I didn't get an answer, so thought I'd try again. I'm interested in 
implementing my own layered sampler with the capability to interpolate 
dynamic levels between layers. I want to know if this is a simple matter 
of mixing signals from adjacent layers at amplitudes R and 1-R.

I think the critical question here is whether the samples are " phase 
coherent"? Is that true?

For mixing between non-phase coherent signals, I think you would use an 
equal power formulation: sin R and cos R. Is that right?

It's piano samples I want to use. I believe they are already phase 
coherent as designed.

Thanks,
Mike

Oeyvind Brandtsegg wrote:
> I'm not sure, but it seems like you are likely to get trouble with
> phase cancellation with any kind of mixing. It seems unlikely to me
> that it would be possible to perfectly align the phases of all
> partials in two (piano) samples. But I might be wrong.
> 
> Just an idea,
> maybe you could split the audio into separate frequency bands, mixing
> each band separately. This would mean (simplified)  that you took e.g.
> the high frequencies of sample 2 mixed with the lows from sample 1,
> and used this as the "interpolation midpoint"
> Hi: 1   2  2
> Lo: 1  1  2
> Obviously, this would not get rid of the phase cancellations of
> frequencies very close to each other. And on second thought, those are
> the ones you would be most keen on getting rid of...
> 
> Oeyvind
> 

You know, this is my thought, too. The situation must be more complex 
that just cross fading between sample due to the creation of peaks and 
nulls. And yet gigastudiio claims to do it. And a piano sampler "engine" 
called Ivory claims to do it. Ivory is amazing... it has pianos with 10 
layers, it uses filters to reduce mechanical noise if you want to, it 
can control dynamic range, and you can reconfigure it in numerous ways. 
Want to use only 6 layers and save memory (speed it up)? Fine. Want to 
use merely 4 layers? Great.

I have a strong preference for the 4 layer version, becuase the quality 
of tone and attack doesn't vary a lot from midi vel=10 to midi vel=120. 
In the 10 layer system however, if you send a note with vel=10, you get 
the tiniest, most muted soud you can imagine. If you hit a note a bit 
too hard by mistake and send vel=120, you get this enormous blast of a 
fortiss-iss-issimo thing spewing from your speakers.

Anyway, my interest is in implementing similar sample systems so I have 
a little more control over what they are doing.

Thanks,
Mike


Send bugs reports to this list.
To unsubscribe, send email sympa@lists.bath.ac.uk with body "unsubscribe csound"