Norm of the inferred jitter amplitude against the simulated (true) one. Top: noiseless case. Bottom: numerical experiment with S/N = 100. There is an almost perfect correlation between the sensed and true jitter amplitudes up to ∼0.″10 when the simulated images are free from noise. The results begin to scatter from there on, with the algorithm tending to underpredict the induced jitter. When noise is included, the algorithm is more stable above 0.″10, but shows more dispersion below∼0.″01.