Each time a new version of iTunes is released, I run a few personal tests. Nothing fancy, and generally nothing worth mentioning here... just throwing a handful of files at the program, to see how it behaves.

One of those tests involves handling of HE-AAC. To encode such files, currently I "Convert to Album Images with Cuesheets or Chapters" in foobar2000 using Nero Digital™ Audio+ 1.0.7.0 (see the actual settings, below) and add album art using neroAacTag, and it was one of those images that I loaded into iTunes 7.3.0.54.

Of course, the individual song titles and track numbers were not displayed (iTunes chose to display the filename in place of a song title), but any tags which remained consistent throughout the album (Artist, Album, Year, Genre and Artwork) were properly shown.

Playback time was shown as being twice the actual length, but the music played at the proper speed. (Pausing the track and then un-pausing caused playback to resume at a point equivalent to half the actual elapsed time. Likewise, using the slider to move to any point in the track would resume playback at a position equivalent to half the distance between that and the beginning of the file.)

iTunes still displayed the track as Kind=AAC Audio File / Bit Rate=64 kbps / Sample Rate=22.050 kHz... but on listening to the track, I was certain I was hearing frequencies higher than an 11.025 kHz cutoff. So I set iTunes conversion to WAV/automatic and decoded the track, which of course decoded as a 22.050 kHz WAV. Then I set iTunes conversion to WAV/44.1 kHz/16bit, and tried again. And again, in spectrographic analysis there was a brickwall cutoff at 11.025 kHz. But I still couldn't shake the impression that I was hearing more than that (and I had already double- and triple-checked to make sure iTunes "Sound Enhancer" feature wasn't surreptitiously turned on!).

So... I began playback again through iTunes, and recorded my soundcard's output. Looked at that in a spectrograph, and lo and behold, there were suddenly frequencies higher than 11.025 kHz showing up. In fact, it almost looked like iTunes was properly decoding the SBR portion of the HE-AAC content, rather than ignoring it as was always done previously.

Now, is Apple seriously working on this and trying to slip it in quietly, until they have the glitches ironed out? That would be a logical development given the advent of the iPhone, but they've been stone silent about any sort of HE-AAC implementation or testing if that's the case. I would really appreciate it if someone else could attempt to replicate what I've described, especially if they happen to work for Nero. (Ivan? Menno? Garf? Help!)

- M.


Here are the neroAacEnc settings I used for this test:

• Encoder: codecs\neroAacEnc.exe
• Extension: m4a
• Parameters: -ignorelength -cbr 61000 -if - -of %d ("61000" is specified to get a consistent 64kbps, since "64000" yields 67kbps)
• Format is: lossy
• Highest BPS mode supported: 32

Here are the neroAacTag settings I used for this test:

• FOR %%f IN (*.m4a) DO neroaactag.exe "%%f" -add-cover:front:"%%~nf.jpg" (with the *.jpg filename being identical to the *.m4a filename, and both files placed in the same directory)

I am not sure if Apple will soon support HE-AAC in iTunes (which may mean a iPod firmware upgrade) but I know they will be supporting HE-AAC encoding and decoding in the next version of iChat included with their next OS. iChat conferences using the iSight will begin using the HE-AAC format. Apple calls their HE-AAC encoder AAC enhanced (or enhanced AAC, I can't remember) but it is still HE-AAC none the less.

Since they are adding HE-AAC support in iChat, I would imagine they would add HE-AAC support in iTunes if one ever recorded the audio portion of their iChat and wanted to play it back in iTunes. Man, all this i discussion has my head spinning.

Apple usually updates their codecs as part of an OS update. Codecs are seen as part of OS X and iTunes is just a frontend using built-in codecs. Smaller codec updates can come with smaller updates but I wouldn't be surprised if a major overhaul like introducing more AAC profiles would come with a major OS update like the upcoming Leopard. A new versions of the codec has been confirmed to come with Leopard, whether it comes with HE-AAC remains a question though.

Personally I would love to have HE-AAC, even if it is just for streaming radio. More and more radio stations are moving to HE-AAC and iTunes is really starting to lack in this department.

By the way, wouldn't LD-AAC make more sense for iChat?

That would make sense... although I'm still hoping some will independently verify what I'm seeing. (Preferably someone who can tell - from a technical standpoint - whether iTunes is actually decoding the SBR component with this release, or adding a pseudo-spectrum of its own to the mix.)

- M.

New iTunes and Quicktime today... Any news?

New version of QuickTime (7.2). Wonder if they made changes to the encoder.

Now using iTunes 7.3.1.3:

Almost identical behavior to what I initially posted. Finally thought of testing a sweep tone encoded as HE-AAC, and it became apparent that iTunes was not yet decoding all of the SBR component, although it appears to either be decoding some of it or possibly adding a slight pseudo-spectrum on its own (on playback only... "decode" still produces a frequency cutoff at the expected 11.025 kHz, regardless of iTunes' decode settings).

My father-in-law gave me his old 3G iPod last night, so I loaded a few weeks worth of HE-AAC albums on it - all encoded using the same settings from my initial post. Playback works (realtime) and is still reproducing some frequencies higher than 11.025 kHz.

Pausing an HE-AAC track on the iPod, putting it to sleep, waking and resuming play picked up where it left off. Didn't expect that, although I should have guessed that the iPod was using some sort of data-length locater rather than basing position off the time stamp. Seeking within a track also seems to be based on a calculated data position, as I was able to scroll to an actual time-location and the iPod would resume play at the appropriate point. Play length is still shown as twice the actual length on the iPod, but that's a minor inconvenience; scrolling past the actual play length of the track simply begins playback of the next sequential track in the playlist.

Again, hoping they improve (or add TRUE) SBR reproduction, but for a 3G iPod this is certainly more than I expected!

- M.

The putting-to-sleep and waking part should have no effect whatsoever, AFAIK it's just like using sleep on a laptop, the code running on it continues from where it left off when you put it to sleep, and the RAM stays powered up.

I can't reproduce. Playing the same file (Nero q0.25) in foobar and iTunes 7.3.1 SpectrumLab shows a clear cutoff at roughly 10kHz for iTunes:

gaekwad2, could you post your SpectrumLab settings so folks can compare?

Q: Is SpectrumLab taking audio directly from the soundcard output while iTunes is playing, or did you use iTunes to first decode to WAV? Did you manually adjust SpectrumLab to 44100 samples per second?

- M.

SpectrumLab takes data from the soundcard. And yes, it was set to 44.1kHz.

Interestingly, iTunes' output was 44.1kHz as well, not 22.05.

Took a little time to learn how to use Spectrum Lab this evening, so I could do proper screen grabs.

First is a reference snapshot of the background noise from the internal soundcard in my laptop.



Second is a shot of iTunes playing Bob Marley's Songs of Freedom box set (taken twelve minutes, forty seconds after the first); this is near the end of the third song on the second disc, titled "Trenchtown Rock (alternative mix)". The gap visible at the 22:57 marker is where the previous song ended. Genre is not shown on the first disc since d1t01-t06 are "Ska" and d1t07-t13 are "Rock Steady," rather than being "Reggae" throughout the album.



iTunes "Sound Enhancer" feature is not used. All settings in Spectrum Lab remained the same for each screenshot.

- M.

Edit Forgot to mention that Artist and Year are not shown, since there are a mixture of tracks by "Robert Marley," "The Wailers," "Bob Marley & the Wailers" and "Bob Marley" spread across his entire career. None of the four discs has a consistent Artist or Year tag throughout the album.

Could you do a comparison with foobar?
And maybe also play a track with a hard ~10kHz lowpass in iTunes to rule out artifacts from processing (by iTunes or your soundcard).

Using linear frequency scale (, smaller fft size, and colour palette shifted towards lower volumes compared to above) I get this:



The red streaks are probably noise caused by clipping, I randomly picked an album for this test which happened to be a (pretty bad) remaster of Phallus Dei by Amon Düül II.

As input I used the monitor mixer of my Audiophile 2496, which doesn't produce any noise when nothing is playing btw. (indicating that your soundcard does do some processing (resampling to 48kHz and back?)).

Edit:
Here's a track that doesn't clip (Can - Sing Swan Song), encoded as -cbr 61000. This time it's iTunes first, followed by roughly the same part of the track played in foobar:

Here is the same track, at approximately the same position, in foobar2000 (and yes, my foobar2000 is configured to look like iTunes...):



By way of comparison, the iTunes screenshot looks more like incremental stair-steps in frequency reproduction, with foobar2000 providing a good deal more spectral continuity.

I'll see about getting some hard-lowpass reference shots later tonight, in both foobar2000 and iTunes.

- M.

Question: I re-ripped "Trenchtown Rock (alternative mix)" from Songs of Freedom disc 2, resampled to 22.050 kHz, and resampled the output file back to 44.1 kHz (both in foobar2000, using the "Ultra" setting). Then I did File > Audio Files > Analyse audio file (without DSP) to ensure I'd gotten a hard-lowpass at 11.025 kHz in the source WAV... but now I can't get Spectrum Lab to read from the soundcard any more. What am I missing? It's got to be something simple, as I didn't do anything else.



Is there a way to turn the spectrum display back on, if it's accidentally been turned off? Now when I start the program, and start a thread, Spectrum Lab only gives me a blank grey screen like the first inch of the spectrum display in the screenshot above. (The remainder of the spectral display shown in this image is the full file analysis of the post-lowpass WAV source.)

- M.

Start/Stop > Start sound thread
(worked for me at least, I never tried the analyze file function before)

No... that's how I was already starting and stopping the sound thread. Something has changed, after I analyzed the file, and now when I tell Spectrum Lab to start a sound thread the spectral display box flickers once but remains a solid grey.

I'll be out most of the day, but will be playing with the program more this evening. If anyone has an idea, please post here or send me a message.

- M.

Coding Technologies has a pack of specially-coded AAC files (in ADTS, MP4 and 3GP formats) that sound noticably different depending on what features the decoder supports.

see:
http://www.hydrogenaudio.org/forums/index....c=56272&hl=
direct link:
http://www.codingtechnologies.com/cgi-bin/developer.pl

... and I wish I'd had those tools last week. It now appears that what I was seeing in the spectrograph was, indeed, soundcard-related noise produced by probably resampling. There were some frequencies being produced, and I was hearing them (which started this whole goose-chase), but iTunes was not producing them or responsible for their appearance.

My iPod is producing a tiny bit of audible sound with an HE-AAC encoded post-11kHz sweeptone, but I am not yet certain whether that is a similar iPod-related soundcard issue or simple aliasing. Going from the iPod to my laptop's line-in (which I had originally done) is unreliable, since the laptop's soundcard is clearly not up to the task.

Still, I hold out hope that Apple will add SBR support in the (relatively) near future. After all, more radio stations are beginning to stream in HE-AAC, and bandwidth for an iPhone on the Edge network is still not as wide as could be desired.

- M.