Re: Quality of CoreAudio SRC
Re: Quality of CoreAudio SRC
- Subject: Re: Quality of CoreAudio SRC
- From: Chris Johnson <email@hidden>
- Date: Fri, 1 Feb 2008 13:30:00 -0500
On Feb 1, 2008, at 10:28 AM, Brian Willoughby wrote:
Second, I'm not sure what you mean when you claim that the
LavryBlack can clock to 'crystal' - The DA10 only offers modes
which are locked to an external clock. It cannot operate on an
internal clock because it has no data source under it's control.
All possible transports are external, and are data push designs,
thus the need for external lock. Thankfully, the LavryBlack offers
three lock modes, but all are compromises compared to the
capabilities of a (theoretical) self-clocked DAC with bidirectional
control over sample data flow, the only way to make a data pull
model possible.
i.e. The LavryBlack IS NOT a local, crystal clock.
Well, with regard to jitter, we do have the option of buffering a
lot of audio and then reclocking it provided the clock is
'steerable' (which the Lavry evidently is) and the range of
correction does not exceed the size of the buffer. This is what the
Lavry does, and that's why I say there's a latency problem- it is
indeed a result of the 'push model' for sample data flow. We disagree
on whether this reclocking with a 'steerable clock' can be
audiophile- I tend to feel that if it lowers the frequency of jitter
artifacts enough, it can be audiophile. You're correct that there is
still deviation from the ideal time-line going on.
I'm a big vinyl record fan so for me, steering a clock with sample-
rate deviations more in the range of wow or flutter is all good ;)
Again, forgive me for the tangent. I have been pursuing this
technology for years, and it's hard to find the knowledge base
anywhere else! In my thinking, this quest is quite related to
CoreAudio, because CoreAudio has finally embraced the proper pull
model for sample data flow that is needed for a truly audiophile
DAC design.
Brian Willoughby
Sound Consulting
I enjoyed the tangent, so thanks- I think these things are good to
sometimes discuss simply to cross-pollinate things a little. I have
so often seen audio DSP that didn't really take into account
audiophile concerns- moving delay taps which didn't sub-sample
interpolate, for instance- and when you make the extra effort you are
rewarded with product sound quality that has a certain hard-to-define
appeal which I think is quite marketable.
For instance, I've seen discussion on how the iPod actually has a
very respectable DAC with low jitter.
As far as SRC, I think an important thing to remember is that
theoretical perfection isn't necessarily the optimum for sound
quality. I've experimented with the 'Secret Rabbit Code' virtual
analog resampler (not perfect but good) using a pathologically bad
test case (increasingly distorted blue noise) to see at what point,
if ever, I could tell SRCed from raw sound in an ABX test.
What I found was, it was extremely tough to do, and the most
effective mode was not the highest quality mode (with the largest
sinc window) but the medium. This is because the highest quality mode
could be driven into artifacts more easily and developed a sort of
'sheen' that could be heard more easily when stressed. I'm trying to
remember who was talking about the audibility of ringing of sharp
brickwall filter implementations but I'm drawing a blank- it's
exactly this that was affecting my ABX tests.
I'd also observe that with the high CD mastering levels of the early
21st century we're actually closer to encountering these pathological
test cases in normal listening, so it's worth paying attention to how
they're handled...
Chris Johnson
airwindows
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Coreaudio-api mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden