Oversampling — Move the Mirror

Our waveshaper created harmonics that crossed Nyquist and reflected back as aliases. Oversampling gives the nonlinear process a larger temporary frequency space — then an anti-alias filter removes what cannot safely return to the host sample rate.

1. Move the mirror

2. But we have to come back down

Suppose the host is 48 kHz but the nonlinear stage runs at 4× = 192 kHz. Harmonics can exist internally up to 96 kHz — but the final 48 kHz signal still cannot contain anything above 24 kHz.

3. How steep does the filter need to be?

A real low-pass filter cannot pass everything to 23,999 Hz and magically remove everything at 24,000 Hz. It needs a transition band.

4. Try 44.1 kHz

Keep the passband near 20 kHz and move the host sample rate from 48 kHz towards 44.1 kHz. The available transition region squeezes:

That reconnects directly to our earlier filter work: sharper boundaries are not free.

5. Oversampling is not a magic “quality” switch

It is particularly useful when processing creates new frequencies that could otherwise alias. More oversampling means more internal samples to process and more filtering work.