Mac Mini, iTunes & high resolution audio

[Jud]

+1

 

In my opinion, anyone that buys HR music on a regular basis should own this application.

 

I don't know, Tom, there are at least some times I'd rather not confirm I've been had.

[kumakuma]

I don't know, Tom, there are at least some times I'd rather not confirm I've been had.

 

So true. I've got a number of suspicious releases that sound pretty darn good... and a lot of true HR releases that sound like crap.

[goldsdad]

Well, yes, I was wrong, because I was using Audacity with the default settings. I plead ignorance.

 

As for my surmising, it's no more a supposition than to say that it's iTunes that's doing the resampling. I'm trying to find out, from a developer I know, if there's any way to confirm exactly where and when the resampling is done.

 

Please make sure you or your developer tests the specific software which I said was resampling, i.e. iTunes 10.2.2, and not some newer version that doesn't resample, and that tests are performed under the specific conditions that I said the iTunes resampling would occur.

[kirkmc]

To calm things a bit here, let me suggest to simply buy Musicscope. It does all the necessary analytics and spits out a great looking report.

 

Sure you can get most of the same with other tools but this is all in one place. I use it all the time.

 

https://www.xivero.com/musicscope/

 

Not something I need, but definitely interesting. Thanks for the pointer.

[kirkmc]

I don't understand why you say it makes no sense that iTunes 10.2.2 (and other versions which I no longer have) was able to resample. Audirvana, Decibel, Fidelia, Adobe Audition, Audacity, etc, etc, etc, are all able to resample.

 

But there's no need for iTunes to do it. For apps that offer the option, yes; that is useful, though I suspect that it's still not the apps doing the resampling, at least on Macs. Assuming that they do resample, that means the audio is being sampled twice, correct? Or do these apps simply do the sampling (or tell CoreAudio to do it) regardless of the AudioMIDI Setup settings?

[kirkmc]

Please make sure you or your developer tests the specific software which I said was resampling, i.e. iTunes 10.2.2, and not some newer version that doesn't resample, and that tests are performed under the specific conditions that I said the iTunes resampling would occur.

 

Yes, that's the difficult variable. He needs to have 10.2.2. So you say that anything after 10.2.2. doesn't exhibit this behavior? Because that's already, what, three years old...

[Daudio]

So you say that anything after 10.2.2. doesn't exhibit this behavior? Because that's already, what, three years old...

 

I run 10.6.3. In fact I downgraded to it - for a better interface, and not wanting any of the Cloud related stuff. I don't know about its internal upsampling behavior as I've never done anything like goldsdad has. It might be interesting to try, but, alas, I don't have Audio Hijack.

[goldsdad]

[...] I still hesitate to say that it's iTunes that does the resampling; I don't think iTunes has the code to do that. iTunes simply tells CoreAudio what to do (at least on a Mac).

 

You already said that, and I explained that an app doesn't need to contain its own resampling code, although it can contain its own code. OS X provides resampling and other audio-processing functions that can be used by any app.

 

BTW, those screenshots in steps 5 and 6 above; is it normal that, instead of a more gradual rolloff, it's so sudden in the examples you show? I don't think that's normal; I would expect that for most audio the frequencies decrease much less suddenly. I saw one file where there's nothing, then a big hump around 30KHz.

 

The extremely abrupt filtering at about 24 kHz plus the mirroring at multiples of 24 kHz, is the very evidence that iTunes was downsampling when you consider that the source audio gradually tapers to 65 kHz before meeting the noise floor!

 

The graphs in steps 5 and 6 are normal for audio that has been upsampled from 48 kHz to 96 kHz and 192 kHz. It shows that iTunes first downsampled the 192 kHz source file to 48 kHz because iTunes had been launched while the system output rate was 48 kHz, then Audio Hijack upsampled the 48 kHz audio from iTunes to the current rate of the output device (96 kHz in step 5, 192 kHz in step 6), just as the OS would do before providing audio to that device.

[goldsdad]

But there's no need for iTunes to do it. For apps that offer the option, yes; that is useful, though I suspect that it's still not the apps doing the resampling, at least on Macs.

 

If audio apps don't resample, why do you think they include resamplers such as iZotope, Goodhertz and Sox, to name but a few?

 

Assuming that they do resample, that means the audio is being sampled twice, correct? Or do these apps simply do the sampling (or tell CoreAudio to do it) regardless of the AudioMIDI Setup settings?

 

The audio isn't necessarily being resampled twice. It can be, but normally it won't be. If the file rate doesn't match the app's output rate then there will be resampling to that rate by whatever algorithm the app uses (for example, iZotope), and only if that app output rate doesn't match the output device's rate will there be another resampling and that will be done by the OS. Normally, the app output rate will match the output device rate. Indeed, if the app hogs the device (i.e. has exclusive access to it), then the device rate cannot be changed to anything other than what the app sets it to be, which will be the same rate as the app is outputting.

[goldsdad]

Yes, that's the difficult variable. He needs to have 10.2.2. So you say that anything after 10.2.2. doesn't exhibit this behavior? Because that's already, what, three years old...

 

I did not say that anything after 10.2.2 doesn't exhibit this behaviour.

 

I said I have iTunes 10.2.2 on OS X 10.6.8, and that iTunes exhibits the behaviour of resampling when the audio file does not match whatever rate the output device was set at (as seen in Audio MIDI Setup) at the moment when iTunes was launched, and iTunes persists in using that initial rate as its target rate regardless of any change that may be made to the device rate while iTunes is open. If the device rate is changed while iTunes is open, then the OS will resample the iTunes output to that new device rate.

 

I said I also have iTunes 12.1.2 on OS X 10.9.5, and I discovered that iTunes exhibits different behaviour following your first response to my initial post in the thread, so thanks for initiating that discovery. That iTunes resamples, too, when the file rate does not match the current iTunes target rate, but, crucially, iTunes changes its target rate to match the rate of the output device if the device's rate is changed while iTunes is open. The OS will never need to resample iTunes' output if the iTunes output rate keeps in sync with the output device's rate.

 

iTunes has changed in this regard at some point between between 10.2.2 and 12.1.2, but I have no idea when.

 

I want your friend to test version 10.2.2 because you've been disputing my claims regarding 10.2.2, specifically. You're welcome to try to prove that no version of iTunes resamples, but I believe they all do when the file rate doesn't match the iTunes target rate.

[kirkmc]

 

I want your friend to test version 10.2.2 because you've been disputing my claims regarding 10.2.2, specifically. You're welcome to try to prove that no version of iTunes resamples, but I believe they all do when the file rate doesn't match the iTunes target rate.

 

I'm not disputing any claims. I simply want to reproduce this, so I can understand it. My initial intention was to write about it, but if it's only affecting older versions, then it's not very timely.

 

Just one thing to point out: regardless of my Audacity settings, only some of the tracks I analyzed in Audacity showed the cutoff at low frequency rates; many showed the correct curves up to about 48 KHz (for 24/96 files). So changing that setting only affected a handful of files. If that setting was wrong, then why did other files show the correct frequency curves?

[goldsdad]

I'm not disputing any claims. I simply want to reproduce this, so I can understand it. My initial intention was to write about it, but if it's only affecting older versions, then it's not very timely.

 

That's a tad disingenuous! You absolutely told me several times that I was wrong. But let's put that in the past, eh?

 

I thought I made clear, in my first response to you (post #15), that the dual-resampling (in certain circumstances) appears not to be an issue with modern iTunes, or at least not 12.1.2. [in that post, I wrongly wrote "no iTunes resampling" at the end of the first paragraph, where I should have written "only iTunes resampling".] I'd have left matters at that, but you continued to tell me I was wrong and my tools were faulty, hence all the wasted time and effort that followed.

 

Just one thing to point out: regardless of my Audacity settings, only some of the tracks I analyzed in Audacity showed the cutoff at low frequency rates; many showed the correct curves up to about 48 KHz (for 24/96 files). So changing that setting only affected a handful of files. If that setting was wrong, then why did other files show the correct frequency curves?

 

The curve is displayed for frequencies whose level is above the minimum level that the graph can display (which is controlled in preferences, as you now know). If the levels for all frequencies are above that minimum level, then the curve will be displayed across the full width of the graph.

[kirkmc]

 

I thought I made clear, in my first response to you (post #15), that the dual-resampling (in certain circumstances) appears not to be an issue with modern iTunes, or at least not 12.1.2.

 

There's still one thing I'm trying to understand. Is there dual resampling, or is it simply that iTunes is overriding the settings in AudioMIDI Setup? In my mind, that makes a big difference. I don't see why any audio should be sampled once, then sampled again.

 

Essentially, it looks like this: there was a bug in iTunes than prevented the correct sample rate from being used. Was that because of iTunes performing what would be two sampling operations on audio (in other words, routing it through CoreAudio twice), or simply because iTunes' settings affected the way CoreAudio sampled the files? They way you have described it suggests the former, but I think it's more likely that it's the latter.

 

And regarding the Audacity settings, isn't it because the volume of certain files was low, not frequencies? Because, as I said, many files I have do show the full frequency range; the ones that didn't turn out, now that I look at them, to be quite low volume.

[goldsdad]

There's still one thing I'm trying to understand. Is there dual resampling, or is it simply that iTunes is overriding the settings in AudioMIDI Setup? In my mind, that makes a big difference. I don't see why any audio should be sampled once, then sampled again.

Essentially, it looks like this: there was a bug in iTunes than prevented the correct sample rate from being used. Was that because of iTunes performing what would be two sampling operations on audio (in other words, routing it through CoreAudio twice), or simply because iTunes' settings affected the way CoreAudio sampled the files? They way you have described it suggests the former, but I think it's more likely that it's the latter.

 

I'm tired. If you come up with actual evidence on which you can base your conjecture, I'll be extremely interested, and will be happy to discuss further.

 

And regarding the Audacity settings, isn't it because the volume of certain files was low, not frequencies? Because, as I said, many files I have do show the full frequency range; the ones that didn't turn out, now that I look at them, to be quite low volume.

 

The volume is low when the frequencies have low levels!!!

[Jud]

And regarding the Audacity settings, isn't it because the volume of certain files was low, not frequencies? Because, as I said, many files I have do show the full frequency range; the ones that didn't turn out, now that I look at them, to be quite low volume.

 

That's what Owen said. Suggest reading carefully for understanding rather than while trying to think of a reply.

[kirkmc]

Some backstory as to why I'm trying to understand the whole sample rate thing...

 

A few years ago, I was asked by Macworld to review one of these so-called audiophile audio player apps, one of those that many people on this forum use. I tried it out for a while and found no difference between its playback and that of iTunes. I had a call with the lead developer of the app, and in talking with him for 30 minutes, he was unable to explain why it should sound better, and what it did differently from iTunes.

 

The only thing that was clear was that it was able to change the sample rate on the fly, which, because of the AudioMidi Setup settings, iTunes didn't do. Other than that, and other than the pompous marketing claims of the app, the developer was not able to tell me what made his app different from any other audio app.

 

I don't believe that things like "hog mode" or whatever each app calls it, or the loading of tracks into RAM, make any difference in playback. Again, the developer was unable to even explain, other than in very amorphous terms, why he thought those sorts of features made a difference.

 

This was several years ago. I looked at the claims of a couple of this type of app today, and it looks as though they do their own audio processing now (which certainly wasn't the case back then; I'm pretty sure the app I was looking at was simply using CoreAudio0). So there could be more advantages now.

 

But I'm still curious about the sample rate thing. I don't believe that an app routing audio through core audio would do it twice; it would make no sense.

[Jud]

I've only checked in on this thread occasionally, so I may be missing something. What I think I see are repeated posts along the lines of "Due to my preconceived ideas about such and such, this makes no sense to me."

 

I should think the solution would be obvious.

 

OK, a little good natured teasing. But probably what I suggested before about reading for understanding rather than in order to reply (or fit it into some abstract idea of what "makes sense") might help.

[kirkmc]

That's what Owen said. Suggest reading carefully for understanding rather than while trying to think of a reply.

 

Actually, what I missed there was the use of the term "levels" to mean volume. I was assuming it mean frequency levels. It's one of those terms that can be applied to different things, I am more used to volume (or gain) rather than the less clear levels.

[jhwalker]

But I'm still curious about the sample rate thing. I don't believe that an app routing audio through core audio would do it twice; it would make no sense.

 

Have you ever heard of a bug?

 

This double processing was discussed thoroughly a couple years back and it was proven at the time iTunes was resampling to whatever had been set *when iTunes started*, then Core Audio was resampling to whatever rate was set in Audio MIDI *currently* - obviously, an oversight / bug by Apple. It has apparently since been corrected.

 

This is an example of where your preconceived notion ("doesn't make sense") is blinding you to proof of a real-world happening.

[kirkmc]

Since it doesn't make sense, I'm looking for proof. Do you have a link to something that gives evidence for what you're saying? Not the fact that the sample rate is different, but the fact that it's iTunes resampling, as opposed to simply overriding the AudioMIDI Setup settings? Do you see what I'm after; it's the difference between the result and the cause. I'm interested in confirming what the cause was. And my "preconceived notion" is not blinding me to proof, because you've shown a result, not a cause.

 

Unfortunately, Apple's bug reporter doesn't let you search for bugs filed by others (I have a developer account), and their developer forums have nothing about this issue.

[jhwalker]

Since it doesn't make sense, I'm looking for proof. Do you have a link to something that gives evidence for what you're saying? Not the fact that the sample rate is different, but the fact that it's iTunes resampling, as opposed to simply overriding the AudioMIDI Setup settings? Do you see what I'm after; it's the difference between the result and the cause. I'm interested in confirming what the cause was. And my "preconceived notion" is not blinding me to proof, because you've shown a result, not a cause.

 

Unfortunately, Apple's bug reporter doesn't let you search for bugs filed by others (I have a developer account), and their developer forums have nothing about this issue.

 

The effect is the same, in either case - the music was being resampled twice: once at the rate iTunes saw and stored when it started up, then again to the current rate set in Audio MIDI. Whether iTunes itself was doing the resampling or there was a bug in Audio MIDI that first accepted the inbound source from iTunes and resampled to the iTunes rate, then resampled *again* before output, it was still an obvious bug.

[kirkmc]

The effect is the same, in either case - the music was being resampled twice: once at the rate iTunes saw and stored when it started up, then again to the current rate set in Audio MIDI. Whether iTunes itself was doing the resampling or there was a bug in Audio MIDI that first accepted the inbound source from iTunes and resampled to the iTunes rate, then resampled *again* before output, it was still an obvious bug.

 

Right, but the curious me would like to know. As far as I know, neither do any resampling; they only tell CoreAudio what to do (I know AudioMIDI Setup is just a settings app). And I'm pretty sure iTunes doesn't do any processing either. In any case, I'm surprised that I can't find much mention of this problem anywhere, except in this forum. I don't contest that it happened, but it seems that not many people talked about it. I would have expected to see a mention of it in Apple's developer forums at least.

[Jud]

Right, but the curious me would like to know. As far as I know, neither do any resampling; they only tell CoreAudio what to do (I know AudioMIDI Setup is just a settings app). And I'm pretty sure iTunes doesn't do any processing either. In any case, I'm surprised that I can't find much mention of this problem anywhere, except in this forum. I don't contest that it happened, but it seems that not many people talked about it. I would have expected to see a mention of it in Apple's developer forums at least.

 

Of course all applications are going to work through the facilities the OS provides. No one says "Photoshop told such and such function in Windows/OS X to apply this change to a PNG file," they say "I applied such and such change to a PNG file with Photoshop." In that same sense, iTunes resamples [using the low level audio functions built into OS X].

[kirkmc]

Of course all applications are going to work through the facilities the OS provides. No one says "Photoshop told such and such function in Windows/OS X to apply this change to a PNG file," they say "I applied such and such change to a PNG file with Photoshop." In that same sense, iTunes resamples [using the low level audio functions built into OS X].

 

I don't think that's correct. Photoshop alters files. The comparison would be the way files are displayed, which uses the system graphics framework.

[Jud]

I don't think that's correct. Photoshop alters files. The comparison would be the way files are displayed, which uses the system graphics framework.

 

Oh dear. Yes, not a perfect analogy. But the point is, no one talks about the interaction between an app and OS lower level functions as somehow separate and divorced from the function of the app. Of course if you want to hang pages worth of arguing with various people in this thread on that fine semantic distinction, feel free.