← Algorithms · Buffer subtraction

Buffer subtraction

Purpose. Solute scattering is I_sample − s·I_buffer. The hard part is choosing the scale s so that solvent / capillary contributions cancel while the solute signal remains. Visually, sample and scaled buffer should match in the high-q “tail”, especially past the Porod region / before a detector “knee”.

What autosaxs does.

1. Detect the detector knee on the buffer alone (SNR-weighted descending-edge contrast on rolling-median log I). Set q_max just before that cliff.
2. Take the late pre-knee band (default: last 35% of [q_min,data, q_knee]). Optional manual q_min / q_max override this.
3. In that band, gate low-SNR points, form raw ratios I_sample/I_buffer, take sliding-window medians, then s = approach_factor × min(medians).
4. Write I_sub = I_sample − s·I_buffer with propagated errors. Flag over-subtracted if any point has I_sub+σ_sub < 0.

A fixed scaling factor (Subtraction wizard) overrides automatic scaling.

Engines. Fully custom (no ATSAS).

← Azimuthal integration · Guinier analysis → · Main window (buffer recipes) →