trenches
create account | help/FAQ | contact | links | search | IRC | site news
[ Everything | Diaries | Technology | Science | Culture | Politics | Media | News | Internet | Op-Ed | Fiction | Meta | MLP ]
We need your support: buy an ad | premium membership

[P]
MP3 and Ogg Vorbis: A User's Perspective

By regeya in Technology
Thu Jan 25, 2001 at 06:32:43 PM EST
Tags: Software (all tags)
Software

A fairly new audio codec has been getting a fair amount of attention lately, possibly because it's developed under an open source model. While that may be enough of a selling point for many, others need more convincing than mere Free Software or Open Source rhetoric. To that end, I decided to do pit LAME, an open source MP3 encoder some claim to be the best MP3 encoder available against OggEnc, the Xiphorous project's official Ogg Vorbis encoder.


ADVERTISEMENT
Sponsor: rusty
This space intentionally left blank
...because it's waiting for your ad. So why are you still reading this? Come on, get going. Read the story, and then get an ad. Alright stop it. I'm not going to say anything else. Now you're just being silly. STOP LOOKING AT ME! I'm done!
comments (24)
active | buy ad
ADVERTISEMENT
What is Ogg Vorbis?

According to the authors of Ogg Vorbis, it's an audio encoder designed to replace MPEG 1 Layer 3 as an Internet audio exchange format. The reason behind this is that, while the specification for MPEG 1 Layer 3 is available (as is reference source) MPEG is covered by several patents. That may not seem to be a problem, except that Thompson, the company that owns the patents, charges licensing fees for MPEG-related processes. That may not be a problem if you use software that's already had its licensing fees paid for (such as MusicMatch Jukebox, a commercial CD ripper/encoder/portable device manager), but you get into legally ambiguous territory if you use a non-commercial encoder such as LAME. Add to this that Thompson also owns patents on processes related to streaming MP3 over a network, and the fees pile up.

As for the name Ogg Vorbis: Vorbis is the name of the audio format. Ogg is a wrapper file format, roughly analogous to QuickTime. In the future, Ogg will be used for video (the upcoming Tarkin video codec) as well as other projects. For now, though, Ogg serves as a wrapper for Vorbis.

Not being a mathmetician myself, all I can say is that, according to the authors, Vorbis uses MDCT (modified discrete cosine transform), the same as MP3. However, that is apparently where similarities end. Vorbis is being designed not just for encoding/decoding local content, but also for network streaming. The authors claim that their format will be able to, among other things, allow a streaming server adjust its output bitrate according to the networking capabilities of the client. With MP3, that's covered by a patent, and therefore gets into legally dodgy areas if one wanted to write such capability into an open-source streaming server such as Icecast. See the January 2001 Linux Journal for an article written by Jack Moffitt, or the interview with Monty on Binary Freedom for more details.

Right now Ogg Vorbis is early in development. As the authors say, their math shows that their codec can do low bitrates, but they're not there yet. So for now, heres the choices you get with OggEnc:

INPUT FILES:
OggEnc input files must currently be 44.1kHz, 16 bit stereo WAV files. You can specify taking the file from stdin by using - as the input filename. Alternatively, the --raw option may be used to use a raw PCM data file, with the same restrictions as above.
In this mode, output is to stdout unless an outfile filename is specified with -o
MODES:
OggEnc currently supports 5 different modes. Each of these is a fully VBR (variable bitrate) mode, but they vary in intended average bitrate. The bitrate option (--bitrate, -b) will choose the mode closest to the chosen bitrate. The 5 modes are approximately 128,160,192,256, and 350 kbps (for stereo input. Halve these numbers for mono input). The default is the 160 kbps mode.

So, since Ogg Vorbis's main use is apparently to produce CD-quality content for the time being, I'm going to compare OggEnc's output quality to the MP3 output quality of LAME. LAME has a ridiculously long list of features, and I won't post them here for brevity's sake. The real problem with this test is that it's too easy to skew the results either unfairly towards Ogg Vorbis or toward LAME MP3 output. Therefore, I'm going to compare default variable bitrate output from both encoders. At the time of this writing, while LAME has a build option to allow one to encode tracks as Ogg Vorbis and to convert files to/from MP3, libvorbis has apparently undergone major changes, which has broken LAME's Ogg Vorbis support.

Some proponents of both claim that these settings are sufficient to achieve CD quality. To judge the results, I'm using only my ears, and am using xmms as the audio-playback software. The platform I'm using for both encoding is a K6-300 with 64MB of RAM. This will account for the slow encoding times; I'm told that OggEnc will run at greater-than-realtime on moderately newer machines. Slackware Linux 7.2 is the OS.

The gory details:

Glenn Gould
Bach: The Well-Tempered Clavier
Track 1 - No. 1 in C major, BWV 870 (3:00)

oggenc:
Done encoding file "gould.ogg"
File length: 3m 00.0s
Elapsed time: 8m 48.6s
Rate: 0.3410
Average bitrate: 175.2 kb/s

LAME version 3.88 (alpha 8, Jan 17 2001 19:39:24)
(http://www.mp3dev.org/)
Using polyphase lowpass filter, transition band: 17960 Hz - 18494 Hz
Encoding gould.wav to gould.wav.mp3
Encoding as 44.1 kHz VBR(q=4) j-stereo MPEG-1 Layer III (ca. 9x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
6902/6902 (100%)| 9:18/ 9:18| 9:19/ 9:19| 0.3227x| 0:00
average: 154.5 kbps LR: 104 (1.507%) MS: 6798 (98.49%)

-rw-r--r-- 1 shane users 3478842 Jan 22 21:13 gould.mp3
-rw-r--r-- 1 shane users 3947349 Jan 22 16:13 gould.ogg
-rw-r--r-- 1 shane users 31799084 Jan 22 15:55 gould.wav

Comments: no difference

Miles Davis Quintet
The Best Of Miles Davis Quintet, 1965-68
Track 4 - Gingerbread Boy (7:46)

oggenc: Done encoding file "miles.ogg"
File length: 7m 46.0s
Elapsed time: 23m 13.3s
Rate: 0.3349
Average bitrate: 173.5 kb/s

lame:
Encoding as 44.1 kHz VBR(q=4) j-stereo MPEG-1 Layer III (ca. 9x) qval=2
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
17865/17865 (100%)| 22:09/ 22:09| 22:10/ 22:10| 0.3511x| 0:00
average: 200.2 kbps LR: 9230 (51.67%) MS: 8635 (48.33%)

-rw-r--r-- 1 shane users 11667434 Jan 22 20:58 miles.mp3
-rw-r--r-- 1 shane users 10120455 Jan 22 18:30 miles.ogg
-rw-r--r-- 1 shane users 82315340 Jan 22 18:06 miles.wav

Comments: Very little difference. I could tell no difference between the mp3 and the wav. There seems to be a slight amount of clipping on the Ogg Vorbis file 53 seconds into the song. The original track shows signs of clipping, but in the Ogg Vorbis track, it's more pronounced. I will note that I had to listen extremely carefully to hear it.

Jerry & Joel Goldsmith
Star Trek-First Contact
Opening Title (4:19)

oggenc:
Done encoding file "startrek.ogg"
File length: 4m 19.0s
Elapsed time: 12m 16.4s
Rate: 0.3522
Average bitrate: 152.0 kb/s

lame:
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9930/9930 (100%)| 12:11/ 12:11| 12:12/ 12:12| 0.3545x| 0:00
average: 136.9 kbps LR: 1107 (11.15%) MS: 8823 (88.85%)

-rw-r--r-- 1 shane users 4433020 Jan 22 23:51 startrek.mp3
-rw-r--r-- 1 shane users 4928129 Jan 22 23:36 startrek.ogg
-rw-r--r-- 1 shane users 45751148 Jan 22 22:41 startrek.wav

Comments: There is no audible difference between the wav and the ogg file. However, in the mp3, there seems to be a somewhat artificial quality to the horn section, and the string section sounds as if some of the members are out of tune. This doesn't occur in the ogg file.

Godflesh
Songs Of Love And Hate
Track 11 - Almost Heaven (5:41)

oggenc:
File length: 5m 41.0s
Elapsed time: 16m 51.6s
Rate: 0.3371
Average bitrate: 163.7 kb/s

lame:
13055/13055 (100%)| 14:38/ 14:38| 14:44/ 14:44| 0.3881x| 0:00
average: 175.0 kbps LR: 281 (2.152%) MS: 12774 (97.85%)

-rw-r--r-- 1 shane users 7453475 Jan 22 20:34 godflesh.mp3
-rw-r--r-- 1 shane users 6976726 Jan 22 19:59 godflesh.ogg
-rw-r--r-- 1 shane users 60152444 Jan 22 19:07 godflesh.wav

Comments: This track surprised me. I couldn't tell a difference between the wav, the ogg vorbis file, or the mp3. I expected to hear characteristic MP3 distortion in this track, because the track contains a fair amount of cymbals and high-frequency noise/distortion. I heard no artifacts in either encoding.

Slayer & Ice-T
"Judgement Night" Soundtrack
Track 5 - Disorder (4:59)

oggenc:
Done encoding file "slayer.ogg"
File length: 4m 59.0s
Elapsed time: 15m 40.2s
Rate: 0.3184
Average bitrate: 166.7 kb/s

lame:
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
11463/11463 (100%)| 11:36/ 11:36| 11:37/ 11:37| 0.4297x| 0:00
average: 211.6 kbps LR: 2263 (19.74%) MS: 9200 (80.26%)

-rw-r--r-- 1 shane users 7912782 Jan 22 23:03 slayer.mp3
-rw-r--r-- 1 shane users 6238709 Jan 22 22:38 slayer.ogg
-rw-r--r-- 1 shane users 52814204 Jan 22 22:00 slayer.wav

Comments: Both the mp3 and the ogg vorbis file had audible artifacts on this one. About 1:59 into the song, the track is heavily-laden with cymbals and sounds as if it were printed hard to tape. In the mp3 track, characteristic mp3 distortion is audible. In the Ogg Vorbis track, the clipping in the original track becomes much more pronounced.

Conversion of MP3 tracks to Ogg Vorbis tracks:

To do this, I did a

for i in `ls *.mp3`
do
mpg123 -w blah.wav $i
oggenc blah.wav
mv blah.ogg $i.ogg
done

Listening results were encouraging. I fully expected these tracks to sound terrible, but in most cases, I couldn't tell a difference between the MP3 and the converted Ogg Vorbis track.

Final Thoughts:

For an end user, there may or may not be a reason to switch. In my test, I didn't allow lame to use the highest quality psychoacoustic analysis, which perhaps would have reduced or eliminated the artifacts I heard in the MP3s. Filesizes in these tests are also a bit unfair, because it should be possible to tune LAME's upper and lower bitrate thresholds to match those of OggEnc. Also, in informal polling, I have found that some people see no reason to switch. The most common answer I get is "even if I switched from using LAME to oggenc, and I made all my rips Ogg Vorbis, my music collection would still be illegal." My own stumbling block is that I own a Diamond Rio PMP300, which has no facility for flash-upgrading the system to support Ogg Vorbis.

The people this may appeal to most are content providers. With Shoutcast, for example, to be 100% legal about your operation (and ignoring copyright issues for a moment) one would have to use a licensed encoder, and for Shoutcast to stream MP3s of varying bitrates in a different bitrate, Shoutcast has to use licensed technology. This prevents projects like Icecast from doing on-the-fly bitrate conversion.

With Ogg Vorbis, the format is patent-free, which means no patent licensing fees. On top of that, the source code is free. In short, there are no fees for using Ogg Vorbis! It may well be that content providers, not end users, will help determine whether or not Ogg Vorbis sets the world on fire, or merely fizzles.

Sponsors

Voxel dot net
o Managed Hosting
o VoxCAST Content Delivery
o Raw Infrastructure

Login

Related Links
o Xiphorous
o licensing fees for MPEG
o MusicMatch Jukebox
o Vorbis
o January 2001 Linux Journal
o Monty
o Binary Freedom
o Also by regeya


Display: Sort:
MP3 and Ogg Vorbis: A User's Perspective | 54 comments (47 topical, 7 editorial, 0 hidden)
A couple of critiques (3.77 / 9) (#7)
by CanSpice on Thu Jan 25, 2001 at 05:05:25 PM EST

Excellent comparison, but I have a couple of comments/criticisms about the comparison technique:

1) You should've done a doubleblind comparison. You even said that you were expecting a certain result out of a test (even though you didn't find what you expected). This can lead to bias in determining the best, even if you think you're being unbiased.
2) What sort of audio equipment did you use? What kind of soundcard? What kind of speakers/headphones? If you're using crappy equipment, odds are you wouldn't be able to discern subtle differences between the different files.

Still, a really good comparison that'll hopefully sway more support into the Ogg Vorbis camp.
--- I don't have a sig.
I considered it to be a real-world test ;-) (4.00 / 2) (#22)
by regeya on Thu Jan 25, 2001 at 07:12:44 PM EST

Frankly, I doubt that most people have Bose powered speakers on their computers with a sweet Turtle Beach card, with an ultra-quiet power supply and gold cabling going from that sweet Turtle Beach card to the speakers. Truth be told, I have a mediocre sound card (an ES1371) some speakers I picked up at Radio Shack in 1997 for $30 that originally sold for $130. I listened through those and a set of cheap-ish Jensen headphones.

Yeah, there are subtle problems that you might pick up with a nice setup. How many home users have a supurb audio setup? How many college kids with 89 hours of MP3s? Chances are, you're going to have a somewhat loud power supply fan going, and perhaps you'll be doing some typing or some other chore while you're listening to music (like I'm doing right now. :-) I can honestly say that most people I have talked to say that 128kbps MP3s sound as good as CD quality on their home setup; I've yet to find an encoder that can do CD-quality at 128kbps on my home setup.

[ yokelpunk | kuro5hin diary ]
[ Parent ]

Uh... (4.00 / 1) (#26)
by Refrag on Thu Jan 25, 2001 at 09:42:16 PM EST

How is Bose going to help you? Remember: friends don't let friends buy Bose.

I'm sorry, but Bose just isn't very detailed. The speakers connected to my computer are Boston Acoustics' BA735s, and right now I'm using a Labtec headset for VoIP to listen to MP3s. However, were I going to conduct a listening test between MP3 and OggVorbis I would find a way to hook up my AKG stereophones.

I know when I was in college I had a respectable audio setup, however you are right that probably 99% of the people out there are using a noisy SoundBlaster rather than something cleaner like the Turtle Beach cards.

By the way, I plan to convert all of my music files to OggVorbis once the encoder (or WinAmp plug-in) is mature enough to create meta-data on the files.

Refrag

Kuro5hin: ...and culture, from the trenches
[ Parent ]

Heh (4.50 / 2) (#29)
by regeya on Fri Jan 26, 2001 at 01:43:50 AM EST

You're right, I suppose. What I found in dorms is that people either a.) stuck with what shipped with their machine, as far as speakers went or b.) got whatever went "BOOM" when they fired their rocket launcher.

Other than the low end, I won't complain too much about my PC speakers or soundcard. They're not incredibly noisy, at least not for casual listening, and I really don't miss the extremely low frequencies or the extremely high frequencies most the time. Dear God, I used to have an old SoundMan WAVE card. Talk about noise--the campus I was on had a public radio station, and I could use it to listen to the broadcast, whether I wanted to or not. :-) I guess after years of cassettes and vinyl I really don't care about a little hiss.

[ yokelpunk | kuro5hin diary ]
[ Parent ]

Get the newest version (5.00 / 1) (#38)
by jsoderba on Fri Jan 26, 2001 at 12:09:04 PM EST

By the way, I plan to convert all of my music files to OggVorbis once the encoder (or WinAmp plug-in) is mature enough to create meta-data on the files.
Beta 3 supports the same metadata as ID3 (well, not genre). Partial output of oggenc --help on Win2K:
-c, --comment=c Add the given string as an extra comment. This may be used multiple times.
-d, --date Date for track (usually date of performance)
-N, --tracknum Track number for this track
-t, --title Title for this track
-l, --album Name of album
-a, --artist Name of artist
If multiple input files are given, then multiple instances of the previous five arguments will be used, in the order they are given. If fewer titles are specified than files, OggEnc will print a warning, and reuse the final one for the remaining files. If fewer track numbers are given, the remaining files will be unnumbered. For the others, the final tag will be reused for all others without warning (so you can specify a dat once, for example, and have it used for all the files)
The Winamp plugin does not support editing, but it does show the data in the playlist.

[ Parent ]
Oh yeah... (2.00 / 1) (#27)
by Refrag on Thu Jan 25, 2001 at 09:51:02 PM EST

I also wanted to say that I regret that I didn't have a chance to vote on this article while it was in the moderation queue. I would have voted it to the frontpage, Kuro5hin needs more empirical research such as this.

Refrag

Kuro5hin: ...and culture, from the trenches
[ Parent ]

Bose = suck (4.00 / 2) (#41)
by fluffy grue on Fri Jan 26, 2001 at 02:06:17 PM EST

You don't want Bose. You don't even want speakers, unless you're willing to spend thousands of dollars on a pair. What you want are good headphones. Get the Grado SR-60s; they're only $70 (a bargain at that price!) and sound fucking incredible. I reviewed them on InTune recently if you want my detailed report on 'em. :)

As far as the mediocrity of your soundcard, ESS's soundcards actually tend to be quite clean. My ES1868 is probably the best-quality soundcard I've ever had (and believe me, I've spent well over $600 on soundcards trying to find the Perfect One - and ironically, a $6 ES1868 was it). It still has some noise, but not nearly as much as, say, the SB AWE64 Gold. The only card I've had with less noise was the Hanmesoft Soundtrack 128 Ruby ISA, and that one's drivers were so incredibly BROKEN (under Windows, which is where I still do most of my sound work) that its crystal-clear audio wasn't worth the $230 I paid for it. :P
--
"Is not a quine" is not a quine.
I have a master's degree in science!

[ Hug Your Trikuare ]
[ Parent ]

Grados (none / 0) (#48)
by Refrag on Sun Jan 28, 2001 at 12:16:46 AM EST

Grados are nice, but they aren't very comfortable because of the hand-cut foam surrounds that they use. I'd much rather stick with my AKGs.

Refrag

Kuro5hin: ...and culture, from the trenches
[ Parent ]

Uncomfortable? Huh? (4.00 / 1) (#49)
by fluffy grue on Sun Jan 28, 2001 at 02:23:45 AM EST

Aside from being a little heavy, my Grados are quite comfy, even after long (i.e. 6 hour) listening periods. :) Admittedly, I've never even heard of AKGs, so I'm gonna go look 'em up on AudioReview right now...

Wow, they get pretty high marks and seem to have a pretty low price. Still, having just paid $70 for my Grados, I'm not about to buy more headphones. ;) The only thing I don't really like about the Grados is that they are open-air (which is an advantage much of the time too), so I can't just shut out the outside world when I want to. If they just had two sets of earpads, one open-air and one closed-air, I'd be happy as a grue who just ate an adventurer. :)
--
"Is not a quine" is not a quine.
I have a master's degree in science!

[ Hug Your Trikuare ]
[ Parent ]

[OT] speakers (none / 0) (#54)
by Refrag on Tue Jan 30, 2001 at 07:55:30 PM EST

Fluffy, I just bought some B&W Nautilus 805s over the weekend. Those speakers rule! :)

BTW, I've always loved your name, even thought I always thought of grue's as having thick leather like skin, rather than furry.

Anyway, I'm going to check out your music at MP3.com. Have you earned any money from there?

Refrag

Kuro5hin: ...and culture, from the trenches
[ Parent ]

This doesn't seem quite right... (3.50 / 10) (#8)
by WesHertlein on Thu Jan 25, 2001 at 05:17:21 PM EST

I have a minor gripe with the testing procedure here...

Conversion of MP3 tracks to Ogg Vorbis tracks:

To do this, I did a
for i in `ls *.mp3`
do
mpg123 -w blah.wav $i
oggenc blah.wav
mv blah.ogg $i.ogg
done

For those that don't speak shell... this seems to indicate that mp3 files were converted to wav files, which were then converted to ogg files.

To be a bit more scientific about this, shouldn't the process have been started with wav files that were converted to mp3 and to ogg? The wav files are pretty much lossless... well, actually lossless I guess, given that you trust the process taking the wav from a CD or wherever. I was under the impression that both mp3 and Ogg Vorbis "threw out" some info to get the compression they achieve.

This then leads to the question of why the wav files occasionally sounded better... is it possible that something wasn't quite right with the mp3 player being used?

Wes



they /are/ both lossy. this makes no sense (2.50 / 2) (#10)
by Defect on Thu Jan 25, 2001 at 06:00:52 PM EST

I'm glad someone else caught that too, i was thinking that i was just misunderstanding something. Essentially what this process is is bloating a compressed file and the compressing it again, which makes no sense. It's like converting a compressed jpeg to a bitmap, then back to a jpeg again and looking for compression issues.
defect - jso - joseth || a link
[ Parent ]
actually... (3.00 / 2) (#14)
by UrLord on Thu Jan 25, 2001 at 06:28:16 PM EST

It's like converting a compressed jpeg to a bitmap, then back to a jpeg again and looking for compression issues.

I think it is more like going from jpeg to bitmap to png.

We can't change society in a day, we have to change ourselves first from the inside out.
[ Parent ]

wrong (4.50 / 2) (#19)
by regeya on Thu Jan 25, 2001 at 07:02:26 PM EST

PNG is lossless. It uses clever methods to make zlib compress the image better, but it throws nothing away. JPEG does.

[ yokelpunk | kuro5hin diary ]
[ Parent ]

Well... (4.00 / 1) (#25)
by UrLord on Thu Jan 25, 2001 at 09:41:49 PM EST

I stand corrected... ditto to the person below you ;)

So it was more of a jpeg to bitmap to another lossy image file type.

We can't change society in a day, we have to change ourselves first from the inside out.
[ Parent ]

Nope (4.00 / 1) (#21)
by Eimi on Thu Jan 25, 2001 at 07:11:08 PM EST

png is lossless, like bitmap, and unlike jpeg.

[ Parent ]
Lossy to lossless to lossy (4.50 / 2) (#28)
by pin0cchio on Thu Jan 25, 2001 at 10:11:30 PM EST

I think it is more like going from jpeg to bitmap to png.

No, that'd be more like MP3 to wav to FLAC, a fast lossless audio codec. What the OP tried in the original experiment is more like jpg to bmp to JPEG2000, a wavelet compression format.


lj65
[ Parent ]
Exactly (5.00 / 2) (#20)
by regeya on Thu Jan 25, 2001 at 07:05:14 PM EST

I should have made that more clear. When LAME is fixed, one of the neato features you'll have is the ability to convert an MP3 to Ogg Vorbis, or an Ogg Vorbis file to an MP3, with one command-line option. Since LAME's Ogg Vorbis support seems to be broken at the moment, I did it anyway. One thing I've heard people ask over and over is "Why would you do that? That'd sound like shit." Well, in this case, it didn't. You wouldn't want to start with a 96kbps MP3 and convert it to Ogg Vorbis, but if you start with a good MP3, it'd be OK.

Sorry about that. I actually posted the story when I got tired of messing with it. ;-)

[ yokelpunk | kuro5hin diary ]
[ Parent ]

Guess (none / 0) (#45)
by sety on Sat Jan 27, 2001 at 01:28:18 AM EST

I'm guessing that this would be just an added command switch and there wouldn't be anything in the backend besides a GPL decoder. So the question is: Is there any advantage of this way over a shell pipe? Probably only for newbies. You would still be doing mp3-->wav-->ogg even if it does it in all one step with the data stream.

[ Parent ]
going via wavs seems wrong (none / 0) (#50)
by jovlinger on Mon Jan 29, 2001 at 01:16:44 PM EST

If I may read between the lines and speculate for a bit:

The switch on the codec to translate from mp3 to ogg directly is to avoid the qualitification errors that all DCT transform based encoders have. I think.

Let's stick with the JPG example: if I have a JPG and want to rotate it 90 degrees, I could decompress it, rotate it, and then resave it. However, that would impose two rounding errors and one quantification error into the rotated image that weren't there in the original. This s inherent in the DCT approach. Alternatively, it is possible to rotate images by just reshuffling the encoded blocks of the JPG picture -- without decompressing them. That way, the rotated images have exactly the same quality as the originals.

Likewise, I suspect that oggenc is able to reuse the blocks from the mp3 file (at least partially) thus avoiding the DCT errors that are introduced by going via wavs.

Mind you, this is pure speculation and is likely wrong in at least the details, if not the broad strokes, so add salt to taste.

[ Parent ]
That's just one test (4.50 / 2) (#11)
by Paul Crowley on Thu Jan 25, 2001 at 06:11:14 PM EST

Most of the tests were done clean, but he gave this procedure a go to see how bad it would be. It was better than he expected.

Rusty just fixed a formatting error that made it easy to read the article the way you did, so it should be easier to see it's just one test now.
--
Paul Crowley aka ciphergoth. Crypto and sex politics. Diary.
[ Parent ]

All clear (4.00 / 1) (#13)
by WesHertlein on Thu Jan 25, 2001 at 06:23:11 PM EST

Good call on that one... the formatting fix makes the test and explanation much clearer.

Guess I should've made my comment editorial. *sigh*

Wes



[ Parent ]
Separate Experiment (5.00 / 2) (#12)
by thedward on Thu Jan 25, 2001 at 06:15:12 PM EST

Its unclear, but I think he was reffering to a separate experiment. He was converting mp3 to ogg to see how well that'd work; I assume he was starting with fresh ripped wav's for the first part. If you read carefully he has a brief comment about the quality of the files that were converted to ogg via wav. The earlier comments on quality were based on wav->mp3 and wav->ogg;

[ Parent ]
stupid point (5.00 / 3) (#32)
by mikpos on Fri Jan 26, 2001 at 09:37:41 AM EST

Also, I feel obliged to point out the UUOL (useless use of ls). for i in *.mp3 is clearer and marginally more efficient (i.e. superior in every way) to for i in `ls *.mp3`.

[ Parent ]
Existing MP3 collections? (4.00 / 1) (#44)
by magney on Sat Jan 27, 2001 at 12:26:10 AM EST

I think the intent of this test was to determine how badly you'd hose your existing MP3 collection if you converted it all to Ogg Vorbis. The results suggest that the answer is "not too badly".

In any case, if you're a serious audiophile, you aren't using MP3 anyway. :-)

Do I look like I speak for my employer?
[ Parent ]

Audio tests.. (4.50 / 4) (#9)
by molo on Thu Jan 25, 2001 at 05:56:35 PM EST

There are two things you can do.. listen and measure.

Listening is only half of the test. While I'm glad you did take the time to do this and explain your reasons for your choices, as CanSpice said, we need more information on your audio setup. He is right in asking for a double-blind test, as your expecations do figure into your perceptions. In addition, most audio critics will not accept just a double-blind test, but they instead want a blind A-B test.

An A-B test is a listening comparison between two audio sources whose levels have been matched. The comparison is done by rapidly switching from one to another. The listener has a chance to listen to the playback as many times as is needed to determine which is better.

In addition to listening, we need to properly analyse the output of the two codecs. This probably means decompressing and piping the output through a spectrum and level analysis. You have a nice and wide range of music to look at, but we need additional more technical audio to test with. Think sweeps, steps, etc.

You definitely have a good start for analysis, but there is much more to be done. I will eventually be making an archive of all my CDs and am yet undecided on what format to store them in.

If anyone has any pointers on more technical measurements and analysis of the Vorbis encoding, I'd love to see some links. Also, does anyone know of any Free audio analysis software for *nix? Thanks.

--
Whenever you walk by a computer and see someone using pico, be kind. Pause for a second and remind yourself that: "There, but for the grace of God, go I." -- Harley Hahn

are you sure? (4.00 / 1) (#30)
by dabadab on Fri Jan 26, 2001 at 04:50:06 AM EST

Using measurements as a tool of comparing different lossy compressions does not seem to be a good idea, as mp3/ogg is about producing a sound different from the original, a sound, that we, humans (with our ears, and not with spectrum analizers) still perceive to be the same as the original.
--
Real life is overrated.
[ Parent ]
Absolutely! (none / 0) (#31)
by molo on Fri Jan 26, 2001 at 08:48:08 AM EST

Just because you didn't notice it, doesn't mean that the difference isn't there. Obviously the ideal would be to completely match the original signal. If one encoder gets you closer to that original signal, isn't it (by definition) better? Also, what if one of the encoders stops at 16khz and doesn't encode anything higher (high frequency sounds are more expensive to encode)? You want to know about it. You also want to be able to back up your listening results with measurements.

Check out r3mix.net and click on 'Analysis' for a comparison like I am talking about. This comparison was done with Xing, Lame, Blade, AudioCatalyst, MP3Enc, N2MP3, SoundJam, Nero and a few other encoders. I think you'll find the results of most mp3 encoders, even in 'high quality' mode surprisingly bad.

--
Whenever you walk by a computer and see someone using pico, be kind. Pause for a second and remind yourself that: "There, but for the grace of God, go I." -- Harley Hahn
[ Parent ]

not at all (none / 0) (#37)
by Ender Ryan on Fri Jan 26, 2001 at 11:02:09 AM EST

"If one encoder gets you closer to that original signal, isn't it (by definition) better?"

Not at all. Both these audio compression schemes remove sounds that we as humans cannot hear, but are nonetheless there. It has nothing to do with being close to the original, it's about making it sound (to humans) close to the original.

At least that's how I understand it...


-
Exposing vast conspiracies! Experts at everything even outside our expertise! Liberators of the world from the oppression of the evil USian Empire!

We are Kuro5hin!


[ Parent ]

Sorta.. (none / 0) (#40)
by molo on Fri Jan 26, 2001 at 02:04:49 PM EST

"it's about making it sound (to humans) close to the original."

I think it would be more accurate to say that its about removing hard-to-percieve sounds. Different people hear things differently and there is no absolute about what can and can't be heard. Achieving a signal closer to the original is guranteed to satisfy more people than one skewed from the original. Besides, how can having a signal closer to the original be a BAD thing?

--
Whenever you walk by a computer and see someone using pico, be kind. Pause for a second and remind yourself that: "There, but for the grace of God, go I." -- Harley Hahn
[ Parent ]

close to the original (none / 0) (#43)
by Ender Ryan on Fri Jan 26, 2001 at 05:15:08 PM EST

The closest to the original won't necessarily sound the best. It could almost exactly like the original except have some cracks and pops. But you're right, being close to the original obviously isn't bad at all.


-
Exposing vast conspiracies! Experts at everything even outside our expertise! Liberators of the world from the oppression of the evil USian Empire!

We are Kuro5hin!


[ Parent ]

pareto (none / 0) (#51)
by jovlinger on Mon Jan 29, 2001 at 01:57:42 PM EST

Economist Paretto formulated the relations of paretto optimality as A is paretto optimal to B if A is equivalent to B in every way, except for those that it is better in.

This is clearly induces a partial ordering -- if higher numbers are better, then (10,3) is optimal to both (8,3) and (10,1). However, we are unable to say anything about how (8,3) relates to (10,1).

Likewise, if encoding A is closer to the original than the encoding B in some aspects, while NEVER being worse, then paretto optimality allows you to say that A is paretto optimal to B. which is damn close to being better.

However, determining what to measure in order to determine this is nigh impossible. Lenght of playback seems wrong. The pointwise difference in amplitude from the orginial? Maybe. Perhaps the fourier spectra?

So I think I just argued myself out of a point. Damn.

[ Parent ]
What about Lame vs. Lame? (3.83 / 6) (#15)
by thedward on Thu Jan 25, 2001 at 06:32:35 PM EST

Isn't this more a test of oggenc vs. lame than mp3 vs ogg? Recent versions of lame will encode to ogg as well as to mp3. It would be interesting to see the comparsion between lame encoded mp3s and lame encoded vorbis files. Then of course there is the comparison of oggenc to lame for just vorbis files. Regardless, an interesting article. Thanks.

Well... (4.00 / 1) (#16)
by regeya on Thu Jan 25, 2001 at 06:51:58 PM EST

I think I mentioned it briefly, but when I tried to build lame with vorbis support, the vorbis support was broken. libvorbis had gone through some fairly extreme changes and lame hasn't caught up yet. I looked at it myself, but frankly I found it to be too mind-boggling ;-)

[ yokelpunk | kuro5hin diary ]
[ Parent ]

Ah; Right you are. (4.00 / 1) (#17)
by thedward on Thu Jan 25, 2001 at 06:55:45 PM EST

Sorry, I missed that part. That makes sense.

[ Parent ]
An irrelevancy but... (4.28 / 7) (#18)
by Vulch on Thu Jan 25, 2001 at 07:01:20 PM EST

Seeing as no-one has mentioned Terry Pratchett yet...

Vorbis, the high priest from his book Small Gods and Ogg from Nanny Ogg, the Lancre Witch, and her extended (in time as well as space) family who feature in several of the series, Wyrd Sisters and Witches Abroad to name but two.

Re: An irrelevancy but... (none / 0) (#47)
by WWWWolf on Sat Jan 27, 2001 at 11:11:38 AM EST

Vorbis, the high priest from his book Small Gods and Ogg from Nanny Ogg, the Lancre Witch, and her extended (in time as well as space) family who feature in several of the series, Wyrd Sisters and Witches Abroad to name but two.

In Ogg/Vorbis name, "Vorbis" indeed refers to the guy in Small Gods, but "Ogg" comes from gamer sense that's described in the Jargon File (something that the RTS players some time ago would probably have referred to as "grunt rush"...)

I think it was in their FAQ, too...

-- Weyfour WWWWolf, a lupine technomancer from the cold north...


[ Parent ]
Ogg will improve over time (4.50 / 8) (#24)
by SEAL on Thu Jan 25, 2001 at 07:32:19 PM EST

First, I think it needs to be stated that Ogg Vorbis is still beta. They are still making improvements, and I'll be watching for more comparison tests after the official release.

Second, because they are still beta, much of Vorbis isn't fully optimized yet. Using Winamp with the vorbis plugin, I compared CPU time used during playback versus mp3 files. On a 450mhz machine, the mp3 files were consistenly low: 5 - 10% CPU. With the Ogg Vorbis files, playback used around 25% CPU. This is something that will improve as the code matures. Reading the vorbis-dev list is encouraging. They have some SERIOUS audio and math geeks tuning this thing.

Finally I wanted to say that this project is really exciting for game developers. Many of us are using mp3 compressed audio in our games to get more speech packed into a CD. But, of course, we pay royalties. When we can use Ogg Vorbis as an alternative, with about the same CPU hit, we'll be all over it.

Best regards,

SEAL

It's only after we've lost everything that we're free to do anything.
Listening to the files (5.00 / 1) (#33)
by Gyles on Fri Jan 26, 2001 at 10:28:30 AM EST

Wouldn't it have been better to, having encoded the files, use the player to convert back to raw WAV files and burn to CD and play back through a hi-fi. There are plenty of problems with using the computer to to the play back (electrical noise, sound card quality, speaker quality etc.).

Legality of Ogg/MP3 collection (4.25 / 4) (#35)
by Sneakums on Fri Jan 26, 2001 at 10:39:29 AM EST

Also, in informal polling, I have found that some people see no reason to switch. The most common answer I get is "even if I switched from using LAME to oggenc, and I made all my rips Ogg Vorbis, my music collection would still be illegal."

As I understand it, your MP3 or Ogg collection is perfectly legal once they were sourced from CDs you own; it falls under the "fair use" provisions of copyright law (at least in the US). It is the distribution of these files that would constitute an illegal act.

As for switching, I did so for two reasons: freedom, and the fact that 128k Oggs sound better to me than 128k LAME-encoded MP3s.

legality (3.75 / 4) (#36)
by Ender Ryan on Fri Jan 26, 2001 at 10:50:33 AM EST

From what I've heard time and time again, Ogg Vorbis will still be illegal because of patents. It is basically impossible to create any kind of audio or video compression without infringing on a patent.

Not only that, but Frauenhaufer (sp? I don't care) and friends plan to file a lawsuit against these guys. Sorry, I don't have a url.

There was an article on slashdot about this recently, again, sorry I don't have the url.

My feeling is that Ogg Vorbis won't survive. But I REALLY hope I'm wrong and would gladly donate money to a legal fund to help these guys out when it's needed, and it will be.


-
Exposing vast conspiracies! Experts at everything even outside our expertise! Liberators of the world from the oppression of the evil USian Empire!

We are Kuro5hin!


FUD (4.33 / 3) (#39)
by jsoderba on Fri Jan 26, 2001 at 12:29:22 PM EST

From what I've heard time and time again, Ogg Vorbis will still be illegal because of patents. It is basically impossible to create any kind of audio or video compression without infringing on a patent.
I can't believe that will hold up in court. (IANAL, tough) Either the patent is to general or there are ways to make a noninfringing codec. According to xiphmont's comments in this /. thread there is no infringement, and Xiph.org will fight a suit in court. There has been no suit yet, so what Thompson said was probably FUD anyway.

[ Parent ]
so? (5.00 / 1) (#55)
by Anonymous 7324 on Tue Jan 30, 2001 at 09:12:24 PM EST

when has the fate of court cases been decided on merit? Fraunhofer has much $$, and isn't afraid to use it to harass the individuals involved until the whole things shuts down.

[ Parent ]
Legal maneuvering (4.00 / 3) (#42)
by regeya on Fri Jan 26, 2001 at 04:34:34 PM EST

Fraunhofer is just (IMHO) trying to protect their assets. They've offered no proof, but instead claimed that, in their opinion, Ogg Vorbis could not have been created without patented MP3 technologies. My bet is that Thompson has seen that the Xiph team is without funding, and is planning to take them to court to bankrupt them (thus, perhaps, eliminating the threat to their "cash cow.")

[ yokelpunk | kuro5hin diary ]
[ Parent ]

Untrue! (5.00 / 2) (#46)
by WWWWolf on Sat Jan 27, 2001 at 11:05:43 AM EST

From what I've heard time and time again, Ogg Vorbis will still be illegal because of patents. It is basically impossible to create any kind of audio or video compression without infringing on a patent.

Not only that, but Frauenhaufer (sp? I don't care) and friends plan to file a lawsuit against these guys. Sorry, I don't have a url.

No, Xiphophorus isn't having any patent issues! Basically, Fraunhofer thought "Oopsy-daisy, those 'free software' types have made a better encoding system than we did, they got to be using our technology 'cuz it's the best, ain't it? Only we (and Microsoft) can 'innovate'..." and sent out some FUD =)

I've heard that the Xiph's legalese-cutting folks have been really busy trying to make sure they're not duplicating someone else's mistakes...

MP3 and Ogg/Vorbis are different approaches to same problem, that's all - they're irrelated and don't use the same algorithms. Vorbis is closer to TwinVQ (or whatever its name was) than MPEG Layer III, but it's different from TwinVQ, too.

-- Weyfour WWWWolf, a lupine technomancer from the cold north...


[ Parent ]
mp3-to-ogg, ogg, my choice (5.00 / 1) (#52)
by leonbrooks on Tue Jan 30, 2001 at 10:15:27 AM EST

1. There is a Xiphorous utility to go straight from MP3 to OggVorbis;

2. Ogg is not from Nanny Ogg, their web site says so. Vorbis is definitely from Small Gods, though. He was eventually taken out by a six-inch shell. (-:

3. I choose OV because

  1. it's free (libre and gratis);
  2. it will get better and is, MP3 won't and isn't;
  3. the later OV will also do bitrates which MP3 will only stare at in confusion - how you say, ``digital telephony?'';
  4. the better OV will produce smaller files and can include things like cueing information (think automated slideshow), MP3 doesn't and won't.

-- If at first you don't succeed, try a shorter bungee
Sorry to do this... (none / 0) (#53)
by regeya on Tue Jan 30, 2001 at 11:13:27 AM EST

...especially since we both seem to have chosen Vorbis. :-)

1. There is a Xiphorous utility to go straight from MP3 to OggVorbis;

What's it called? I find no mention of it on xiph.org and I was under the impression that they wouldn't touch MP3 with a 10-foot pole.

[Ogg Vorbis] will get better and is, MP3 won't and isn't;

Well, I'll certainly agree that Ogg Vorbis will get better, but MP3 is just a format. Have you tried LAME lately? It's got excellent output quality. I'm no math-head (reading just some of the MPEG specs made my head detatch from my neck and spin in circles) but my ears tell me that LAME is damned close to CD quality on 128kbps...and they're still working on it. Sure, Bladeenc still sounds sorta crummy at low bitrates but it's just one encoder.

[ yokelpunk | kuro5hin diary ]
[ Parent ]

MP3 and Ogg Vorbis: A User's Perspective | 54 comments (47 topical, 7 editorial, 0 hidden)
Display: Sort:

kuro5hin.org

[XML]
All trademarks and copyrights on this page are owned by their respective companies. The Rest 2000 - Present Kuro5hin.org Inc.
See our legalese page for copyright policies. Please also read our Privacy Policy.
Kuro5hin.org is powered by Free Software, including Apache, Perl, and Linux, The Scoop Engine that runs this site is freely available, under the terms of the GPL.
Need some help? Email help@kuro5hin.org.
My heart's the long stairs.

Powered by Scoop create account | help/FAQ | mission | links | search | IRC | YOU choose the stories!