Abstract
Classical waveforms generated by oscillators in analog synthesizers have widespread harmonic spectra. While they create bright and rich sounds, they can cause aliasing when they are trivially implemented in the digital domain because of the harmonic contents that are present above Nyquist. Digital subtractive synthesis is a popular synthesis technique which requires designing oscillators which are perceptually alias free. It is challenging to generate computationally efficient waveform genÂerating algorithms which produce similar sounding analog waveforms which are free from audible aliasing. One such implementation is BLEP (BandLimite stEP) where a tabulated correction function is used to manipulate the trivial oscillator. This research presents a solution to the aliasing problem by finding an ideal window using a combination of 17 different windows to manipulate the input function of the BLEP method. A genetic algorithm is used to perform a heuristic search based on the evaluation model suggested by Mariana Bosi. The idea was to optimize the algorithm to perform for every MIDI note of an 88 key keyboard. In this research, we will also derieve the mathematical analogy between the BLEP method and the Fourier series proving why the aliasing is reduced and justifying how the BLEP implementation achieves the results it does.