[sdiy] How to make wavetables? What size?

[ASSI]

On Donnerstag, 19. Mai 2005 01:01, Metrophage wrote:
> Simply building something which can work is probably easy enough for
> me, though I do not yet quite understand the details of the
> addressing and the tables themselves. I noticed that the Digi and CGS
> use 64 byte waves, whereas the MW uses 256 byte waves.

The MW uses symmetric waves with 128 samples that are 8bit wide (it 
resynthesizes the wave to some better resolution internally) and it 
stores just one half-cycle of the wave (64 samples, 8bit with sign-bit 
at the MSB).

[making waves and tables...]

I've created a few Octave (a free Matlab clone) functions that can deal 
with resampling and resynthesis of waves that have 2^n samples length 
so you can convert one format into another (like VS waves into MW). You 
still need to unpack/pack them from and to their binary 
representations, something that I usually do in Perl.

[Nyquist Theorem...]

What it says really is that a bandlimited signal can be sampled at 
double the bandwidth without losing information. It does not state that 
the reconstructed signal does contain no frequency components above 
fs/2. The spectrum of the sampled signal is in fact infinite and 
periodic and this is why you need the reconstruction filter (or 
anti-alias filter) to remove the unwanted parts of the spectrum when 
doing the DA conversion. This has many important consequences. One is 
that some seemingly innocent waveforms like square and saw may look 
good on the screen, but they alias badly when reconstructed because the 
samples have not been bandlimited. Another is that the reconstruction 
filter gets more and more complex as you approach the fs/2 boundary 
with the bandlimit, which means that most of the time the 
reconstruction filter does not remove all of the excess spectral 
components even if the samples itself are correctly bandlimited.


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

DIY Stuff:
http://Stromeko.Synth.net/DIY.html