Re: Quality of CoreAudio SRC

On Jan 31, 2008, at 11:54 AM, Brian Willoughby wrote:

Hi folks,

I just read a disturbing article in Stereophile. Elias Gwinn of Benchmark Audio claims that, while the SRC in iTunes is of very high quality, the SRC in CoreAudio is horrible and will cause significant distortion. This goes against everything I thought I knew about the SRC in CoreAudio. Is there a mistake here? Elias claims to have consulted Apple about this. Is it possible that the Benchmark Audio driver for their USB hardware is using an inferior SRC, and they are blaming it on CoreAudio? When I speak of CoreAudio SRC, I am referring to the AudioConverter API, and I understand it to be of the best quality that the industry offers today - certainly for upsampling, and most likely for downsampling, given the constraints of the generated low-pass filter.

I have appended the relevant quote, which I would like an educated response to. I have also included another comment which seem inaccurate. I ask for comment because these are the sorts of quality issues that I have been testing in my own audio application development, and, further, I have long planned on developing a sample-rate following player.

IMHO, it is bad form to be changing the audio hardware's sample rate for each track. If you're gunning for the absolute highest quality output then maybe it's okay but it's pretty obnoxious to the rest of the system and should be clearly communicated to the user.

Brian Willoughby
Sound Consulting

2) OSX still does not have the ability to follow sample-rate changes. We consider this a nuisance for most users and a show-stopper for users who want to play a mixed list of 44.1kHz and 96kHz material. An OSX sample-rate mismatch will invoke the very-poor-quality sample-rate conversion that is built into OSX. The iTunes 7.5 bug seems to be locking on this sample-rate conversion at all sample rates other than 96kHz. Until this is fixed, OSX is not the operating system of choice for audio playback.

This is a bug in iTunes but mis-stated above. iTunes grabs the hardware sample rate at launch and runs its internal processing chain at that rate for the life of the app. It has nothing to do with 96kHz per se. Because of this, if you change the audio hardware's sample rate *after* launching iTunes, then there will be a double SRC operation going on. The assertion about the "very poor quality SRC in OS X" is overstated -- the main problem is that there are two of them going on.

[...] After extensive testing and communicating directly with the engineering team at Apple, some of these initial observations have been explained. We now know the reason for the poor performance observed in the initial tests, and we have conclusive information about the operation of iTunes 7.x on Mac OS X.

I'm not sure who they talked to but it wasn't me. ;-)

[...] 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.

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.

Hope that clears things up a little. And now the cat is out of the bag with respect to who I am... :-P

References:
	>Quality of CoreAudio SRC (From: Brian Willoughby <email@hidden>)