KVR Audio

LeVzi · Post by **LeVzi** » Fri May 11, 2012 2:17 am

I dunno what that graph is all about, but I just loaded up 2 for ableton live and they were both different......

hibidy · Post by **hibidy** » Fri May 11, 2012 2:24 am

It's just a load bug. If you go back and load them again, they are identical.

1-2-Many · Post by **1-2-Many** » Fri May 11, 2012 3:17 am

jancivil wrote:frankly the whole exercise reads like someone looking for a solution to a problem they don't actually have.

Not trying to invent a problem - Just trying to understand what this means IRL. There's been a lot of talk here about how the host doesn't really matter re the output quality of the rendered audio. Well here's evidence which might say differently.

ermi · Post by **ermi** » Fri May 11, 2012 8:36 am

SRC is a process which you can do in many different ways, you definitely can't say all hosts are the same if you're using SRC. If you're just using the same sample rate throughout things are much more simple, I believe.
If you're interested in this stuff I recommend you to visit Hydrogenaudio.

aciddose · Post by **aciddose** » Fri May 11, 2012 9:01 am

if any host does sample rate conversion in real-time it's pure insanity. most provide the option for offline resampling where you can use much better filters.

it also is only needed if you don't work in a fixed rate. for example say you sample at 44.1k but work at 96k, that's a problem.

in that case you should either sample at 96k, or offline resample any material you're going to use in a project up to 96k in order to work with it.

that said, up-conversion isn't very difficult at a fixed ratio.

down-conversion is the issue. what reason you'd ever have to down-convert in real time i have no idea. other than a sampler which usually might use 2x oversample plus a hermite/linear combination i really don't see any reason you'd do this. hermite even works ok up to about 1.5x, the 2x oversample is only needed if you want really clear results. why would you sample less than per-semitone anyway? otherwise you'd need a more refined method like sinc interpolation which works perfectly at any rate.

this is one of those things you should notice though. "i'm playing this sample at 22.05k in my 96k project". that will be kind of obvious. 44.1k -> 48k is less obvious and can have a huge degrading effect. only real solution is to sample at 48k instead, or resample offline if that isn't possible. 96k -> 48k or 48k -> 44.1k is trivial but expensive and you should notice the cpu hit once you get around 20 tracks going like that. offline resample and you'll convert that cpu hit to zero.

jancivil · Post by **jancivil** » Fri May 11, 2012 9:08 am

1-2-Many wrote:
jancivil wrote:frankly the whole exercise reads like someone looking for a solution to a problem they don't actually have.
Not trying to invent a problem - Just trying to understand what this means IRL. There's been a lot of talk here about how the host doesn't really matter re the output quality of the rendered audio. Well here's evidence which might say differently.

In Real Life? The thread title singles out a particular vendor. At KVR this begs a flame war.

*IRL* are you hearing a difference between hosts? or are you looking to identify a problem (that is in context to other remarks you make here, NB) with one vs another.

Or, would you want to assert you can hear a difference between sample rates? (Per se, all things [SRC] being equal.) If so, would you subject yourself to blindfold testing on that? etc...

as just said, this is 'meaningful' for people that have the belief they need to work at more than one rate. I don't really buy into that, it's a waste of time. That's my POV. I think the ideation that a person at the end-user level, ie., creating content can discern the difference of sample rates is driven by marketing hype. I can grok why a developer of audio software or recording samples would care to go to this extent but that's a very technical matter.
If you *must* downsample considerably it's something to study. But KVR 'Hosts' as the particular forum for this arcane type of discussion, isolating a vendor in the beginning isn't the best idea I've seen today.

Life [in digital audio] is as simple or as convoluted as one wants it to be I reckon.

1-2-Many · Post by **1-2-Many** » Sat May 12, 2012 6:55 pm

Yes - IRL - In real life I use Cubase5. In real life I just want to be sure that I'm not doing things within my host that could be compromising the output quality of my projects. I have sometimes upped the sample rate to 96 to get a certain VSTi to sound better, but it looks like I should just render this out at 96 and then use a better SRC to bring it down to 44.1 and then reimport it with the project set at 44.1

In this respect, aciddose's post was far more helpful than anything you've added to the conversation - IRL (Thanks Aciddose). Seems you're the one trying to make problems where there aren't any.

hibidy · Post by **hibidy** » Sat May 12, 2012 7:00 pm

Everyone should just go with protools. Everyone knows protools is the only way to make giga bling.

jancivil · Post by **jancivil** » Sat May 12, 2012 7:29 pm

1-2-Many wrote:Yes - IRL - In real life I use Cubase5. In real life I just want to be sure that I'm not doing things within my host that could be compromising the output quality of my projects.

So work in 44.1 and render in 44.1. Upping the sample rate to that, as I said, do you have a belief that it improves anything? It doesn't. Upsampling does in no wise add to the information you have. (As far as editing audio in the project, theoretically it's a good idea to have some headroom such as Cubase provides with 32-bit fp. Theoretically. This isn't that.) You have a non-issue you are choosing to worry about:

1-2-Many wrote:I have sometimes upped the sample rate to 96 to get a certain VSTi to sound better, but it looks like I should just render this out at 96 and then use a better SRC to bring it down to 44.1 and then reimport it with the project set at 44.1

It looks like that since you found a "problem" you don't understand. IN MY OPINION a waste of time. Do you hear the improvement, really? Again, would you care to blindfold test that belief?

1-2-Many wrote:Seems you're the one trying to make problems where there aren't any.

As you now reveal it, it is you looking for a problem (which you're only amplifying for yourself, in public). I don't care what you choose to waste your time on. I'm commenting as a member of the forum with an opinion.

hibidy · Post by **hibidy** » Sat May 12, 2012 9:00 pm

I see the quatloos are not quite rolling in, but it's quite possible that will change soon.......come to papa, hibidy needs a new pair of shoes (literally)

aciddose · Post by **aciddose** » Sat May 12, 2012 9:24 pm

jancivil wrote:So work in 44.1 and render in 44.1. Upping the sample rate to that, as I said, do you have a belief that it improves anything? It doesn't.

jan, you're so wrong here it hurts.

there are a large number of reasons oversampling will produce better results. you'd likely even want to work fully at 96k, then only resample down to 44.1k for the post production process where you really "master" for the cd version.

in other cases you'll want to use 48k or 96k for example on blu-ray or game systems. cd is still a popular format but you might be surprised to check what the ideal rates to run an ipad or ipod audio output at actually are.

a short explanation of one interesting issue: plugins which use oversampling filters to eliminate aliasing will build up phase shift near the cutoff of those filters. 20k is a common value to use when you're running at 44.1k. in those situations you have only 2khz bandwidth for the transition band of the filter. at 48k this doubles to 4khz. 96k you get more than a full octave!

you won't be capable of understanding anything i just said unless you have the knowledge and experience. in any case, there are very valid reasons to run at higher rates.

the phase shifting isn't the only reason, there are a large number of related issues and it gets awfully complicated to explain. you'd need to research the subject for months at least. ideally i'd have you not believe anything and do your own research. maybe it isn't worth it for you, but if you choose to "believe" one point of view at least believe the correct one.

jancivil · Post by **jancivil** » Sat May 12, 2012 11:15 pm

Run the project at 96k

to get a certain VSTi to sound better

because of the way a plugin oversamples to obviate aliasing problems. Ok then.

"believe" a point of view... I have a point of view that this esoterica is not as relevant as you find it. This is your purview. On the numbers, I would tend to grant you're right, you're right and I'm wrong. Fortunately I do music with most of my time. Which you have been so very wrong on, consistently argumentum ad culum, trying to fake knowledge of. Are we even do you think?

aciddose · Post by **aciddose** » Sun May 13, 2012 12:15 am

not sure, i have no idea what you're talking about. i've tried to fake what?

no, not only will running at 96k be an improvement "to get a certain plugin to sound better", it can potentially improve the quality of a signal passed through multiple plugins in the high frequency range. that's what i was saying. these are sort of extremes and sure we can make the argument "why not just forget all this and make your music on two spoons? that's what i do and it works fine for me." that can be a perfectly valid argument. you have to throw away the context though and if you take that into account it isn't a valid argument. to someone else it might be important even if it isn't to you.

Steinberg's NotSoHot SRC...