COMPARISON: AAC v. MP3 Files. It's the size AND what you do with it. Big.. err long.
Jun 23, 2003 at 4:14 AM Thread Starter Post #1 of 7

blessingx

HeadFest '07 Graphic Designer
Supplier of fine logos! His visions of Head-Fi
Joined
Mar 27, 2003
Posts
13,179
Likes
28
I wanted to do a few "real life" file size comparisons and thought I'd share them with you. Although this should probably be in the Source (Components/Format) forum, it seems most of the compression discussions have occurred in the Portable forum, so I'm posting it here. Also although quality is discussed often, file size is less so, so I want to spend some time with it. I think in terms of albums and as such wanted to see the real size differences in "usable" settings of the two compression formats I can use on my desktop and iPod (MPC, OGG, FLAC, etc. are out for that reason). "Usable" is certainly subjective, but since discovering this site and buying a decent pair of headphones I've had to get rid of a lot of low quality (highly compressed) MP3's. Even AAC 128, which in my opinion sounds very good for 128 kps, isn't quite there (though close). On the other end, I think 320 kps mp3's are excessive and certainly are beyond my hearing/equipment and many think still containing the format "problems" while taking up a lot of space. I'm also not testing overlapping settings (192 ABR) that are better covered by presets (--APS). I'm using what are generally considered the best two encoders for the formats, LAME (which is the only MP3 encorder with these presets) and Quicktime. In this case LAME 3.92 and Quicktime 6.3 (with iTunes). In addition because I'm on a Mac, I used the AACelerator front end for testing Quicktime's higher ("best") setting for 160 kps and 192 kps. iTunes reportedly uses the lower setting of "good". For MP3, I'm using the iTunes-LAME app. This is not the horrid iTunes built-in encoder and its poor VBR support. I tested the "-Y" command to limit the sound to 16kHz- something recommended to save space with portables because of environmental noise. Also I'm using two albums as MP3 VBR file sizes can dramatically change based on the songs complexity. AAC, though technically VBR, performs more like CBR in that a four minute jazz piece is going to take up the same amount of space as four minutes of silence (I've tested). This is primarily a test of space, but I'll make some quality comments later.


Original: Radiohead's Hail To The Thief 56.6 minutes 571.3 megs

* AAC 160 kps 65.5 megs
AAC 160 kps "high" (AACelerator) setting 67.7 megs
MP3 LAME "-alt-preset standard -Y" (each songs average kps was 147-206) 70.0 megs
* MP3 LAME "-alt-preset standard" (each songs average kps was 154-237) 77.8 megs
AAC 192 kps 78.5 megs
AAC 192 kps "high" (AACelerator) setting 78.6 megs
AAC 224 kps 91.5 megs
* MP3 LAME "-alt-preset extreme" (each songs average kps was 181-267) 91.8 megs

Original: Tamas Vasary's Frederic Chopin: Nocturnes 80 minutes 733.7 megs

* AAC 160 kps 84.1 megs
AAC 160 kps "high" (AACelerator) setting 84.3 megs
MP3 LAME "-alt-preset standard -Y" (each songs average kps was 178-186) 94.8 megs
* MP3 LAME "-alt-preset standard" (each songs average kps was 187-193) 98.6 megs
AAC 192 kps 100.7 megs
AAC 192 kps "high" (AACelerator) setting 101.0 megs
* MP3 LAME "-alt-preset extreme" (each songs average kps was 208-222) 111.6 megs
AAC 224 kps 117.4 megs

* My recommendations.

As you can see the album sizes can jump quite a bit, 25-40% depending on album, between the settings.

So what about quality? Well let me start by saying my equipment is a G4 to Xitel HiFi/AN1 to Acoustic Research Mini-RCA to X-Feed MK111 to AR RCA-Mini to Trigger Meta Mint to Grado SR-80's with Senn pads. The weak points are the Xitel and my ears. A few readers may note I was a gunnersmate in the Navy. Slight hearing loss has probably occurred.

Not surprisingly the larger the file, the better the sound. But at some point you do cross a line of diminishing returns. Each format has its soft spot and understandably the newer AAC's aim is at a lower end. AAC doesn't suffer the standard MP3 artifacts at 128-160 kps. It is very crisp. If you're on a Mac and are holding off trying the Apple Music Store because of the relatively low 128 kps, I think you'll be surprised (if though you don't want to spend that much for a lossly file I'm with ya). Seemingly though because of its aim at very good quality at around 128-160, the overall quality increase as you move up the line quickly becomes baby steps. I believe I MAY have been able to tell the difference between 192 and 224 (when A/Bing) with the Chopin, but I wouldn't swear to it. I definitely couldn't with the Radiohead and in fact would have difficulty telling the difference between 160 and 192 in that case (though this is partially a compliment because of just how good the 160 kps sounds). Another problem with its aim is that in trying for analytical clarity at such a small kps (in headphone world think "Studio" like the V6's), there's a slight warmth/musicality missing (when compared to MP3's) that never shows up even at 224. Because of this, I do believe MP3 definitely takes the overall quality lead at preset standard, but is it worth the cost of about 15% file size (over 160 AAC)? That warmth may not seem missing (especially with portable phones or in portable environments) unless A/Bing. Also the higher Quicktime settings achieved by AACelerator makes a very slight overall difference. Again I believe it's there (especially at 160), but doubt it will be noticed if never used. I was unable to tell the difference between preset standard and preset extreme on either of the albums.

Which format would I recommend? I think this is based on how you use your files and what your "standard" is.

If you use your compressed files only with a portable and go back to CD's at home (like I think many, but not all MiniDisc users), 160 AAC is pretty hard to beat. It should give you nearly everything you need when out and about. If though you also use your files with a better home setup (and most likely larger hard drive), the increase in file size would be a minor sacrifice to keep the warmth mentioned above. Also depending on your upgrade path you may find yourself in the position to rerip your CD's, never a fun task.

By "standard" I mean do you start with MP3 as the marketplace default unless AAC gives some overwhelming reason to switch, or do you start with AAC because it's much quicker (at least on a Mac) and better sounding at smaller size unless MP3 gives you an overwhelming reason to encode slower and larger. MP3 playback is nearly everywhere now. I share files with friends whose main player is a MP3 capable DVD player, etc. Plus everyone has their favorite player and I like to use Audion (no AAC playback) some of the time. LAME encoding though is very slow even using the "fast" and "multi" (if dual processor) tags, and you do have to move up the kps a bit to shine. Also AACelerator is a slower process (though not as slow as LAME).

I guess I'm mostly MP3 camp as I've got the hard disk space and like to introduce music to friends without AAC playback options (at least until iTunes is introduced in on the PC). But more importantly at the upper end I think MP3's definitely sound better when A/Bing. Since this is a headphone community whose members can sometimes spend a decent amount of cash for best sound, this seems very important. I can certainly see where AAC is useful (especially in that sweet 128-160 space), but I would call that "specialized" use and for "general" (including home setups), LAME mp3's are going to continue to be my default. In fact I often encode at preset extreme for those future headphone upgrades. That might be wasted space, but there's a security of knowing the files are about as good as MP3's are going to sound (or AAC's at this point- though in the future who knows, look how far MP3 has come). You decide what's best for you, but don't think the older format has necessarily been replaced by a new one.

Hope this has all been useful.
 
Jun 23, 2003 at 12:03 PM Post #2 of 7
Very helpful! You got words for things I may have observed, but could not articulate in written or spoken language. Thanks!
280smile.gif
 
Jun 23, 2003 at 2:45 PM Post #4 of 7
Quote:

Originally posted by austonia
great write-up. I'd like to something similiar with a 2nd person switching the sources behind my back.


I forget how multiple Quicktime players/windows are handled on a PC, but on the Mac with 8-9 (when using CD) simultaneously running players (and clicking on each for its sound to take over) all having the same title/file name can get confusing fast. Although not designed as a blind test, the confusion created something like as I lost track of which was which. Had to use file info to rediscover the file path. You may try something similar if you want a one person "blind" test.
 
Jun 24, 2003 at 9:14 PM Post #6 of 7
I made several requests for Apple to support OGG a year ago (like their agnostic graphic/font file support in OSX), but with their push of AAC, it seems unlikely and as such isn't part of the available formats supported on the iPod (and the reason for this test). I've never heard a claim that OGG beats MP3 for sound quality on the upper end (only in the other direction) so you MIGHT be able to assume that at least if you can afford --aps space, there would be little reason to switch to OGG (though I'd like to see a test also). Like AAC/MP4, WMA, MP3Pro, etc. it seems the target is at a lower end. MPC is the exception, though its support is scarce at this point.

But your iRiver doesn't support OGG either right?
 
Jun 30, 2003 at 7:44 PM Post #7 of 7
I did a bit of testing of quicktime mp4 160k against lame -aps on my computer and on an ipod. Testing on computer was blind with winABX, on ipod was also blind. My little band of testers were unable to reliably distinguish between QT mp4 @ 160 and lame -aps in blind testing. ymmv
 

Users who are viewing this thread

Back
Top