# Physical flushing comparison

The physical procedure below has not been performed. Its computational run-through is complete: see `outputs/virtual-bench/report.html`, `completion.json`, and `virtual-bench-inputs.json`. Those results are synthetic/theoretical and identify adverse cases as well as passing conditions.

1. Measure actual chamber and probe dimensions. Include real fittings, side ports and changes in cross-section in a revised model if they are present. The current CAD and calculation cover only a uniform annulus.
2. Choose a passive tracer with independently measured diffusivity. Confirm it does not adsorb on the actual surfaces or change the liquid viscosity. Measure density and viscosity at the test temperature.
3. Fully flood and de-air the chamber. Supply a stable 5 mm3/s reference flow and verify it independently by timed mass collection. Record flow uncertainty and any pump pulsation.
4. Fill the chamber with a known tracer concentration. Switch to tracer-free liquid without changing flow. Measure the inlet transition to establish its actual step duration and time zero.
5. Record outlet concentration continuously. Use an independent near-probe optical or conductivity measurement only if its sampled volume corresponds to z=8 to 12 mm and r=2 to 2.2 mm. Report spatial averaging and sensor time response.
6. Estimate remaining whole-chamber tracer mass from the known initial mass minus calibrated integrated outflow, accounting separately for any inlet transition and tracer taken by sampling. Do not equate a single point sensor with whole-chamber concentration.
7. Report the first crossing below 1% for whole mass, the declared probe region and flow-weighted outlet. Repeat at least three times per geometry; include uncertainty and any sensor lag. Compare complete response curves as well as t99.
8. Measure static pressure across the modeled section. The reference budget is 10 Pa, so use a differential transducer and setup with enough resolution for this range; fittings and elevation head must be accounted for independently.
9. If bubbles, adsorption, leakage or non-Newtonian behavior appear, revise the model before interpreting disagreement as a geometry effect.