Here is another comparison from Bias (makers of Peak), but unfortunately it does not include Voxengo.
I thought you might find it interesting.
I've seen that, thanks.
One another strange thing I've noticed in that Flash SRC comparison is that ALL 1kHz sinewave tests contain spikes in the higher frequencies, clearly visible on all high performance results, including iZotope, r8brain PRO and free and Saracon. Seems like their original sinewave already contained that kind of noise.
Your comments on our tests has caused us to have another look at the method used with minimum phase resamplers. This from Aleksey Lukin, the designer of our test:
"I've found a small inaccuracy of our testing occurring with steep
minimum-phase filters. This is specifically the minimum-phase property
that made our calculation of filter's response inaccurate. I've
corrected this problem and attaching the redone graphs of
minimum-phase resamplers. My apologies to Aleksey Vaneev and our
web-page visitors for this mistake."
The inaccuracy was confined to the r8brain Pro 1.3.1 Minimum Phase results only. Our website has been updated with the new graphs. I am sorry that this has caused you to feel our tests are awkward and clumsy.
In answer to your question regarding the 2MHz range of the graphs, Aleksey Lukin had this to say "We get 2 MHz reading in our test by analysis of the original resampling filter's impulse response (this filter operates at the least common multiple sampling rate of 44100 and 96000)."
I hope this helps to clear things up. I am a proud user of r8brain Pro and love the new version.
Thanks! And thanks for the comment about 2MHz+ range - this still is not very explanatory, but I have concluded for myself that you are performing some kind of aliasing 'unwrapping'.
I suggest you to review 1kHz test tone test as well (please see my note above). Please check out BIAS Peak 5 and Digital Performer 1kHz tests - they are identical - probably misplaced. The same applies to iZotope and Logic Pro 1kHz tests - they look identical while the overall response is different.
Ah, good catch. I have corrected that and the new version is on the site (you may need to hit refresh.) I was a little rushed when I posted the Logic Pro and Digital Performer results. :)
We are having a look at your concerns regarding the 1kHz test. I will post when a conclusion has been reached.
Our testing method for the overall response is legitimate, and was approved by the engineers at Weiss prior to being used. I would be happy to discuss it with you via email. Please feel free to contact me.
The extraneous spikes are quantization effects in the original file. We don't think these are a problem, as they are below -165dbfs. I have made a note to this effect and posted a graph of the original file: http://src.infinitewave.ca/Original.html
I will make this information available in the help section of the site.
Good read guys! Lots of good information gathered thanks to the ongoing discussion here and elsewere, thanks!
As I'm slowly moving into professional (read: making a living, I might totally suck at it!) mastering, we just got hold of a 92 square meter flat in the middle of Helsinki, I'll be purchasing R8brain Pro in the coming weeks. Will take it for a demo ride right now.
Aleksey - from the test as it stands right now, it looks like the iZotope SRC comes out on top, but the only discernible advantage is S/N ratio... as you've alluded to, it could be because the iZotope SRC is a 64-bit algorithm (as indicated on the iZotope page: http://www.izotope.com/tech/src/)
Would you consider implementing the 64-bit mode as you've suggested, as a means of clawing back that slight advantage iZotope seems to have at the moment?
Sure, I still do have plans to implement 64-bit conversion, but this is not of top priority. The quality difference is negligible (from real-world acoustics side of things). Beside that as you see, iZotope went 'licensing' route. Probably they will try to release it as standalone only if their 'licensing' approach fails. Nobody contacted me to license r8brain PRO sample rate conversion, while there are many individual customers. :) Let's see how it goes.
Just a couple of thoughts.
This topic was last updated 180 days ago, and thus it was archived. Replying is disabled for this topic.