Apple just released iTunes 4.5 for Mac OS X and Windows. Among the new features are WMA importing (Windows only, it seems) and a new codec Apple Lossless. The iPod software has also been updated with support for the new codec.

Apple iTunes, iPod software

Edit:Its AAC encoder seems to be updated.
bit to bit comparison (skipping silence)

Seems they've been listening to their customers - lots of usefull additions!

They're still missing one extremely necessary feature, to clean up the music database. i.e if files are deleted through a file manager, they remain there in the itunes music library. For people for organze their files in directories, and move them for whatever reason, end up getting lots of entires in itunes pointing nowhere.

Hehe, I would be rather pissed if I started my test today. It would have been born outdated.

You can do a "consolidate library". And you can also see files that are not present, they are marked with an icon. But there is no Smart Playlist feature to only see the missing files (maybe now in 4.5)

Make sure if you own an iPod or iPod mini you get Apple's latest firmware update for your player so it will work properly with iTunes 4.5 and Apple's Lossless format:

http://www.apple.com/ipod/download/

So the iPods can now play lossless too? Yet another plus in their favour now it seems. I may just have to get myself one (though I was looking at an iPod mini, and lossless isn't that useful for it).

Has anyone tested the new lossless codec yet? Does it store the files in it's own container? Is it vaguely competitive compressionwise? Compared to FLAC, say?

take a look here for the discussion.

edit: spelling (doh!)

Does anyone know if this new upgrade (to the iPod software) fixes the occasional Nero tagging related issue for VBR files, or do Nero generated .m4a/.mp4 files still occasionally skip and have problems as described in other threads here on HA.

I tried playing two problem files encoded with Nero 2.6.1.9 that would cause the iPod to skip over them with firmware 2.1. The same files play fine with firmware 2.2 so it would appear that the nero issues are fixed according to my somewhat limited tests.

That's great news. Glad to see Apple is listening to customer issues... Anyone else confirm that this has been fixed?

I made a couple of m4a files with the new apple aac codec. The resulting files show VBR activity when played back in foobar2000. I had set the bit rate for 192, and the files would vary in playbavk from 160 range up to 230-240 kbps. So it looks like apple aac is now vbr-and all the fuss here today is about the lossless encoder!:)

I read elsewhere that someone thought the low-pass filter might have changed. To test this I encoded the same song with Nero at 128k and the new iTunes at 128k. I did spectrum plots of the same region on the original and the decoded AAC. The results are a bit strange. The Nero file shows a cliff at 16 kHz as expected. The iTunes file shows a 16 kHz cliff followed by a mound from 17-19 kHz.

http://home.columbus.rr.com/tcristy/freq_compare.jpg

@mdmuir: Very interesting about what you discovered about the apparant VBR encoding in the new iTunes. A new listening test and/or results posted by others will hopefully clear this up.

Also, you are right, this VBR support is not really being talked about yet. Perhaps you are the first one to discover it. I think the new Lossless audio feature is worthwhile as well. I wonder why Apple isn't telling us more of what they changed/which codec they used for both the AAC Encoding and Lossless encoder.

Also I read somewhere today on a news site that Apple upgraded the quality of their MP3 encoder (something about 320k support), but I have yet to confirm that an improvement has been done to lower MP3 bitrates.

All in all it sounds like a very good upgrade. It is sure getting a lot of press coverage and a lot of bandwidth here on HA talking about the new features in this upgrade. If I didn't know better I'd think some of us are like kids let loose in a candy store today

@tcristy: Also very interesting discovery you made. Wonder what's up with that..

Well, that certainly is interesting. I wonder if it's audible? Maybe it's data that's not meant to be heard, like a watermark or something.

ff123

Hmm... and how long until there's a listening test to determine whether there is an actual increase in audio quality? "Improvement" is - especially on Hydrogen Audio! - a subjective term, until properly verified. For some reason, I am still a little suspicious... and I think I'll archive a copy of my iTunes 4.2 installer.

- M.

One week. Hopefully.

No, the codec isn't VBR yet. At least, it's not using their VBR codebase. It's probably more like an ABR.

And yes, lowpass management was changed.

Here are two spectral views from OrdinaryWorld.wav:

iTunes 4.2:
http://pessoal.onda.com.br/rjamorim/OrdinaryWorld-42.png

iTunes 4.5:
http://pessoal.onda.com.br/rjamorim/OrdinaryWorld-45.png

Regards;

Roberto.

Interesting, you see the same 16-17Khz gap followed by more information.

My first thought was that they might be adding that lump of high frequency information to fool those complaining about the low-pass cutoff into hearing more than is really there. This might be a situation like PNS where anything up there is perceived about the same as long as the amplitude is close to the same.

Is that lump of high frequency audible and more importantly, beneficial to the quality?

That will take listening tests with people with very good high frequency hearing. For those who can hear it, it might well be detrimental. Typically, if you can see intermittent black spaces in time in spectral view, it corresponds with what people call "ringing."

bAdDuDeX used to complain about ringing which occured above 16 kHz, so I'm sure somebody somewhere will be able to hear a difference; whether it's better or worse than just a plain 16 kHz cutoff is the $64 question.

ff123

erm, was iTunes 4.2 the last windows version before 4.5?

yes

@ mdmuir:

hi!
there is no improvement in this regard - take a look here:

I think it would be beneficial especially considering I and many others have a 3000+ kb/s conection, that apple consider allowing options in quality of download lowfi/hifi etc...

If I am to pay a dollar + for songs I would prefer a format that is cd quality. That way when I make it into a cd or transcode to a different format for my cd player I don't lose a lot of quality.

I found something interesting when I updated. The music sharing feature is not backwards-compatible with previous versions of iTunes. Anyone who has not yet upgraded is displayed in grey text under the "Shared Music" and selecting them gives a nice message saying "The shared music of 'user' is not compatible with this version of iTunes."

I was trying to figure out what would prompt Apple to do something like this when other upgrades didn't break compatibility, then a thought hit me. What if Apple changed the way it streams the music or encrypts it to break MyTunes compatibility (MyTunes is software that allowed you to download others' music instead of just sharing it. Apple forced the maker to remove it from the web, although it can still be found pretty easily on google). Sure enough, MyTunes no longer works with the latest version.

EDIT: I found a page that details the new encryption for music shares. So iTunes is trying to kill off MyTunes and other applications of the sort. Even funnier, it's already been cracked.

EDIT: I wish Apple had been kind enough to disable conversion from compressed audio to lossless or at least warned that you weren't getting better quality by upmixing.

I've done a bit of AAC encoding with iTunes 4.5 and noticed this "ringing" effect almost immediately. I wasn't sure if it was just me at first, but seeing this post helps to solidify my own findings.

I've posted a sample demonstrating this effect which is particularly evident with the symbols: iTunes 4.5 Ringing Sample

My own personal opinion is that this effect is a detriment - the ringing is too pronounced and for me is hard to listen to. I prefer the MP3 sample in this instance.

Hi 32bitwonder,

Thanks for your input. It seems to fit the idea that including extra high frequency information is not always a good thing.

The ideal way to demonstrate a sound quality regression would be to test on multiple samples (and multiple people) and compare 4.2, 4.5 against the original wav (using ABC/HR). But that may be inconvenient to set up. I'm sure inquiring minds want to know, though.

I hope the iTunes engineers actually conducted listening tests before deciding to add the extra high-freqs, and didn't just assume they would sound better!

ff123

BTW: what setting did you use to encode your sample. It looks like you used something higher than a 128 kbit/s setting. The extra high frequencies are centered around 19 kHz. So I wonder if you're hearing the effects of that or maybe something else? Anyway, methodical regression testing is called for.

I hope so too!

However, if other samples and other listeners can confirm this behaviour, this thread would serve very well to prove two things:

1. Never automatically assume a newer version of a codec is always better. Do some listening tests first.

2. Never use spectrographs to judge the quality of a psychoacoustic codec.

If 32bitwonder's findings are confirmed, this can become a thread we could link to in case any of these issues come up again in the future.

But then again, I hope it's an isolated issue or something else is wrong, because it would be bad to see the iTunes AAC codec getting worse instead of improving...

The AAC file was encoded at 224 kbit/s using iTunes 4.5. The source file was FLAC encoded from EAC and later transcoded into WAV (using foobar2000) in order to encode using iTunes. I felt that at a sufficiently high bitrate (224 kbit/s) there would be less debate regarding what's causing this effect.

Did you compared with previous QT encoder?
I've download the file, encoded it, and then searched a way to use the previous encoder. No way, except by reinstalling again iTunes 4.2 (thing I won't do).

I've tried to perceive ringing with QT 128. I'm using my notebook (AC97), and sound is far from excellence. I can't hear anything wrong. The file you uploaded is encoded at 224 kbps : does it mean that you heard ringing at 224 kbps ?! Nice hearing...

Have you tried mpc --standard ? The HF are not constant too, and ringing is surely audible for people with good hearing.

Anyway, I will test this new encoder on my main computer (with better soundcard and headphone) as soon as I can.
There are some samples or instrument (harpsichord came in mind) where extra-frequency might improve the overall quality (QT 16Khz lowpass is audible with this instrument without difficulty).

The purpose of lossless encoding is to spare bandwith and make download/upload less fastidious for small connection. Please, don't decompress file before uploading. Thanks

I've not compared with the previous QT encoder either as I'm not about to start reinstalling 4.2 either.

It's been suggested to me that the ringing effect may in fact be more pronounced at higher bitrates. In order to help determine if in fact this is the case, I've updated the sample page and removed the WAV's to include the source FLAC and iTunes AAC (.m4a) files at bitrates ranging from 128 - 320 kbp/s respectively. It should also help a bit for those with "bandwidth issues".

The page again is: iTunes 4.5 AAC Ringing Sample

My initial perception now is the effect is most audible at 192-256 kbp/s followed by 160 and then 128 & 320 kbp/s as the least affected.

I've also noticed that if I queue up these samples in winamp 5.03, this effect isn't as noticable as it was in iTunes. Something to do with the decoder perhaps?

Either way, my ears are somewhat tired now (been at it all morning). Time for lunch.

A good way to determine if the extra high freq information is responsible for degradation, without having to re-install iTunes 4.2, is to low-pass filter the encoded sample in something like Cool Edit (or SOX, if you don't have Cool Edit). And then compare the normally encoded file against the low-pass filtered encoded file.

ff123

I'll try to set up a test for you later on today. An ABC/HR comparison would give you an unbiased result of your preferences as far as degredation vs. bitrate goes.

Also, I can upload filtered versions for you to listen to and compare against the others.

ff123

Hello,
I’ve just ripped one CD with EAC, encoded it to ACC (m4a) using Nero, and loaded it to my iPod (firmware 2.2) using i-Tunes 4.5.
All in windows
Until now I believed i-Tunes does not load AAC to ipod in windows, or was I wrong?

Now I will try that again all in i-Tunes, and see if it works.

it works

itchy

iTunes 4.2 encodes:

http://www.u.arizona.edu/~mikeg/128k4.2.m4a
http://www.u.arizona.edu/~mikeg/192k4.2.m4a
http://www.u.arizona.edu/~mikeg/256k4.2.m4a

Decoded with foobar, no replaygain, no dither.

Mind if I ask what what song that is?

Here is a setup to directly compare the iTunes version 4.2 and 4.5 files that 32bitwonder and Mike Giacomelli provided:

http://ff123.net/export/aac_ring.zip

1. unzip to a test directory
2. go to the Bin subdirectory and execute the setup batch file
3. run abchr.exe and load up the config.txt file.

The config file sets you up for a 128 kbit/s comparison, but you can easily change this to compare the 192 kbit/s and 256 kbit/s files.

ff123

My take on this issue is that these high frequencies are being added to please graph watchers, and not to increase quality

Well, it's certainly a weird way of including high frequencies. There is this big dip a bit above 16 kHz and a burst of information afterwards. I mean, can't they just alter their low pass filter from 16 kHz to 18 or something simple like that so you don't get the dip?

Actually, these graphs reminds me of some Xing frequency responses on some site....the burst of info after a dip.

Did a test encode with the From Dusk Till Dawn soundtrack to check how Apple's Lossless implementation did it against WMA9 Lossless, Monkey's Audio 3.99 and FLAC 1.1.0. This soundtrack consists of most old and good rock (Tito & Tarantula, Stevie Ray Vaughan, ZZ Top, The Mavericks, The Blasters and more):

Apple Lossless (iTunes 4.5/Windows) : 324 MB (340 440 684 bytes)
WMA9 Lossless : 315 MB (330 587 120 bytes)
Monkey's Audio 3.99 (Extra High) : 307 MB (322 920 984 bytes)
FLAC 1.1.0 (Level 8) : 322 MB (338 156 561 bytes)
WinRAR 3.30 (Best, Solid Archive) : 350 MB (367 372 419 bytes)
Uncompressed WAV : 494 MB (518 181 672 bytes)

I always use the best compression when going lossless (or creating archives). Too bad you can't select the compression level in WMA9 Lossless or Apple Lossless.

Been sticking with Monkey's Audio since 3.97, and from these results, i think i'll still stick with it (i've also done alot of tests like this on a few Pink Floyd albums and metal albums, and Monkey's Audio is the format that always has best compression).

Here's another ringing test at 128 kbit/s:

http://ff123.net/export/applaud_ring.zip

1. Unzip to a test directory
2. go to the Bin directory and double-click the setup.bat file
3. run abchr.exe and load the config.txt file

ff123

@ ff123: Thanks for the tests, I'll go over the samples in some detail when I get adequate time to run through them.

A side comment on testing samples like this (generally) is, I worry a bit about being able to narrow down or reproduce my initial first impressions. The more I listen, the less evidient it becomes. The one thing that assures me that it may not all be just me is the fact that my initial first impression was that of "...what's with all this ringing!" prior to reading this thread. The fact that it's been proven that this high frequency range exists also helps. Still..it's hard to hold on to first impressions...

@ Mike Giacomelli: Thanks for the 4.2 samples. The song is "Push" from The Cure's 1985 album "The Head on the Door".

Now i'm disappointed
According to my results with the past Roberto's tests the old iTunes AAC was very good with similar hf problems (nor warbling nor ringing artifacts).

It seems apple screwed up on 4.5. I guess I'll end up testing 4.2 then...

iTunes 4.2 is not available anymore on Apple website. I don't think that testing the previous release of iTunes is a good thing. Even if quality is now lower (but we need more than one or two negative report to be sure about this regression).

I still have iTunes 4.2 on my intranet server (just the setup.exe file) and installed on my crashbox. Opened it for the first time today in at least two weeks to buy one song and definitely not upgrading to 4.5. If I actually used their AAC encoder then I might consider it, but it fails to meet my standards as both an encoder and player.

You can try this:

http://a1408.g.akamai.net/7/1408/9955/2003...iTunesSetup.exe

Connecting to a1408.g.akamai.net:80... connected!
HTTP request sent, awaiting response... 200 OK
Length: 19,979,192 [application/octet-stream]

I think that is still 4.2 since 4.5 is 20,450,232 bytes.

Thank you, but I have iTunes 4.2 on my hard drive.
The problem is not to find links for the software. Roberto's test is a comparison between popular encoding solutions, which mean:
- availabiltity of the software
- non-tweaked command line

Most people will download iTunes on Apple website. That's IMHO that version that must be tested.

The bits allocated to the frequencies above 16 kHz have to come from somewhere, so I think it's quite possible that quality below 16 kHz may be degraded.

ff123

I don't even know what my tests are representing anymore.

Some say they should represent each format at it's best, no matter what lenghts you go to obtain it.

Others say they should represent what most people use for their personal encoding.

For instance, if Nero won the AAC test, would iTunes be featured in the Multiformat test because it's obviously much more popular, let alone free?

I think it was mostly out of luck, for me, that at each test I conduced the winner also happaned to be the most popular. Surely saved me from lots of headaches to take decisions.