Re: Quality of CoreAudio SRC
Re: Quality of CoreAudio SRC
- Subject: Re: Quality of CoreAudio SRC
- From: Brian Willoughby <email@hidden>
- Date: Sat, 16 Feb 2008 17:29:09 -0800
On Feb 16, 2008, at 12:05, Herbie Robinson wrote:
New results are available at http://src.infinitewave.ca/ for
CoreAudio and Logic 8.
All of these seem to allow some amount of aliasing, at least
compared to the "iZotope 64-bit SRC Steep, No Alias"
I don't see any aliasing in the Leopard sweep. I do see some sort
of noise at multiples of 100 Hz in the 1K tone test, but that can't
be aliasing unless there was noise in the input signal (which is
why I would have liked to see the input signals before SRC). It
looks more like truncation modulated by the sampling rate (44KHz +
100Hz).
Check out the transition results, Herbie. Even maximum quality
Mastering mode allows signals above 22050 Hz to pass at around -42
dB. It's hard to tell exactly, given the resolution of the graph,
but the aliasing is at least louder than -48 dB.
To be fair, it seems that rather few of the SRC implementations
perform perfectly in this regard, except for the iZotope No Alias.
Merging AMR, Minnetonka AWE, Pyramix, r8brain Pro, Reaper,
Samplitude, Secret Rabbit Code (Best Sinc), Sequoia, Sonar, and Sonic
HD (Steep) also have no aliasing, but have other shortcomings in
tradeoff.
Depending upon your source material, aliasing filters may not be very
important or even needed at all, which is why some SRC
implementations make it optional.
Can anyone tell me whether G4 and G5 results would be bit-
identical? Are there examples of Altivec code which is G5-
specific? My assumption is that the PowerPC model set everything
in stone before the G4 or G5 implementations, and the only
difference would be speed.
I think IBM dropped a couple of instructions, but nothing that
affects audio processing.
Thanks! I was about to set up the same DAW on the same OS and pit G4
against G5.
If someone else has Intel Mac, Infinite Wave might appreciate careful
results to be submitted. Every SRC output I created was realized on
G4 or G5 processors, so we haven't seen whether Intel SSE can beat
PowerPC Altivec on any of the quantization noise measurements. I
suppose that some of the SRC implementations on the existing chart
are Intel only, e.g. Adobe Audition.
Brian Willoughby
Sound Consulting
_______________________________________________
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