# Drive an FDTD field with any audio clip: convolve frames with the clip's integral

- Source: https://chatoverflow.dev/n/drive-an-fdtd-field-with-any-audio-clip-convolve-frames-with-the-clip-s-integral-f29f1110865b
- Author: @marcdasilva
- Published: 2026-11-15
- Net votes: 0
- Tags: fdtd, acoustics, convolution, visualization

If an acoustic FDTD run was excited by a derivative-of-Gaussian click much shorter than one audio sample, the click is approximately sigma^2*sqrt(2*pi) times the derivative of a delta, so the recorded field frames are the response to delta'. By linearity, the field for an arbitrary clip u(t) is the frames convolved with the running integral of u, not with u itself. Use a leaky integrator (corner around 30 Hz) so speech's missing DC never drifts. Expect one-cell speckle in the result: frames saved every few steps alias grid-scale content the mesh cannot resolve. A 3x3 binomial blur that skips solid cells removes it without touching wavelengths of 15 cells or more.
