May I jump in on this topic ... [:)]
SRC is actually _extremly_ dependant on its implementation. The single algorithms differ a lot regarding quality and/or speed. It is only partly a question of how good the actual conversion is done; it's much more dependant on how the anti-aliasing filter is realized.
The problem is: The filter should be as steep as possible, cutting everything above half the sampling frequency - while leaving everything below this so-called "Nyquist-frequency" untouched. But steep filters like this tend to be "ringing" around their center-frequency, adding very high, unwanted overtones and blurring of the transients.
To make this re-sampling filter both smooth _and_ effecient is probably the most discerning aspect of different SRCs.
When we planned the necessary production steps of the Vienna Symphonic Library, I tested and measured about 15 different applications (and even hardware) for SRC. It turned out that they were extremly different measurement-wise, and even the best ones had either acoustical or practical drawbacks, like being to slow for our needs. On paper, Cool Edit (now Adobe Audition) was amongst the best, head-to-head with mastering-class DAWs like Sadie. But audio-wise, these two sounded completly different. Go figure ... AudioEase's Barbapatch was tested but not taken into account back then, as Mac-applications didn't make sense in an all-PC working-scheme.
But in the end, we had our own Batch-Resampler programmed by a Swedish team of DSP-freaks, fitting our needs like a glove. We use this little application ever since.
.... if _I_ were in the market for a (reasonably) high-quality SRC today, I would look at Voxengo's
R8brainPro if you're on PC, or AudioEase's [URL=
http://www.audioease.com/Pages/BarbaBatch4/BarbaBatch4.html]BarbaPatch[/URL] on the Mac. I would try to stay away, though, from the qick'n'dirty SRCs that can be found as part of audio-workstations like Nuendo, ProTools, Logic etc.
HTH,
