[sdiy] FW: Kawai K3 - Bandlimited Waveforms

[ASSI]

On Sunday 17 January 2010, Plutoniq9 • wrote:
> I've created a graphic showing frequency analysis of both the K3's
> bandlimited waves, and then my waves which simply downsample (w/no
> filtering) the original 512-byte waveform. I tried using a
> low-pass filter before resampling but the resulting waves are
> different then what the K3 shows. It's making me think that the
> K3's waveforms were created using Fourier synthesis, because they
> seem to be just reducing individual harmonics over nyquist.

Low-pass filtering is not going to work well.  Just make sure you have 
a proper periodic waveform to start with (the derivative across the 
cycle point needs to be continous as well), go to the frequency domain 
and then synthesize the shorter waves with fewer harmonics.
 
> I have a question, how come normal wavetable synths don't need to
> go to this extent to reduce aliasing in the higher octaves
> (normally they just use a single waveform right?)

They have a much higher sampling rate, so the typical 128byte 
waveforms are already bandlimited for the highest octave.  These 
aren't really bandlimited of course, so if you listen carefully enough 
you still hear lots of aliasing.  The aliasing in the lower octaves 
that is much easier to hear comes not from the waveform itself, but 
from the quantization noise.

> secondly.....since I'm trying to use PCM waves, what can I do to
> at least reduce the aliasing in the higher octaves, just a steep
> low-pass filter before resampling?

See above.  The Ensoniq SQ80 waves (from the ROM at Rainer Buchty's 
site) would make excellent fodder for this as they are very carefully 
sampled and cleaned.  Let me know if you want help with this.


Achim.
-- 
+<[Q+ Matrix-12 WAVE#46+305 Neuron microQkb Andromeda XTk]>+

Factory and User Sound Singles for Waldorf Blofeld:
http://Synth.Stromeko.net/Downloads.html#WaldorfSounds