Into the void

In my early DSP days, I had to address aliasing in my distortions. Like anyone starting out, linear oversampling was all I could access, as other solutions were too heavy in both math and CPU for my plugins.

Years passed, and when I joined NI, I had the chance to discuss this topic with my colleagues and friends Vadim Zavalishin and Mike Daliot. The status quo was that this was a well-known topic, already covered by people smarter than myself, and it had no perfect solution.

Locked down in Wuhan before it was cool

In 2016, I was in Wuhan with my wife and had to stay in a flat for a couple of days with not much to do. To keep my brain busy, I decided to think about the issue of aliasing. Due to my graphics background and way of thinking, I started picturing it as a geometrical problem to be solved. After a few sessions of just thinking it through, I decided to try and calculate the area of the space between two samples before and after they cross the threshold of a hard clipper.

The result was very satisfying. Right after that, I figured out that by recording the cumulative output of the function, I could store the values in a table. This table, when subtracted from the result of the same table over a consecutive sample input, would give me the area between the two samples for a nonlinear function that wasn’t just a clipper.

I later learned this was called an integral from my colleagues at NI, to whom I presented my results with great excitement on my way back.