I know giancarlo said there would be no change and that the algo is good, but ...
A few years back I used the site http://src.infinitewave.ca/ and came to the conclusion that actually the freeware src algorithm from SoX http://sox.sourceforge.net/SoX/Resampling (with VeryHighQuality Linear Phase parameter) are one of the best out there, and even beat most of the very expensive and commersial src algorithms. And of course I know it is more into it than just reading the analyzing charts.
Giancarlo says the algoritm for NAT and Nebula is very good, but does it compare to SoX? SoX is freeware so I just wonder if Nebula in any way could take advantage and be of higher quality in regard to src if the SoX algorithm was implementet in some way?
By the way, if someone want to use SoX on their MAC for normal file conversion, I wrote a script some time ago making it easy to use SoX (why did not someone do this years ago . I postet the script on Gearslutz forum and you can find it here http://www.gearslutz.com/board/masterin ... t-osx.html
fuseburn wrote: You might wanna take a look at http://src.infinitewave.ca/ You'll be surprised how weird things can get with SRC algorithms.
About Mac, Logic Pro sample rate conversion is quite good, much better than Weiss Saracon. I think the little artifacts showen on http://src.infinitewave.ca/ site's graphs are quite unaudible. Sox is good too. Or, (not free) Izotope. The problem on Logic is 32 bit (NRGui). You should use Soundtrack Pro, included in Logic Studio, or another audio editor like Steinberg Wavelab. Or Wave Editor. But that change of phase is quite big. For recording your own samples I can understand working at 96.000. But working with sound libraries or software synths (Machfive, Exs24, Scarbee, Rex etc.) the sample rate today is always 44.100 24bit. I know 96000 sounds better. But I don't see the point to convert them to 96.000 (and no NAT on mac), when quality stays the same, and then downsample all for mastering at 44100 (And then considering that, I think, the vintage gear sampled didn't work at 96000). Someone says it's better start from an higher sample rate, (the double law), someone not. This is not a sample rate conversion war. But I find much more quick and less heavy for the system working at 44100 (Not really a problem for me, luckly I've a new Mac Pro). But the graphs you kindly posted reveal a problem. Maybe not big, (It may be corrected with some tricks) but... But maybe this It's all an exaggeration. At the end, it sounds great. Perfection is not of this world.