KVR Audio

tony tony chopper · Post by **tony tony chopper** » Fri Sep 17, 2004 5:25 pm

gol, the page states there are offline renderings, available in the same line of the product image link.

yes, and I have a futurama bender toy, and on the box it says "Bender Robot, guaranteed* to provide you with several minutes of distraction from your otherwise dull and meaningless life."

and at the bottom, in tiny, it's written
* Not actually guaranteed

WilliamK · Post by **WilliamK** » Fri Sep 17, 2004 5:28 pm

gol wrote:
gol, the page states there are offline renderings, available in the same line of the product image link.
yes, and I have a futurama bender toy, and on the box it says "Bender Robot, guaranteed* to provide you with several minutes of distraction from your otherwise dull and meaningless life."

and at the bottom, in tiny, it's written
* Not actually guaranteed

I love futurama, where can I get one? ... must ... have ... Bender Robot ...

Wk

tony tony chopper · Post by **tony tony chopper** » Fri Sep 17, 2004 5:35 pm

it sucks (it even has one leg longer than the other) as much as any licensed toy.. but it's bender..

Mc farlane should do a bender toy. But it'd surely end up all bloody and gore.

René · Post by **René** » Fri Sep 17, 2004 5:45 pm

It is really interesting to hear everyone's point of view, I believe that our beliefs drive us in everything we do.

I for instance, couldn't care less about aliasing in a synthesizer (even when all the synthesizers I've written include some bandlimiting algorithm).

A synthesizer is a device defined as 'a-box-which-makes-noises-musically-useful'. So on synthesizers, everything is valid: aliasing, uber aliasing, and extra-uber++ hyper-mega-aliasing TM. There will always be some genius who find a great musical use for that noise.

Perhaps the only exception to this would be the vintage-emulation synthesizers, which should perfectly recreate something, no matter by what means. If we'd do a Poly-800 emulation with no aliasing, it wouldn't be cool.

And I say Perhaps because the 'good thing' about emulations seems to be that we can 'bend' the original design, and always state that the deviation was to have 'all benefits from the old thing and none of the disadvantages'. Amongst the best emulations I've seen there's always the 'and this has this also' thing, which makes the thing richer.

Samplers, on the other hand, are defined as 'a-box-which-plays-a-sample' in a certain way.

That task, is a -technical- task, not a -musical- task, even when we later decide to use that downpitched cymbal as a snare. After we've found a solution which is included in the pretended specifications, THEN we can move to the musical arena.

Just to end my participation in this thread, I would like to let everyone know that the discussion behind technical qualities of samplers has ever existed.

Back in the old good days, discussion was centered on if compressed 12-bit samplers sounded better than 16-bit samplers. The flames on the old Compuserve were almost as high as in this thread.

However, I have to agree 100% with a comment stated above in this thread: users get huge benefits from this kind of thread. Developers push themselves harder when all is about to show the world their compiler size. Even harder than they do to please customers

So, great thread.

-René

Muon Software Ltd · Post by **Muon Software Ltd** » Fri Sep 17, 2004 5:45 pm

but surely linear interpolation (!), hermite interpolation and the rest are all effectively FIR filters

No, not at all. They're not LTI processes at all and can introduce spectral elements not in the original signal.

Regards
Dave

WilliamK · Post by **WilliamK** » Fri Sep 17, 2004 5:45 pm

Muon Software Ltd wrote:So I sort of have to wonder if all of this is just p155ing into the wind. I would be happy to have DS404 near to the bottom of the chart because it only uses linear interpolation for speed and I guess it doesn't sound all that bad Regards Dave

I'm really glad you jumped into this discussion. Nice to know that those are not exactly aliasing, but something else. But could that still be a problem for the sound-quality?

But I wonder now, and see, I'm using the word "wonder", not stating anything ok? Should George use the words "High quality" - "Avarage Quality" - "Poor quality" ? I mean, maybe you should let the users think if the quality is "Poor" or not? But that's what I think.

Also, everything should be listed using the highest-quality possible. Even if is not real-time. But mention which quality was selected. Just one idea.

Wk

george · Post by **george** » Fri Sep 17, 2004 6:04 pm

Well, I have updated the page, it has a few changes so I've taken the PDF down until I can update it.

URL is http://www.buzzxp.com/samplers in case you are too lazy to search for it

Post by **Ben [KVR]** » Fri Sep 17, 2004 6:13 pm

So what does the output from this look like:

http://www.kvr-vst.com/get/1213.html

tony tony chopper · Post by **tony tony chopper** » Fri Sep 17, 2004 6:13 pm

btw as far as I remember, halion also renders at a higher quality (or I should say, process at a higher quality depending on which thread the host is, because there's no concept of rendering in the VSTi specs)

Muon Software Ltd · Post by **Muon Software Ltd** » Fri Sep 17, 2004 6:18 pm

But could that still be a problem for the sound-quality?

Well, if the interpolation algorithm is creating noise, that's noise that shouldn't be there. If you look at the frequency response of linear interpolation, you'll see that it has some impact in the band of original harmonics that you want to keep - there's a slight roll off of the highest harmonics. It also doesn't do a particularly good job of removing the spectral images inherent to interpolation. Hermite does a better job of both. So I guess you could definately say the hermite interpolation has less undesireable side-effects compared to linear interpolation....and if you re-arrange the formula a bit you can implement it in such a way that it costs only a little bit more than linear.

I'm not sure the noise generated by linear interpolation is all that offensive though.

george · Post by **george** » Fri Sep 17, 2004 6:23 pm

WilliamK wrote:Should George use the words "High quality" - "Avarage Quality" - "Poor quality" ?

I agree with you, William. Poor is not such a good description, because having that quality doesn't automaticly make it a bad product, so I have replaced it with "Low" (looks more consistent that way, tho).

autloc · Post by **autloc** » Fri Sep 17, 2004 6:24 pm

Muon Software Ltd wrote:No, not at all. They're not LTI processes at all and can introduce spectral elements not in the original signal.

Regards
Dave

I think we must be talking at cross purposes. The equation y[n] = 0.3*x[n] + 0.7*x[n-1], for example, (a linear interpolation or 1st order FIR) cannot add spectral components. Because it is accompanied by a sample-rate conversion, however, aliasing occurs. I did a google for this and turned up JoS:

http://www-ccrma.stanford.edu/~jos/Inte ... inued.html

Anyway, as I said, I think we're saying the same thing but going about saying it in different ways. I'm sorry that I've made a mess out of this thread by trying to be precise and doing the opposite

I was just trying to say that the interpolation noise is in fact aliasing (of the filter) caused by the sample rate conversion.

WilliamK · Post by **WilliamK** » Fri Sep 17, 2004 6:24 pm

Muon Software Ltd wrote:
But could that still be a problem for the sound-quality?
Well, if the interpolation algorithm is creating noise, that's noise that shouldn't be there.

Muon, I must say that this is the best explanation ever. So the whole problem is not really aliasing, but noise caused by the anti-aliasing codes such as Linear-Interpolation. Good to know. I'm glad I added Hermite on my latest update.

One users just told me that he thinks the audio is much cleaner now. And I agree, as we removed the noise that LI was creating. But it wasn't aliasing.

Good to know that. Thanks a lot, really.

Best Regards, WilliamK

autloc · Post by **autloc** » Fri Sep 17, 2004 6:25 pm

george wrote:
WilliamK wrote:Should George use the words "High quality" - "Avarage Quality" - "Poor quality" ?
I agree with you, William. Poor is not such a good description, because having that quality doesn't automaticly make it a bad product, so I have replaced it with "Low" (looks more consistent that way, tho).

Low sounds the same to me as "poor". Why not just say "noisy" and "clean"?

WilliamK · Post by **WilliamK** » Fri Sep 17, 2004 6:49 pm

Muon Software Ltd wrote:and if you re-arrange the formula a bit you can implement it in such a way that it costs only a little bit more than linear.

Indeed, using SSE in Paralel and Cache-Prefetch it costs almost the same. That's why Wusikstation default quality is now Hermite.

I did an extreme test. 6 Layers with 16 voices playing a total of 128 internal voices (W1/W2 counts as doubles) = 26% of CPU usage (Linear Interpolation) on my Athlon XP 1.6Ghz. With High-Quality (Hermite) = 30%.

Wk

The aliasing thread