Ogg vs. MP3, anyone doing A/B testing?
Apr 8, 2002 at 9:47 PM Thread Starter Post #1 of 12

phidauex

Sees your every move. Eats your potatoes.
Joined
Mar 19, 2002
Posts
203
Likes
10
Well, I've been reading a lot about Ogg Vorbis, and have heard good things about its sound quality.

For those of you who haven't heard about it, Ogg Vorbis is an Open Source audio compression system (read: free, unlike mp3s, which are licensed. If you've ever encoded an MP3, and not paid a royalty, then you are technically a thief), it is intended to compete with, and be better than, WMA, MP3, MP3pro and the like. The sound quality is supposedly better than any of the above, and the file size can be smaller. It uses new generation algorithms and technology to further compress audio, while retaining high sound quality. Winamp supports Ogg Vorbis (both decoding and encoding, though you may have to download a plugin), as well as many other players. The encoder software is still in RC3 beta, but now fully conforms to the Ogg Vorbis 1.0 specifications (earlier versions worked, but didn't fully conform to specs, so the sound quality was pretty bad, they were basically intended as 'proof of concept' encoders, and developer tools). The format is upgradeable, and so as the technology gets better, sound quality will go up, file size will go down, and encoding will get faster through optimization. All upgrades will be fully backward compatible.

This all sounds right as rain, however, most of what I've heard when it comes to sound quality basically amounts too "It sounded better." or "It was more... um.. Rich sounding.." Basically reviews by people who don't really know what to listen for, and how to describe what they are listening too. I've read a few things that suggest that Ogg has fundementally superior technology, and can produce higher sound quality, but I'd like to do some testing, and also hear some reviews by other people, not just people listening through their little computer speakers or their Sony V700DJ headphones, but people who can really get into the sound with high quality phones and amps and whatnot, and who are used to listening and evaluating what they hear.

I'm going to go home and encode some Ogg Vorbis stuff, and some MP3 stuff, and do some back to back testing.

But if anyone else has been doing this sort of testing, or would like to do some, please do. I've always been a big supporter of the Open Source movement, and would love to support an open source, higher quality compression format. I love my mp3s, but if ogg can sound better at similar file sizes, then I'll switch.

A few notes, when talking about the 'quality' of Ogg Vorbis files, please use the 'quality' level, IE, "Quality 2" or "Quality 4.97". The bitrate is inherently variable, and isn't a good measure of the quality of .ogg files. Also, other things besides bitrate change with the 'quality number', such as cross channel encoding and such. As a reference, "Quality 2" is supposed to be roughly equivalent to 128kbs mp3s. Most people can't distinguish Quality 3 or 4 from their CDs. "Quality 7" and above are said to be so close that only the best of the best golden eared folks can distinguish it from the source CD.

Also note that there is a file size spike at "Quality 5" and above. This is because at 5 and above, Ogg doesn't perform the cross channel coupling. I'm going to focus on qualities below 5, because file size IS important to me, because I plan on using these files on portable players (Such as my SlimX, once the firmware supports it (which is coming soon!)).

So anyway, I just wanted to get some people talking about this, and see if anyone is interested in performing, or has already performed, some true A/B testing with Ogg.

For your reference, visit www.vorbis.com for more information on the Ogg Vorbis format.

Peace,
Phidauex
 
Apr 9, 2002 at 2:37 AM Post #3 of 12

phidauex

Sees your every move. Eats your potatoes.
Joined
Mar 19, 2002
Posts
203
Likes
10
I haven't given them .mp+ a try yet, but I should. One problem with it, which is part of the problem with .mp3, is that its licensed, and royalties must be paid to the company that invented the codecs. If you've ever encoded an mp3 file and you didn't pay a royalty to Fraunhofer, you are technically a thief. Its never really come into play until now. Fraunhofer (along with evil RIAA) are now trying to force internet radio sites to pay back royalties for encoding and streaming. In some cases, these back royalties are exceeding a 1/4 of a million dollars! So internet radio sites are being shut down, the next target will be end users. Its time to develop an open-source compression format, that will be free of the legal shackles that MPEG is currently suffering from.

Anyway, not to preach, but let me just say I'd feel more comfortable with an open source solution, and I think ogg is a good candidate.

On to practical stuff. Ogg is currently encoding right now. I used CDex to rip and encode a few songs, specifically, the first three tracks from Faithless's "Outrospective" album. It has good bass, good highs, and plenty of male and female vocals.

The system I'm using isn't phenomenal, but hopefully it will be good enough to expose the differences in the encoding.

PII400MHz
400Mb Ram
Teac SCSI 40X CD-ROM
Soundblaster AWE64 Gold (yeah, not the greatest) :p
Cmoy amp w/ Linkwitz crossfeed
HD590 cans
CDex ripper/encoder (I've got EAC, but it doesn't do .ogg (yet!) so I thought it would be more fair to use the same ripping program, so only the encoding codec is different)

I'm using LAME codec (latest version), with VBR, 160kbs min, 192kbs max.

The file sizes are 5.60Mb, 8.05Mb and 11.37Mb for the three songs.

The first Ogg rip at Quality 2 resulted in files sized 2.86Mb, 4.12Mb and 5.75Mb. These are about half the size of the .mp3 files. For what its worth, the Ogg encoding was about 50% slower than the Lame encoding. How do they compare in quality?

Well, in this case, the Ogg files are clearly inferior to the LAME files. The highs sound a bit 'robotic', and everything is just a bit muddy. However, they are still half the size of the LAME encoded mp3s. This quality level would probably be quite dandy for Joe Consumer. The bitrate is displaying between 80kbs and 100kbs during play. For being half the bitrate, they certainly sound more than half as good.

The next round of encoding gives us Ogg files encoded at Quality 4. Encoding has slowed down quite a bit, its encoding at about 1X, instead of the mp3s, which were encoding at about 3X
frown.gif
However, this is one of the things they say will be changing soon, as the code becomes finalized, and then optimized.

The file sizes for the Quality 4 .ogg files are 3.59Mb, 5.11Mb, and 7.24Mb

The bitrate varies between about 115kbs, and 140kbs. At this level, I have a very hard time telling the difference between the .ogg and the .mp3 files. In fact, the .ogg seems to have a tighter bass, while the .mp3 still sounds a bit muddy.

The .ogg file isn't quite what I would call 'CD-Quality' at this point, but its very good. I'd say its on about equal footing with the MP3 file. And its file size is still about 65% the size of the MP3s. Not too shabby.

Lets kick it up to Quality 5. Cross Channel coupling is no longer used at Quality 5 and above, this causes a jump in file size, and also a jump in quality. At Quality 5, the speed has also increased a bit, probably due to the lack of cross channel coupling. Its running at almost 2X.

File sizes at Quality 5 are now 4.54Mb, 6.52Mb and 9.22Mb
This is a bit of a jump from the Quality 4, but its still less than the .mp3s! Bitrates are showing up around 150-170kbs. At this point, I think the .ogg is better than the .mp3. The highs are clearer, the soundstage feels wider (yes, I tried it with and without the crossfeed). There are no artifacts, and the noise floor is unchanged from the original.

One last extraction at 4.97 Quality, just to see what the size jump is, and if there is any noticeable soundstage change with and without cross channel coupling. I might not notice anything, because I'm used to a bit of crossfeed as it stands (and in fact, prefer crossfeed).

There is quite a size jump from 4.97 to 5! The file sizes I've come out with here are 3.77Mb, 5.43Mb and 7.82Mb. 4.97 is fully 20% smaller than 5. And I can't tell a difference. Back and forth, A to B, I can't differenciate. The lack of cross channel coupling in the 5 quality probably makes some difference, but with my equipment and ears, I can't tell for the life of me.

So anyway, in conclusion: I quite like Ogg Vorbis! Quality 2 is too low for me, I can tell whats going on. Quality 4 is still distinguishable from the CD, but barely. Quality 5 is practically indistinguishable from the CD, given my hardware. I'm sure that people with better systems will continue to hear differences all the way up the spectrum, given the fundemental nature of a lossy compression format, but I've reached the peak of my gear/ears.

The .ogg format at Quality 4 exceeded the quality of the 160kbs/192kbs VBR Lame MP3. At level 4.97, I feel I have a very good balance between file size (still smaller than the MP3s!) and sound quality. Also, I have the ethical benefit of using an open source format. As soon as iRiver releases the firmware upgrade supporting Ogg, I'll be switching my prefered encoding style.

However, just a note on various 'transcoding' programs out there. There is software out there to convert .mp3s to .ogg. The process involves decoding the .mp3, and then re-encoding as .ogg. While this seems good, its not advised. Ogg removes different aspects of the audio data than mp3 does, and so once its been encoded as an mp3, if you reencode as ogg, then you'll end up with a file that is missing everything that mp3 takes out AND everything that ogg takes out. So you have a doubly poor file.
frown.gif


Also, if you do switch to Ogg encoding, remember that 'bitrate peeling' is supported, which means that files can be lowered in bitrate without additional penalty. A file encoded at quality 7, and then peeled down to quality 3, will sound just as good as a file encoded initially at 3. This means you could keep a higher quality version on you computer, and then peel the files down to a lower bitrate to fit more on your portable player.

Also remember that while Ogg is currently in RC3 beta, encoders now fully conform to the Ogg Vorbis 1.0 specifications. The format is inherently modular, so as upgrades are made, the sound quality will continue to improve, and the file size will go down. The system will be fully backward compatible as well, so don't worry that today's Ogg 1.0 encoded files won't work on tommorow's Ogg 3.0 decoders.

Don't take my word for it though, I encourage other people to try it too. I'm not super experienced, and I'm sure others can add a lot more to my simple tests.

Peace,
Phidauex
 
Apr 12, 2002 at 2:44 AM Post #4 of 12

The Quality Guru

Blah! he says.
Joined
Jan 27, 2002
Posts
887
Likes
10
I used to be very into this, and this is the order of sound quality, which is FACT, not opinion.

1=highest quality codec, 5=Worst codec relative to those mentioned

1. MPC/MP+ (Mpeg Plus) - Not in development, currently the BEST
2. Ogg - soon might beat MPC in sound quality;it's in development
3. AAC - Good, but has a few minor flaws; inferior to Ogg and MPC
4. Mp3 - MANY flaws, unacceptable by audiophile standards
5. WMA - A horrible codec; unacceptable by audiophile standards
 
Apr 12, 2002 at 9:23 PM Post #6 of 12

phidauex

Sees your every move. Eats your potatoes.
Joined
Mar 19, 2002
Posts
203
Likes
10
By what measure did you determine the quality? Isn't sound quality a subjective measurement by its very definition? I agree with your findings, but I wouldn't claim them to be FACT, because of the inherent nature of what it means to have 'quality sound'.

Mpeg Plus is pretty nice, and I like the technology. BUT, it still has the glaring issue of closed-source technology. I'm not keen on RIAAs plans to shut down internet radio, and they are using the royalty issue as a strongarm technique. Ogg is free from these restrictions.

With ogg being the quality it is and still in RC3 (but meets full 1.0 specs), it has a good chance to overtake mp+, we'll see though, eh?
wink.gif


Peace,
phidauex
 
Apr 13, 2002 at 8:54 AM Post #8 of 12

Joe Bloggs

Member of the Trade: HiBy / EFO technologies Co
His Porta Corda walked the Green Mile
Joined
Oct 21, 2001
Posts
11,240
Likes
3,013
Location
Hong Kong and Melbourne
Quote:

Originally posted by phidauex
By what measure did you determine the quality? Isn't sound quality a subjective measurement by its very definition? I agree with your findings, but I wouldn't claim them to be FACT, because of the inherent nature of what it means to have 'quality sound'.

Mpeg Plus is pretty nice, and I like the technology. BUT, it still has the glaring issue of closed-source technology. I'm not keen on RIAAs plans to shut down internet radio, and they are using the royalty issue as a strongarm technique. Ogg is free from these restrictions.

With ogg being the quality it is and still in RC3 (but meets full 1.0 specs), it has a good chance to overtake mp+, we'll see though, eh?
wink.gif


Peace,
phidauex


Don't think it matters really... using Ogg instead of mp3 or mp+ doesn't mean it gets any harder for RIAA to crack down on your copied music does it?
frown.gif
 
Apr 14, 2002 at 2:30 AM Post #9 of 12

pigmode

Truck-Fi
Joined
Jun 26, 2001
Posts
5,372
Likes
97
I have a bunch of music on my hdd encoded to mpc. It's used occassionally for background music through a Videologic Crossfire. I tried Ogg briefly when I had some time on my hands, and it was good enough for my needs. There are no plans to do any critical listening.
 
Apr 18, 2002 at 2:28 AM Post #10 of 12

Mr.PD

Lives to Take It Outside.
Joined
Mar 14, 2002
Posts
6,581
Likes
13
I just downloaded ogg winamp plugin and the winLAME encoder with ogg. Right now I am listening to Beauty by Motley Crue encoded with ogg set at 9.1 and comparing it to the wav file and an mp3 file encoded at extreme. The ogg file is 12.5MB at 465.4kbs, the wav file is 38.0 MB and the mp3 is 7.30MB at 271.2kbs.
I admit that this is not an ideal song to test with. But it does have a lot of bass in it. I am having a hard time telling one file from the other two. I wish I had that better sound card I keep putting off. So, this is not a good test of audio quality. File size is interesting.
I will do more testing when I have time.
 
May 2, 2002 at 6:45 PM Post #11 of 12

edog

100+ Head-Fier
Joined
Oct 21, 2001
Posts
112
Likes
11
You say that MP3 has many flaws. If you use the highest-quality LAME encoding (--alt-preset extreme), can you hear those flaws in a portable listening environment (i.e. away from your home setup)?

Also, what specific quality flaws does MP3 have? What bugs you the most?

Finally, how does MPC compare to Sony's ATRAC Type-R, which is primarily used in their decks and seems to be highly regarded by many audiophiles?

Quote:

Originally posted by The Quality Guru
I used to be very into this, and this is the order of sound quality, which is FACT, not opinion.

1=highest quality codec, 5=Worst codec relative to those mentioned

1. MPC/MP+ (Mpeg Plus) - Not in development, currently the BEST
2. Ogg - soon might beat MPC in sound quality;it's in development
3. AAC - Good, but has a few minor flaws; inferior to Ogg and MPC
4. Mp3 - MANY flaws, unacceptable by audiophile standards
5. WMA - A horrible codec; unacceptable by audiophile standards


 
May 3, 2002 at 9:07 AM Post #12 of 12

gavinbirss

Head-Fier
Joined
Oct 19, 2001
Posts
83
Likes
11
Hello,

I found with my own evalutation of Ogg Vorbis that only V 0.4 at "-m 6" quality setting : 350 K ABR was the most true to the original file.

I would recommend this version above any other.

The compression ratio is 1:4 however, and if file sizes is not a issue but quality is you may also try FLAC. (1:2.1)

Gavin
 

Users who are viewing this thread

Top