TwoAnalyses

The two rates side by side, the ideal limit they share and the loss where they part

8/8 passed in 2ms

Test	What it does	Result
test_bound_order	The loss-tolerant rate is never below the standard one across 30 flaws and 24 losses.	✅ pass
test_clamped_floor	Both rates clamp at zero with the fourth slot still reporting: a phase error above 0.48, a coin of exactly 0.5.	✅ pass
test_crossover	The two analyses stop at 28.766343 and 57.775328 dB at tilt = -delta and at 29.659930 and 57.843443 at tilt = 0, gaps of 29.008984 and 28.183513, against a shared 57.855244 dB at zero flaw.	✅ pass
test_flaw_sweep	At 20 dB the loss-tolerant rate falls by 1.1491x from zero flaw to 0.126 rad where the standard one falls by 22.096x.	✅ pass
test_ideal_bb84	At zero flaw both rates equal bb84_rate to 1e-15 and the phase error equals the bit error to 2e-16.	✅ pass
test_loss_sweep	The loss-tolerant rate falls a decade per 10 dB, 2.3991163e-01 at 0 dB against 2.3902135e-04 at 30, where the standard one is exactly zero by 30.	✅ pass
test_not_conflatable	The two fourth slots are different quantities, a coin imbalance against a phase error, and the bound built from the coin sits above the exact value.	✅ pass
test_shape	The loss-tolerant rate is bb84_rate's Shor-Preskill shape on this engine's observables to 1e-15.	✅ pass