[sdiy] Digital accumulator VCO core?

cheater cheater cheater00social at gmail.com
Tue Feb 16 02:34:30 CET 2021


On Tue, Feb 16, 2021 at 2:28 AM Brian Willoughby
<brianw at audiobanshee.com> wrote:

> I just realized that I should have written sinc as cos(x)/x because that's symmetric around x == 0.

cos(x)/x is not sinc and is a completely different function,
unrelated, and does not have any of the same properties.

> How do you avoid canceling out the step itself?

You don't. You want it to cancel out.

> Sub-sample alignment of the MinBLEP will create some variation. Also, if the engine runs at high resolution and then dithers down to 24-bit to avoid correlated quantization, that will add some randomness by definition.

I'd want more variation than that.


On Tue, Feb 16, 2021 at 2:31 AM Brian Willoughby
<brianw at audiobanshee.com> wrote:
> On Feb 15, 2021, at 14:09, Mike Bryant wrote:
> > And you don't just use an FFT to do autotune, that's merely a starting point if you want to do it in real time, as it would be for a real time spectral gate.  Some of the best systems don't even use an FFT at all nowadays.
>
> You would perhaps be surprised by how many algorithms use FFT behind the scenes, without making it obvious. Impulse-based convolution (incl. reverbs) use FFT to vastly reduce the CPU usage below the textbook convolution math.

agreed, Autotune is a terrible starting point for something that just
merely requires FFT.



More information about the Synth-diy mailing list