Analog and Synthetic VCO

The datasets include recordings from a real analog VCO and a digital VCO emulation, featuring data from the Korg Monologue and the Ableton Live Operator. Three different waveshapes have been collected from the VCO of the Monologue: triangle, square, and sawtooth. Instead, for the Operator, we have collected sinusoidal and sawtooth waveshapes. 

The analog dataset includes 72 pitches in the range [E0, E6]. All the VCO signal post-processing and modulations were deactivated in order to record the clean VCO output. The audio was recorded at 96 kHz with a total duration of 8 seconds per pitch and subsequently downsampled to 48 kHz when training the models. The recordings were postprocessed, obtaining a separate audio file for each pitch and waveshape value, with their first sample aligned with phase 0. The length of the files was limited to 262,144 samples at 48 kHz, equivalent to 5.41 seconds.