My wife and I are converting our 1300 CD collection to MP3 and putting it up on a networked HD. It's a long slow process because we buy or borrow 5 or 6 new CD's a day and we only convert about a 10 a day because the MP3 tags often require lots of hand-editing, especially for classical music which is 1/3 of our collection. We've done about 4000 "songs" (tracks) so far.

Here's the problem: iTunes needs to do a "determining gapless playback" for every song it did not personally rip itself, and on our wireless network that's SLOW - about 200 songs an hour. So we want to consider using the iTunes MP3 encoder to avoid that.

The iTunes MP3 encoder has generally gotten panned in tests I've seen, but all the serious tests I've seen were from 2004 or 2005 and used low settings like 128 or 160 kbps/CBR. In my own tests at 192 kbps/VBR/"Highest Quality" I can't hear any difference between the iTunes MP3 encoder and LAME or WMP, but I don't have audiophile gear to test on. Just the same I'm using 224 to have an extra margin - we're trying to compromise between qiuality and file size for our portable devices.

Does anyone know any recent (iTunes 7.x) tests at bitrates >= 192 kbps/VBR comparing MP3 encoders?

PS - Yes it HAS to be MP3. We want to play our music not only on our iPods, but on our cellphones, car stereos, boomboxes, my wife's iRiver, etc, and they don't do AAC.

Nobody can tell you what you will hear if you upgrade your gear in the future.

Almost offtopic, but have you considered FLAC, or some other lossless codec?
This way, you can always create "temporary" encodes of all your music. Like 64kbps (or less) for a phone ringtone, 128 for portable mp3 player/car stereo, 256 for a boombox, and lossless for the home stereo.

At 224k you have fairly secure encodings. You sound like you care about overall robust quality - if so I'd stick with that bitrate. Also at that bitrate all the encoders are more of less equal if you deduct problem samples. You might want to experiment with lame mp3 -V4 setting which is around 160 k vbr and that could save you lots of space if its good enough for you.

Yeah, Apple Lossless would be the way to go here if you've settled on iTunes.

How essential is iTunes to your setup? Are you open to using a different player and rockboxing your iPod (and possibly your iRiver)? If so, LAME becomes a more attractive option.

I'm also wondering if/how there's anyway to store the iTunes gapless info on the file itself. Maybe you could load it initially on a local hard drive, store it, and then reload it on your external HD. It's adds a bit to your process, but it may speed things up in the long term.

Consider the math - Currently 1300 CD's - say an average of 10 tracks per CD - 13,000 "songs" and adding more every day. The sheer labor involved in creating multiple copies of each one would be monumental. And I don't know what sort of tagging structure FLAC has but unless I could perform one tag edit and apply it to both that would multiply the work even more, since, as I said, a lot of the tags need to be hand-edited.

I've used several MP3 players and haven't found one for tiny portable use that I like better than my 8G Nano, and on the other end of the scale, we're buying an 80G iPod to replace our CD cabinet. That way we can have our entire music collection pocket-size wherever we are. 80G is barely enough - the competing 60G products by Creative and Microsoft won't hold enough.

We're in our 50's so our hearing is certainly not going to get better over time, and as I said, I can't hear any deficits even at 196/VBR/"Highest Quality", (I'm listening on my Nano and on my PC with a separate sound card, using Bose QuietComfort II's, and Etymotic ER6i's) so I think we'll be satisfied with the iTunes encoder, but a lot of the tests from 2004 and 2005 really panned it so I just want to make sure we're not making a dreadful mistake.

It's not really that bad converting, you just load all the FLACs into foobar, then convert them to MP3 or AAC. Sure it takes a while, but you'll only do it once or twice a year. Foobar and dBPowerAMP make good use of dual core CPUs so you'll get about 30 - 40 times real time on even a medium level dual core CPU.

First, I agree that Apple Lossless is the way to go. It is well supported now and sports full iTunes tagging; going from the lossless file to iTunes-encoded MP3 is easy with all tags, art and lyrics kept intact. I have largely abandoned FLAC for this reason, and Apple Lossless clearly works just as well.

The original post indicates that determining gapless playback info is slow, but that is largely due to the use of MP3 instead of AAC. Since MP3s cannot store information about "gaplessness" it has to be calculated for each track by iTunes. AAC supports real gapless information and works far better (e.g. faster) for this application.

Too bad iRiver doesn't support AAC, but at least the gapless status need only be calculated once.

LAME MP3 stores gapless information, and iTunes uses this if its available.

Ah, good info. I haven't used LAME in a while, sticking with AAC.

It's not just iRiver - neither of our cellphones plays AAC, neither of of car stereos play AAC, my boombox doesn't play AAC, my PDA doesn't play it - but they all play MP3. AAC just isn't widely-enough supported whereas MP3 is universal.

Also, I disagree that the gapless problem is due to MP3 - it's REALLY due to the fact that iTunes has no reliable way to turn OFF gapless in version 7.x. This problem doesn't exist with MP3's in 6.x or WMP, etc. And just last week one of the Apple lackey's on Apple.com/support announced that they have no plans to fix this. Some people have gotten so fed up with this feature they've UNinstalled 7.x for 6.05. (I can't do that because the 80G iPod requires 7.x).

Out of my 1300 CD's most are classical, jazz, folk, country, and rock, where I don't need gapless. I have maybe 20 CD's of house, techno, trance, or other dance music where it would be nice - so why do I need to sit through iTune's machinations for the other 1290 CD's?


. . . so getting back to my original question -

Are there any good tests of MP3 encoders including iTunes 7.x that look at (er, listen to) 196 or 224 kbps VBR? I saw a test here on hydrogenaudio that looked at 196 cbr, and it wasn't bad.

Or, is it true, as someone here suggested, that at those bitrates for 99% of the music any differences will just be splitting hairs?

I expect you've got slim pickings here. For example, if you were to look at relevant Wikipedia pages you'd find two listening tests linked, both of which are fairly old and both of which you'd have already found here. If there'd been large-scale tests conducted since it's a fair bet someone would have added a link to them to Wikipedia.

SoundExperts, also linked from the Hydrogenaudio knowledgebase, have a 192kbps test, which includes the iTunes MP3 encoder. On the particular sample they tested it was roughly comparable to LAME, but both MP3 encoders trailed Nero AAC, Ogg Vorbis aoTuV, and WMA 9.1. However, at the bottom was iTunes AAC--the usually good iTunes AAC encoder has problems with samples like the particular sample they'd chosen.

http://www.soundexpert.info/coders192.jsp

I don't know how much one can conclude from tests, however many participants were involved, that are of a single, problematic, sample. But I doubt you'll find any more extensive but also recent tests around on the web. But is there are reason to suppose that the iTunes MP3 encoder would not generally be roughly comparable to LAME at such high bitrates?

Just that the iTunes encoder generally came in close to last in several tests in 2004 and 2005 (admitedly at slower bitrates) so I was afraid it might be a generally inferior encoder.

It has also occasionally been alleged, I think even by some posters on hydrogenaudio, that Appple might have deliberately hobbled their MP3 encoder because they have such a vested interest in getting people to use AAC.

And in general, in discussions on forums like this, you get posters asserting that "everyone knows" it's just a generally-understood fact that the iTunes encoder is not very good (though they usually do so without citing references).

I'm an engineer so I prefer to base decisions on good data whenever possible so I was hoping to find some.

BTW - is 196 or 224 considered a "high" bitrate? 128 kbps MP3 CBR - regardless of the encoder - to my ears, sounds gawd-awful for cymbals, high brass notes, and applause, or in complex orchestral passages. Since jazz and classical are the two biggest categories in my library, this is a problem. 128kbps AAC sounds merely awful on the same material. (I do NOT understand how people can pay perfectly good money for the 128 kbps AAC's they sell on iTunes!) I admit that 196 VBR MP3 sounds OK to me, but that's not much higher than the "gawd-awful" setting.

It clearly is. The question you seem to be asking is "inferior enough that I should care?".



Not bobbled, but they've clearly depreciated it. Though maybe thats the same thing given how poor it was when they lost interest in it.

I guess you write about the common 128 kbit/s CBR files found throughout the internet here, which are usually encoded by technically inexperienced people who often use outdated encoders (Blade, old Fraunhofer...) and even go for full stereo mode because they don't trust the joint stereo one. Someone who knew what he was doing wouldn't ever do evil things like these, hence you can't use these encodings as a valid argument to question the quality of a ~128 kbit/s MP3 file. And disregarding the encoder doesn't make any sense at all, due to the quality of a well-tuned encoder drastically differing from a rudimentary one (ever compared LAME/latest Fraunhofer to Blade/Shine?). That would be the same like replacing your speedy Ferrari's engine by an old Fiat Panda's one and then claiming that the Ferrari was a slow piece of scrap.

Following my personal definition 192 kbit/s do prove being a high bitrate, due to noticeably lower average ones already being transparent to my ears. LAME -V 5, which commonly averages closely above 128 kbit/s, is just fine to me. I'm not alone in this case, as you can see in this poll. CBR isn't taken into account here, because in my opinion this restricted mode doesn't provide representative results concerning an encoder which features both a decent VBR algorithm and a good psymodel.


In the above quote you criticize people not citing references when discussing the quality of an encoder, but in the lower one you make a very subjective, non-referenced statement yourself. A good reference to support your opinion about iTunes-AAC @128 kbit/s would be a log of a proper ABX test.

Besides, have you had a look at the latest 48 kbit/s multiformat test? Lots of people obviously had troubles hearing differences between the source and the 96 kbit/s iTunes AAC high anchor, as Sebastian Mares also stated in his summary. And 128 kbit/s are fairly above this anchor's bitrate, though I have to add that I'm not aware of the exact quality of the iTunes portfolio myself, due to not being a client of online music shops. It might be possible that someone actually screwed something up in the iTunes files' case. If it was like that, the results of your ABXing could provide useful information.


What "gawd-awful" setting? The possibly Blade-encoded, 128 kbit/s CBR full stereo one? Even if the same encoder in joint stereo mode was used for both files, the step from 128 kbit/s to ~192 kbit/s would still include a whoppy 50% rise in bitrate and the switch from CBR to the quality-based VBR. I'm very sure a proper listening test would prove that this made a serious difference.

But if it "clearly" is, then since the point of this thread is to find some recent, decent tests, you should be able to point us at some.



There's that word "clear" again, so could you please cite some tests showing that, at 196 or 224 VBR, iTunes 7.x is "clearly" inferior?

Thanks. That's what this whole thread is about.

Has the encoder even been updated since the last listening test? Maybe you should start by determining if the current tests are out of date.



I don't think you understand what you quoted. It says the mp3 encoder was depreciated by Apple, not that its inferior.

I'm just going by what I've heard personally on files downloaded from the iTunes store. They were terrible. Apple has no incentive to do anything but their best possible job in encoding songs they sell on the iTunes stote, since unhappy customers won't come back for more.

There's no way to know, but LAME, Realplayer, and WMP have all updated their encoders since 2004/2005 so there's no reason to assume iTunes is still using the same one they were then. Apple doesn't announce these things. But even then, the tests from that time were all at slower bitrates, with lower quality settings than I use and they were CBR, so they don't shed any light on my question.

Still, since it said "clearly" I think you should point us to the basis of that.


The prior comment already said it's inferior. In any case, if it's "clearly" depreciated then he should provide a basis for that claim since it's not at all clear.


I made my question clear at the start of this thread and so far almost NO ONE has answered it. To summarize: It would be very convenient for me to use the iTunes MP3 encoder to encode my CD's. I'm working at 224 kbps, VBR, "highest quality". At those settings is there any good evidence that I'm giving up any audible quality compared to competing, less convenient workflows?

I did give you an answer. Any half decent mp3 encoder should be transparent at 224k (take a hint: there are no 224k tests for a reason) Take even an eight year old LAME mp3 encoder [v3.20] - at 224k chances are you can't tell the difference. A modern Itunes encoder 'should' do better.

Whether you sacrifice quality is tough question. Given lossy compression + mp3 limitations , the answer might be *maybe*. If you could hear it or not is another matter. My bet is that you can't tell the difference.

It is generally accepted here that LAME is the gold standard for MP3 encoding. There haven't been any recent MP3 encoder comparison tests that I'm aware of. If you look at things from a business standpoint, Apple (iTunes' developers) has more incentive to promote their own AAC format over the MP3 format, despite the fact that their software supports the latter. LAME, on the other hand, is dedicated to MP3. Given that, LAME has greater incentive to improve MP3 than Apple does. It's therefore more likely that LAME is a better encoder...

However

Any claimed quality difference can be validated by proper ABX testing only. ABX testing is subjective, i.e. the results vary from person to person. If one of us were to answer you directly right off the bat, the reply is unlikely to be specifically applicable to you. The only way for you to get a truly, specifically, meaningful answer is to do some testing yourself. See the wiki here for the LAME switches you might want to test against.

Sorry buddy, but when it comes to quality comparisons there are no short and easy answers.

You may also want to consider iTunes' performance as the number of tracks grows. I've used a friends computer with over 24,000 tracks and iTunes literally takes half an hour to shutdown completely, and browsing through the tracks is difficult. However when I load his shared library on my iTunes, it's fine, so if you're going to be using iTunes's sharing to play music, it might not be a problem. Foobar2000 supports much, much larger libraries efficiently, but of course it be very hard to get it as user friendly as iTunes and there's no efficient easy way to share the library with other machines on the network, that I know of.

They have an interest in AAC in that AAC is MPEG 4 audio. It is a better format and intended by the MPEG as the replacement for MP3.

http://en.wikipedia.org/wiki/MP3#Design_limitations

It is also suitable for enhanced podcasts, because of what you're able to put in it, and Apple is interested in podcasting. (In fact, iTunes is by far the best means of browsing and subscribing to podcasts and synching them to a DAP; some media players don't even have RSS support at all.) AAC is more modern and more extendable and generally gives a better result at a lower bitrate: why wouldn't they prefer it?

But as for deliberately sabotaging their own software ... perhaps some posters on "hydrogenaudio" need tinfoil hats. But, I suppose, as Mike says, they're less liable to interested in the MP3 encoder and resources are finite.



It's dependent on so many things. If blind tests indicate anything, they seem to suggest that most samples become transparent to most people at around 128kbps. But would one be listening on a portable device or piped through a good stereo system? What's the listening environment like? And, as you allude to there, what sort of music?

Some current pop and rock music is so FUBB in production--

http://en.wikipedia.org/wiki/Loudness_war

--that it hardly matters what bitrate it's encoded at. Add in that someone may be listening on a portable device through tinny headphones on a noisy bus or train, and they'd be wasting their disk space and money on encoding at a higher bitrate when they're not going to be able to hear the difference.

In any event, I don't think you'll be able to find any recent comparative tests of MP3 encoders at 192kbps or above and I'd think for the reason shadowking gave. I tried Googling and there's precious little. If it's test results you need, then it looks like you'll need to do the tests yourself.

There isn't a listening test to give you the answers that you want.

However, the previous listening tests give some information about which samples / kinds of sounds caused the most problem for those encoders at those bitrates.

Other threads on Hydrogen Audio mention or list audio samples that give problems for Lame (and, rarely, other encoders) at high bitrates.

What I would do if I were you would be to collect those problem samples, encode them using iTunes at my chosen settings, and listen very carefully.


On another subject: If lame ABR at 128kbps sounds awful to you (try it - it's different from CBR) then you have good ears / perception / are picky. It certainly doesn't sound "awful" to me, and sounds surprisingly good around 140kbps for most content.


Finally, if it's this gapless calculation over a network that's causing a problem, can't you run that calculation locally somehow?

Cheers,
David.

Our plan for playing the music is to play it off an 80G iPod hooked to the Aux input of our stereos.

Using two different over-the-ear headsets (Bose QC II and some Sony model I ferget) and a set of Etymotic ER6i earbuds, I've compared the output of my iPod to 3 different PC's and the iPod trounced all of them. In practice we'll be using speakers but our speakers aren't as good as our headphones for fidelity testing.

Also we don't want to leave a PC running 24/7 just to listen to music anytime we want, especially because we like to go to sleep at night listening to music.

When we're done encoding we expect to have in excess of 10,000 tracks, but probably nowhere near 24,000.

I have 50,000+ songs in itunes (external USB drive connected to 5 year old Dell laptop, connected to Benchmark DAC-1, connected to home stereo.) Opening and closing ITUNES takes a few seconds at most.

Also, can't find the link anymore, but there has been some testing done on the DAC (digital analog converter) within the IPOD itself and the testers were quite surprised with the DAC quality. I was happy with the IPOD as a DAC player in my system for several years. The only thing that ultimately drove me from the IPOD as a player in the home stereo was the disk storage limit. I wanted immediate access to everything.

EDIT: Both the laptop and external drive go into sleep mode if I'm not playing something within 15 minutes or so. Thus, I can play a set of songs and when playlist finished, everything goes to sleep shortly thereafter. But if the playlist is 24 hours long, it will play the entire 24 hours. I know that "sleep" is not the same as off, but the monitor and drives shut down, almost no heat comes from the laptop, and we can't hear any fan, etc.

Everything. I have really broad musical tastes. Among the music I have on my little 8G iPod Nano right now are Mozart, Mendelssohn, Miles Davis, Martha and the Vandellas, Joni Mitchel and Moby. Also John Coltrane, Patsy Cline, the Klezmatics, the Clash, Chick Corea, Johnny Cash, François Couperin, Commander Cody, Aaron Copland, Chanticler, and Cibo Matto. Not to mention Bach, Brahms, Bartok, Benny Goodman, Bing Crosby, Bob Wills, Bob Marley, The Battlefield Band, the Bothy Band, Big Joe Turner, Bad Boy Bill, Billy Joel, Billy Idol, Buddy Guy, BB King, Billy Strange, and Bright Sheng.

. . . etc. You get the idea.


You have to do it on every computer you add the music to. Say you encode a song using a non-iTunes encoder and stick the MP3 up on the NAS. If I go to computer A and "add folder to library" I get the determing gapless playback message. Now if I go to computer B and add the same folder to ITS library, I get the same rigamarole. again. We have 4 PC's.

Holy Cow! How long did it take you to rip that many? How did you do it? What musical genre(s) are they, mostly?

Did you have to edit the tags? About a third of my collection is classical music and classical requires EXTENSIVE tag editing because Gracenote and the other online tag info databases are so screwed up, and also because iTunes has no concept of "movements".

How do you start it up again? Do you have any way to control it remotely? One reason we want to use the 80G iPod is so we can have our music wherever we are - in our bedroom if we're going to sleep; in our living area while we eat or cook or read, in my garage or basement via a docking station if I'm working there, etc. We've considered the Roku but that requires a PC up and running (not sleeping, we think) 24/7, and also we had some reservations about the fidelity of its stream.

At the rate our music collection is growing we'll probably exceed 80G in 2 or 3 years but by then they'll probably have something bigger. Still, 50,000 songs - that's amazing!

Holy Cow! How long did it take you to rip that many? How did you do it? What musical genre(s) are they, mostly?

Took me 3 years on and off. I'd simply take small chunks. For example, I'd think, "this evening while I'm reading, watching TV, etc. I'll rip all my beatles CDs. So I had maybe 12 or so of these and that would take me a while. I'm to the point now where I only have to rip new purchases (yes, I still buy CDs). I have about 3,000 CDs. Maybe 10-12K of the songs are from live shows (legally) downloaded from archive.org. These I didn't have to rip. Mosty rock & roll, folk, blues, some jazz, a very little classical and opera.

Did you have to edit the tags? About a third of my collection is classical music and classical requires EXTENSIVE tag editing because Gracenote and the other online tag info databases are so screwed up, and also because iTunes has no concept of "movements".

This was the saving grace. I didn't have to enter most tag data as most loaded automatically. I used mostly the LAME MP3 encoder and freedb as the tag lookup. I didn't have much luck with my classical or opera either on this count. This will add significantly to your time. Although I did play around with most of my tags in terms of artist name (just Bob Dylan or Bob Dylan & the Band; just Jerry Garcia or Jerry Garcia Band, Jerry Garcia & the Black Moutain Boys, etc.), genre, etc.


How do you start it up again? Do you have any way to control it remotely?

No remote. To restart I simply move the mouse or rub finger along mouse pad. This wakes up the laptop and as soon as I click on play in itunes, the drive "wakes up". If I wanted remote access from many part of house I'd check out a sqeezebox or transporter from slim devices. Read about them here. I do have two IPODS and use them. One I travel with and the other I use out on the deck connected to a portable radio's aux input. If I wanted to listen from bed and control, I'd probably just do the same thing in the bedroom with a small stereo and aux input for the IPOD.

http://www.slimdevices.com/


This is some useful reading as well. pay attention to using ITUNES as a player and note the recomendation to make sure your ITUNES volume is set to max and the settings you need within QUICKTIME. This is all not relevant if you're using the IPOD itself.
http://extra.benchmarkmedia.com/wiki/index...k_-_Setup_Guide