Model Comparison Tests

• Directory Tests/SIE, Tests/SIS
• Purpose validate consistency between models.

cd Tests/SIS
sh sis.sh
sh compare.sh

The first command sis.sh generates images using four different configurations, both Raytrace and Roulette in both analytic and sampled versions.
(See SIE below.)

The second command compare each simulation model against its sampled counterpart and analytic raytrace against analytic roulette. The result is visualised in the montage directories montaige-raytrace, montage-roulette, and montage-compare for visual validation.

The comparisons are not expected to be exact. The sampled lenses will produce numerical inaccuracies, and the roulette formalism only works within the convergence ring. The tester will have to rely on informed judgement.

Similar tests exist for SIE. The lens models are kept separate because sampled and analytic approaches are treated as different lenses and there is no easy way to combine SIE and SIS while keeping sampled lenses separate.

cd Tests/SIE
sh sie.sh
sh compare.sh

Sampling Results

The sampling tests (montage-raytrace and montage-roulette) are designed to validate analytic models by using a sampled version as reference.

Roulette Results

The montage-compare results are created to validate the roulette model, and thus the calculation of roulette amplitudes.

Design Notes

• The code is complicated by the fact that sampling is implemented as a separate lens model, and not a separate simulation model. Hence, we cannot syntactically fix the lens model as SIS (resp. SIE) and try all the different simulation models with and without sampling.