ShotCalibration

The calibration line, and the estimator bias a wrong shot-noise unit makes

9/9 passed in 0ms

Test	What it does	Result
test_bias_signs	An assumed unit above the actual one divides the fitted transmittance by the ratio and lowers the estimated excess noise by (ratio - 1)/(eta*T).	✅ pass
test_conventions_differ	The self-consistent estimate is Jouguet Eq. (8)'s unbiased-slope form times the shot-noise ratio.	✅ pass
test_flat_matches	tlo_estimate's unbiased-slope row equals attacks.calib_xi bit for bit at five ratios.	✅ pass
test_gap_hidden	The excess noise a fixed ratio hides rises as the channel darkens, as (ratio - 1)/(eta*T).	✅ pass
test_jouguet_hides	At Jouguet's T = eta = 0.5 an intercept-resend raising the excess noise to 2.1 SNU is erased by attacks.calib_ratio's shot-noise ratio of 1.525.	✅ pass
test_line_reads	tlo_calib returns the assumed and actual shot noise, their ratio, and the electronic floor in the assumed unit.	✅ pass
test_loss_signature	The fitted transmittance carries an apparent extra loss of 10*log10(ratio) dB, 3.0103 dB at a factor two.	✅ pass
test_monitor_reads	The normalised zero-signal variance is 1.05 when honest and 0.85 at ratio 1.25, the electronic floor unmoved.	✅ pass
test_range_moves	tlo_range is alpha*sqrt(ratio): a half-range fixed in photocurrent moves in shot-noise units.	✅ pass