Snow Leopard / Quicktime X

[cfmsp]

 

New software announced by Apple today.

 

The bulk of the main built-in apps in Leopard have been re-written in 64-bit for Snow Leopard, the new release.

 

There's a 'rumor' that Quicktime 'X' has been rewritten in 64-bit and that Quicktime will provide auto sample rate changes during playback. If true, presumably these features would extend to all iTunes/Core Audio playback, not just Quicktime.

 

If so - yes, there's still a couple of 'if's involved - the gap between iTunes and Amarra might be considerably smaller soon, especially if 64-bit precision is used in audio processing algorithms.

 

enjoy

 

 

 

 

[The Computer Audiophile]

Anybody have a developer's copy of Snow Leopard?

 

[souptin]

Down at the bottom of this page:

http://www.apple.com/macosx/technology/

is the following small print:

 

"All system applications except DVD Player, Front Row, Grapher, and iTunes have been rewritten in 64-bit."

 

Leaves the way open for third party developers I guess, but not so encouraging for those of us wondering what a full 64-bit re-write would bring to iTunes playback. Of course it's still possible that QuickTime X will help, I've never been 100% sure what it does for iTunes. Always assumed it was for video, not audio, but you know what people say about 'assume'.

 

[cfmsp]

 

Souptin,

 

I guess the real question is whether Core Audio has been re-written in 64-bit.

 

iTunes and Quicktime both utilize CoreAudio functions for audio playback - as I understand it, even Amarra uses some CoreAudio functions (unless their dedicated hardware is also installed).

 

iTunes is the file mgmt software and front end, as well as the Apple storefront for songs, apps, video, etc.

 

Quicktime is the (slimmer-than-iTunes) UI for audio/video playback via CoreAudio and the video playback functions. Many people refer to Quicktime as the playback engine for iTunes, you'll see it here on CA occasionally.

 

There would be no need to rewrite iTunes-as-UI/file-mgmt into 64-bit. Rewriting the core video processing engine in 64-bit would seem to be an obvious thing to do, given the amount of processing required by hi-def video. Indeed, I believe I read that hardware acceleration is included in Quicktime X.

 

The question as I see it, does CoreAudio also get 64 bit treatment.

 

 

 

[Audio_ELF]

Not sure how much it means, but from http://www.tuaw.com/2009/06/08/quicktime-x-leaps-forward-in-snow-leopard/ comes this...

 

"[..]In addition to the visible changes, QuickTime X looks like a fundamental rewrite of the application and its underpinnings. Support for Core Audio, Core Video and Core Animation could mean some really interesting things for the future of media playback (not that we weren't promised as much a few years ago, of course). All of this comes wrapped up in Snow Leopard, and takes full advantage of the speed-tuning tech therein.[..]"

 

Anyway, only a few months to wait till September (unless you can get developer's version to see) and at a reported $29 upgrade cost ($10 if you buy a Mac after 8th June) it's not going to break the bank for people with suitable hardware.

 

cfmsp wrote... "Quicktime is the (slimmer-than-iTunes) UI for audio/video playback via CoreAudio and the video playback functions. Many people refer to Quicktime as the playback engine for iTunes, you'll see it here on CA occasionally."

My understanding is that there are two elements to Quicktime ... the playback engine, which it is my understanding is also used by iTunes, and the Quicktime player which is completely separate application. From reading both these appear to have been updated to 64bit versions along with all of the core functionality of Leopard / Snow Leopard which should include Core Audio I think.

 

Eloise

 

[cfmsp]

 

"My understanding is that there are two elements to Quicktime ... the playback engine, which it is my understanding is also used by iTunes, and the Quicktime player which is completely separate application."

 

Perhaps only developers would (need to) know for sure. I tend to think of CoreAudio functions as the 'playback engine', but that could just be me. Depends on the level of abstraction desired I guess.

 

cheers

 

 

 

[Audio_ELF]

Found this ... http://www.ring3circus.com/rce/how-i-cracked-the-itunes-7-drm-pt-ii/ ... which while it doesn't relate to if iTunes uses the Quicktime engine for audio playback (it's about cracking iTunes DRM) does state that, while running iTunes under a debugger "[.. he] started out with a breakpoint on CreateFileW, to catch the point where iTunes loads in the ‘m4p’ file, waited for kernel32.dll to do its thing, then set a conditional breakpoint on ReadFileExW (the extended version is used because the file is read asynchronously). This landed me immediately in QuickTime’s audio-playback engine. [..]" It would also make sense (to me anyway) if Apple had reused the Quicktime Engine rather than having to write a second playback engine.

 

Anyway just all part of the rich tapestry of working out how iTunes work. Ideally we want the 64bit version of QT playback engine (if it is used by iTunes) to load the file into memory before playing it.

 

Eloise

 

[cfmsp]

"It would also make sense (to me anyway) if Apple had reused the Quicktime Engine rather than having to write a second playback engine."

 

Eloise,

we're in violent agreement, I think!

 

What you're referring to as the Quicktime playback engine, I'm referring to as CoreAudio functions, i.e. the lower level library of routines that would be (re)used by anyone needing access to any part of Apple's audio processing capabilities, not accessing via Quicktime.

 

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

 

For anyone interested in such arcanity, Wikipedia explains that CoreAudio IS the audio playback engine used by Quicktime, having replaced Sound Manager in 2005. There's also a Wikipedia entry for CoreAudio. http://en.wikipedia.org/wiki/CoreAudio

 

Indeed, CoreAudio was implemented to allow anyone needing high quality audio processing for their applications to call an Apple provided capability. From the CoreAudio wikipedia entry: "The second objective reflects a shift in emphasis from developers having to establish their own audio and MIDI protocols in their applications to Apple moving ahead to assume responsibility for these services on the Macintosh platform."

 

cheers,

clay

 

 

[jon]

I think I've read somewhere on these forums that itunes does it own upsampling and that this is better than core audio's. This is why itunes needs to be restarted when sample rates are changed in core audio via audio/midi set up.

 

If Itunes uses quicktime which in turn uses core audio - is it itunes or quicktime doing the upsampling or have I just lost the plot?

 

 

 

 

 

[cfmsp]

 

"If Itunes uses quicktime which in turn uses core audio - is it itunes or quicktime doing the upsampling or have I just lost the plot?"

 

No, Jon, I don't think you've lost the plot.

 

I think you've just wandered into the 'misty' or grey area.

 

I think this is an explanation that has been repeated often enough to become defacto response. Perhaps it originated in the Benchmark wiki?

 

I cannot find the source of different info stored in my brain - perhaps I can recall it later - but I remember reading a post from a relatively authoritative source that there are multiple levels of upsampling available in CoreAudio (which I am referring to here as the lowest level - and most complete set - of functions available to be called by iTunes / QT / Amarra / et al).

 

The point being, depending on one's required response time vis-a-vis required quality, either of several algorithms for performing upsampling could be selected. As someone who's used a pro audio tool for upsampling will likely have experienced - real-time processing and the highest quality upsampling are often mutually exclusive goals. Thus, the multiple levels, depending on needs of the application.

 

My 'guess' as to why CoreAudio is blamed in the 'scenario' you mention above is that - for instances in which it may not be known what level of processing is desired, the default might be the least highest quality, and therefore statements like "if Coreaudio does the upsampling, it's quality is poorer than when iTunes does it". Purely a guess, on my part, by the way.

 

If someone has more insight, please do share.

 

clay

 

 

 

 

[CG]

"As someone who's used a pro audio tool for upsampling will likely have experienced - real-time processing and the highest quality upsampling are often mutually exclusive goals. Thus, the multiple levels, depending on needs of the application."

 

Ever wonder why Macs with faster processors and more memory are generally considered to give better sounding output from a DAC?

 

It may just be better to either do the upsampling not in real time or to let external hardware do the job. A couple people have made some pretty convincing measurements that when no upsampling is used (for example, the settings in Audio Midi Setup are set to 44.1 KHz 16 bit output for a CD), when no equalization or other effects are applied within iTunes, and when the iTunes volume control is set to maximum, the output is what everybody seems to like to call "bit-perfect." That might be good idea to apply.

 

[davidR]

I'm working on getting a copy of it myself. Apparently it's extremely buggy so I'm not hopeful on getting my audio interface to work with it properly. Fingers crossed though.

 

[cfmsp]

 

"That might be good idea to apply"

 

Yes, of course.

 

I'm not aware of many folks intentionally invoking upsampling via Audio Midi. The scenario that was posted about CoreAudio upsampling being worse than iTunes is the result of unintentional upsampling, i.e., wrong settings in Audio Midi, and is surely to be avoided.

 

AFAIK, upsampling proponents amongst audiophiles seem more likely to be using Windows/PCs.

At least most of the discussions I've read were PC-based tools and systems.

 

I can only recommend Wave Editor as the tool of choice for intentional upsampling on OS X, not that I can recommend upsampling as a general rule.

 

Perhaps professional upsampling IS more likely to be performed on a MAC / OS X!?

 

enjoy

 

[Audio_ELF]

Interestingly Gordon recommends letting iTunes up-sample to 88.2 when using his DACs with a MAC. The recently referenced dCS document also noted that up-sampling using iTunes was not bad - though did recommend avoiding it.

 

I think the real big problem is when you change the "target" sample rate in Audio Midi without restarting iTunes as then you get iTunes resampling to the initial value, and CoreAudio resampling to the new value set.

 

Eloise

 

Note: these two statements can co-exist when you consider that the dCS system includes a high end up-samper (to DSD) as part of the full setup.

 

[cfmsp]

 

yes, Eloise, that is correct.

 

This is the 'error' that precipitates the poor quality upsampling.

 

Hopefully - a thing of the past - with Snow Leopard.

 

Oh happy day. Clearly this is a nod to audiophiles who have bent the ear of Apple.

 

 

 

 

 

[Audio_ELF]

As far as I can see ... the best solution would be some form of "exclusive access" mode like Vista WASAPI. And for it to be fully and properly supported in iTunes with a selection for multiple supported sample rates.

 

Eloise

 

[cfmsp]

earlier I said.among other things,

 

"My 'guess' as to why CoreAudio is blamed in the 'scenario' you mention above is that - for instances in which it may not be known what level of processing is desired, the default might be the least highest quality, and therefore statements like "if Coreaudio does the upsampling, it's quality is poorer than when iTunes does it". Purely a guess, on my part, by the way."

 

I've found a post on the Apple core audio list that offers some corroborating info:

 

This was a comment posed to the list (which was replayed verbatim from Benchmark's Elias Gwinn)

> [...] If the user changes CoreAudio's sample-rate in AudioMIDI Setup to something different than what iTunes is locked to, CoreAudio will convert the sample rate of the audio that it is receiving from iTunes. In this case, the audio may be undergoing two levels of sample-rate conversion (once by iTunes and once by CoreAudio). (The SRC in iTunes is of very high quality (virtually inaudible), but the SRC in CoreAudio is horrible and will cause significant distortion.) If the user wants to change the sample rate of CoreAudio, iTunes should be restarted so that it can lock to the correct sample rate.

 

This is the response from an Apple employee:

iTunes uses the AudioConverter API internally but we set the quality to "max" and AUHAL probably uses the default (I don't know). One SRC at max quality followed by one at the default quality is not so great when analyzing sine tone playback.

 

 

 

enjoy

clay

 

 

 

[CG]

"I'm not aware of many folks intentionally invoking upsampling via Audio Midi. The scenario that was posted about CoreAudio upsampling being worse than iTunes is the result of unintentional upsampling, i.e., wrong settings in Audio Midi, and is surely to be avoided."

 

As you and others have posted, iTunes decides what to do by reading the information in Audio Midi Setup> If you set the sample rate to anything other than what the original iTunes track is or for a different bit depth, you get upsampling. Or over sampling, whichever you want to call it.

 

But, this has been been done for years in all DACs and CD players except those of the NOS breed. There's potentially lots of good reasons why you'd want to use upsampling. The issue becomes the type of filter used (in digital terms) and the effects each design has on the resulting waveform. There's some discussion of this here:

 

http://www.ayre.com/pdf/Ayre_MP_White_Paper.pdf

 

http://stereophile.com/cdplayers/meridian_8082808i2_signature_reference_cd_playerpreamplifier/

 

http://www.wolfsonmicro.com/uploads/documents/en/Ultra_High_Performance_DAC_whitepaper.pdf

 

There's also papers available that you have to pay for. Probably loads of other free ones, too.

 

 

[cfmsp]

 

I found info to corroborate my earlier post that there are multiple levels of SRC quality within OS X

 

to wit:

"The quality of the sample rate converter on Leopard has been significantly improved and it is faster. In addition there is a new setting (kAudioConverterSampleRateConverterComplexity_Mastering) which gives a whole higher range of quality settings at greater CPU cost. In other words, the minimum quality in mastering mode is greater than the maximum quality in normal mode."

 

These new levels are in addition to the max setting mentioned earlier as being used when iTunes calls AUConverter.

 

clay

 

 

[cfmsp]

 

"But, this has been been done for years in all DACs and CD players except those of the NOS breed. There's potentially lots of good reasons why you'd want to use upsampling."

 

Understood and agreed, on both counts.

 

And I stand corrected by Eloise - who pointed out that Gordon recommends changing the sample rate IN Audio Midi to 88.2k for RBCD playback.

 

cheers,

clay

 

 

 

[Bob Stern]

Clay, is the "Core Audio discussion forum" you refer to the one referenced in Wikipedia?

http://www.lists.apple.com/mailman/listinfo/coreaudio-api

 

Apple also has a web discussion forum for developers, but it requires a $500/year ADC Select membership. If you purchase that, you can also download a copy of Snow Leopard and all documentation. Apple does not require you to allege you're a developer; they only require your $500!

 

There's also a basic ADC membership that is free and gives you access to all developer documentation for Leopard, but not Snow Leopard.

http://developer.apple.com/products/membership.html

 

[cfmsp]

 

Yes, Bob, that's the list I was referring to.

 

A google search with topic (such as SRC) plus coreaudio will help locate archived threads of interest.

 

THere;s also an archive search as well.

 

thanks for the info on ADC. I guess one could glean some good info from the documentation.

 

$500 for a 'buggy' Snow Leopard, versus waiting until September for the final version for $29. hmmmm....

 

Clay

 

 

 

[CG]

"The quality of the sample rate converter on Leopard has been significantly improved and it is faster. In addition there is a new setting (kAudioConverterSampleRateConverterComplexity_Mastering) which gives a whole higher range of quality settings at greater CPU cost. In other words, the minimum quality in mastering mode is greater than the maximum quality in normal mode."

 

This is good information.

 

The rub is the "greater CPU cost." Lots of guys have reported that going to a faster CPU improves sound quality. Same for more memory. Same for using an SSD for the system drive. Could CoreAudio be placing a "quality governor" on the conversion quality to maintain system latency requirements?

 

I guess the second rub (although rubs aren't always bad) is that at least I don't know what the filter design for "bats" quality rate conversion is. Some designs seem to sound better than others.

 

Maybe the first will be fixed in Snow Leopard by using a second core for the conversion. For those machines that have multiple core processors. Not sure about the second.

 

[cfmsp]

 

I'm not sure either, CG.

 

I know Amarra are allocating some of their processor intensive capabilities to dedicated hardware, specifically their new Model Four (based on the Metric Halo ULN-8), but don't know how much of this is necessary (for highest quality sound) versus a desire to sell their hardware. Initial reports were that Amarra would ONLY be sold as software/hardware combo.

 

I"ve heard nothing but great things about the sound quality of the ULN-8, but have not read any reports of the Model Four in use with Amarra. I"m sure it sounds fabulous, based on how recording engineers are fawning over the ULN-8, though is costs $6k.

 

 

[davidR]

I have a developers copy of Snow Leopard. It is definitely zippier overall from booting up to surfing the web. Safari was very quick, just as fast if not faster than the Firefox Beta 3.5 I'm running on regular Leopard, impressive. Quicktime loads very quickly when playing movies and has a different appearance. The video quality looks very nice. Unfortunately I can't comment on the sound of Itunes or music playback period because the audio drivers for my interface aren't compatible with this OS and new drivers won't be released until after Snow Leopard is launched. I really wish I had an external DAC on hand to do listening comparisons of Leopard 10.5.7 vs. Snow Leopard 10.6 Big bummer...