Spectralshape relation to the selected waveform

Question

So i t's been some time since i havent used Kyma, buy i've returned :)

I have been experimenting lately with the additive drone example, and more specifically with creating the wavefroms for spectralshape and i have a couple of questions.

Am i right in my assesment that each sample corresponds to a partial's/harmonic's (depending on the spacing value) level control. And if so should the waveform length in samples reflect the number of sines synthesized? I've noticed that a waveform with less samples than the resynthesis adds progressively more noise/aliasing.

What happens when the samples go below zero? Are they disregarded or do they affect the phase of the corresponding partial? My efforts with a waveform's completely below zero led to some sort of noise oscilation in low frequencies and spacings, while at higher frequencies/spacings i got occational clicks.

cheers

Answer 1

Hi Greg,

The shape in the wavetable can be thought of as something like a formant envelope; it remains fixed even when you shift the frequency of the fundamental and the spacing of the partials under that envelope.

The beginning of the wavetable corresponds to a frequency of 0 hz and the end of the table corresponds to a frequency of half the sampling rate.  The spectral shape acts something like a fixed filter or formant characteristic, attenuating the amplitudes of some frequencies more than others.
The value in the Frequency field is used to look into this table (somewhere in the range of 0 hz at the left edge of the wavetable to half the sample rate at the right edge of the table).  The amplitude at that point in the table becomes the amplitude of that frequency in the OscillatorBank or other generator bank that is controlled by this spectral source.   The frequencies of the other partials are: Frequency + (partialNumber * Spacing).  For example, if you set the Spacing parameter to the fundamental frequency, the partials will be harmonics of the value in Frequency.  You can use smaller or larger values in Spacing to get nonharmonic partials.

The OscillatorBank and other generator banks do not expect negative amplitudes so it would be best to avoid those in your spectral shapes.

Comments

Thanks for your reply, what is confusing to me is the static nature of this envelope. I was under the impression that the frequency sets an offset for the "zero" and the spacing is a factor to multiply the analysis bandwidth (sRate/WindowSize). But you are saying that the frequency and the partials are not related to the envelope? Do I understand this correctly?

---

Yes, the envelope shape is fixed, and you can move the fundamental frequency (and partials defined by the spacing) around under that fixed shape. So for example, the fundamental could be attenuated by different amounts as you move it up or down relative to the spectral envelope.