KVR Audio

headquest · Post by **headquest** » Wed Feb 08, 2006 8:37 am

There's been some interesting independant research into the quality of sample rate conversion in popular hosts, audio editors, and specialised plugin applications.

The most surprising result is that there seems to be considerable difference between the sample rate conversions in the different applications tested. Considering that digital audio consists of "1"s and "0"s, I guess that I would have expected otherwise. Having said that, there are some who for years have disregarded digital audio and plumped for old school analogue equipment, claiming better results. So perhaps we should not be too surprised.

Also there have been a fair few threads here at KVR where it has been argued that one host "sounds better" than another, and it seems to me likely that issues such as sample rate conversion might ultimately provide some possible answers to these questions and debates.

The research is presented here:
http://src.infinitewave.ca/

To see the full results for each application you need to toggle between "Display Modes", of which there are four:

* overall response
* 1 khz tone
* Passband
* Transition

To see each application click the "Next" button under "Navigation" and scroll through the software they tested.

Sequencers generally did not fare too well in these tests. Sequoia (Samplitude) performed the best, followed by Pro Tools. Nuendo (and CUbase SX) performed poorly, but Logic and Sonar were even worse still. This is surprising when you consider that those two are so highly regarded by the pros.

The message appears to be, don't do sample rate conversions within your sequencer - use an audio editor or plugin...

Audio Editors generally performed much better than the sequencers. Bias Peak acheived good results on the mac platform, while on Windows Adobe Audition clearly did better than either Wavelab or Soundforge. In fact Soundforge did rather poorly considering its reputation. Meanwhile Goldwave was the worst performer of all the software tested, by quite a margin!

Other Programs tested included a new 64-bit SRC application coming from Izotope, which overall acheived the best results on test. r8brain Pro (from Voxengo) performed well in all tests except the 1khz tone, where it was behind some of the others. Weiss Saracon (and we're talking mega-bux here!) performed well overall, but not so well in the "Transition" test.

Your thoughts?...

arke · Post by **arke** » Wed Feb 08, 2006 8:49 am

My thoughts ... it don't matter to me. I compose at 44.1 ... I mix at 44.1 ... I export at 44.1 ... it don't matter to me the least bit.

Kingston · Post by **Kingston** » Wed Feb 08, 2006 8:52 am

I think the more important thing to note is that Voxengo r8brain Pro actually handles the absolutely crucial steep cutoff much better than anything.

Compare that to the equally good (if actually slightly worse) Weiss converter software. Price difference?

Around $3000

I mean seriously, which one would you pick?

bmanic · Post by **bmanic** » Wed Feb 08, 2006 9:43 am

AFAIK the iZotope ozone one looks best on paper. It's not released though.

Cheers!
bManic

Kingston · Post by **Kingston** » Wed Feb 08, 2006 11:01 am

Well it's in the same price range so the same thing applies with izotope and weiss, too. Although we're already wanking around -190dB so the difference with the best ones is not only academic, but totally irrelevant.

headquest · Post by **headquest** » Wed Feb 08, 2006 3:30 pm

Just to let you know, I have completely rewritten the opening post.

When I started the thread I was in a hurry and was sloppy in giving details. I have now provided a more thorough analysis and conclusions which I hope will interest others who read the thread.

Funkybot · Post by **Funkybot** » Wed Feb 08, 2006 4:05 pm

Just cause no one's said it yet, and it needs to be said: Way to go man, thanks for the putting the work into this!

nuisance sonore · Post by **nuisance sonore** » Wed Feb 08, 2006 4:23 pm

My first thought? I felt vindicated in using Voxengo free r8brain exclusively after reading about it in newsgroups a few years back, as it seems to be doing as good a job as anything else. Way to go Voxengo and thanks for many years of use of a great and free product.

My second thought: "I should really buy the pro version, if only to support Alex, even thought the free version caters to all my needs."

Followed by: "I only really use sample rate conversion for when I need to post mp3 somewhere, considering mp3 low quality does it really matters the quality of the sample rate conversion at this level?"

Thanks for posting this headquest.

loomchild · Post by **loomchild** » Wed Feb 08, 2006 4:35 pm

headquest wrote: Your thoughts?

Any claim backed up as strongly as yours deserves interest.

bmanic · Post by **bmanic** » Wed Feb 08, 2006 4:46 pm

Kingston wrote:Well it's in the same price range so the same thing applies with izotope and weiss, too. Although we're already wanking around -190dB so the difference with the best ones is not only academic, but totally irrelevant.

Yes I agree but I do hear a difference between the R8brain pro converted file and the original when I do the conversion more than once so "perfect" sample rate conversion is still desireable. As it looks in the test, the iZotope one performes rather much better than the r8brain pro one on the 1kHz test tone. But then again, Aleksey found a whole host of bugs in the test that they corrected so I'll stick with my ears to decide which one I like and so far, r8brain pro kicks some serious ass (call it placebo but I actually LIKE what r8brain does to the samples.. it does something good. Maybe it's just the slight filtering of highs or the tiny phase shift in minimum phase mode but I still like it!).

Cheers!
bManic

nuisance sonore · Post by **nuisance sonore** » Wed Feb 08, 2006 5:15 pm

I'm curious thought, what do you use sample rate conversion for?

I mean, when I work on something, I usually know what the destination sample rate will be, and work from start to finish at the said SR. So, for DVD work, all my projects would be at 48KHz from start to finish.

Unless you work with loops and convert them to the project SR before adding them? Or sound samples that are used in sampler that you convert before using? Would that be the best thing to do?

Spaceman Sounds · Post by **Spaceman Sounds** » Wed Feb 08, 2006 5:25 pm

If like me you work at 88.2khz because vsti's and fx all sound to my ears a lot better in the aliasing department... then sample rate conversion is a must when making media ready for 44.1khz.

I've actually found that recording at 88.2khz andthen resampling using R8 brain sounded better than recording at 44.1khz.

bmanic · Post by **bmanic** » Wed Feb 08, 2006 5:52 pm

I use it when mastering, all the time (except lately, after hearing the difference between r8brain pro and free I can't arse myself to convert anything until I can afford r8brain pro!

). I first upsample the file from whatever format it was in to 96kHz and then apply all the processing I need. Then I downsample it to 44,1kHz for CD burning.

Cheers!
bManic

jupiter8 · Post by **jupiter8** » Wed Feb 08, 2006 5:56 pm

bmanic wrote:I use it when mastering, all the time (except lately, after hearing the difference between r8brain pro and free I can't arse myself to convert anything until I can afford r8brain pro! ). I first upsample the file from whatever format it was in to 96kHz and then apply all the processing I need. Then I downsample it to 44,1kHz for CD burning.

Cheers!
bManic

Wouldn't 88.2 be a better choice??

Borogove · Post by **Borogove** » Wed Feb 08, 2006 6:16 pm

I think it's worth noting, and possibly marking on the graphs, the -96dB limit of 16-bit data, and the -144dB limit of 24-bit data. While some of these conversions are obviously higher quality than others, it looks like almost all of them are "good enough for rock-n-roll".

Sample Rate Conversions - hosts compared!