Pick an example to load the parameters of one of the repository's example scripts. The preview updates at once; press Run for the full simulation.

RTF computes the room transfer function on an explicit frequency grid. RIR adds the impulse response (inverse FFT of the transfer function on a grid that resolves the shorter of T60 and the RIR length) and its energy decay curve. The RIR window shapes the band before the inverse FFT: minimum phase keeps the response causal (nothing before an arrival, nothing folded across the end); zero phase gives symmetric pulses with a few milliseconds of pre-ringing on a grid extended by a guard interval; none is the raw inverse FFT for diagnostics.

Shoebox: a rectangular room given by length (x), width (y) and height (z). Convex: drag the white vertex handles in the 3D view or edit the coordinates; the room is the convex hull of the vertices, so every vertex must lie on that hull.

Click a wall name to highlight its face in the 3D view. Absorption α is per wall in [0, 1]; impedance is the normalised specific impedance ζ (complex). T60 is estimated from the wall materials when it is not given directly; for convex rooms T60 input is not available.

Drag the blue (source) and orange (receiver) dots in the 3D view or type positions. Reflection order is the maximum number of wall reflections per image source; the 3D view shows the preview's image sources up to order 3 only.

Orientation: Z-X-Z Euler angles α, β, γ in degrees; 0/0/0 faces +x. SH order is the spherical-harmonic truncation of the directivity fit. The sphere radius follows the dataset.

RTF mode solves on the start–end grid in the given step. RIR mode derives its frequency grid from T60 and the sample rate (resolution 1/T60 up to the Nyquist frequency), then applies a band-pass window and an inverse FFT; RIR length only pads or truncates the output.

Every image path of length r gets c · N(t · drift, √t · volatility) added, t = r / c, as in DEISM.update_fluctuations(). Volatility is in s½; paths are unchanged when both volatility and drift are zero.

Example values: the four runs of examples/deism_volatility_example.py use volatility 0, 0.5e-5, 1e-5 and 1.5e-5 s½ with drift 0; the "Path fluctuations" presets use 1.5e-5 s½. Drift is relative: 1e-4 lengthens every path by 0.01 % on average (a 100 m path by 1 cm); negative values shorten them.