ClickDps

Phase 5b: DPS QBER derived from hardware, not pinned

20/20 passed in 4963ms

Test	What it does	Result
test_afterpulsing	A 5% afterpulse, landing on either detector equally, lifts the click rate and drags the QBER toward 1/2 without destroying the key.	✅ pass
test_chunk_invariance	Every draw is a pure function of (seed, stage, slot index), so chunk sizes two orders apart give identical records and scalars.	✅ pass
test_curve_discards	A carrier released during hold-off is lost, not deferred: a two-gate curve under a 50 ns hold-off reproduces the afterpulse-free digest, while the flat coin charges its whole probability at any hold-off.	✅ pass
test_curve_reduces	A one-entry afterpulse curve at zero hold-off reproduces the flat coin's record digest, counts and QBER.	✅ pass
test_curve_rejects	An empty curve, an entry outside [0, 1) and a curve summing to 1 or more are refused by name on either detector.	✅ pass
test_dark_count_floor	As $\eta \mu$ falls at nonzero dark probability the QBER climbs to 1/2, the 1e-2 dark probability exaggerated to reach the floor, so the shape is the anchor, not the number.	✅ pass
test_dead_time	Raising dead time strictly lowers the click rate and cannot lower the QBER.	✅ pass
test_delay_scales_walk	A two-period delay line doubles the integrated phase variance, and the QBER follows the closed form with the longer arm.	✅ pass
test_doubles_carry_blinding	Dark counts push the receiver toward doubles, as blinding does, and under squashing the QBER climbs with them.	✅ pass
test_doubles_squashed	Under squashing, doubles are a strict subset of the sifted key, the click total counts each twice, and the QBER lands on the squashing closed form.	✅ pass
test_holdoff_reaches	The flat coin's QBER ignores hold-off where a 20-gate curve of the same 3% decays to the bare floor, so the shipped QBER is pessimistic, 57% relative at 160 ns.	✅ pass
test_linewidth_raises_qber	QBER rises with linewidth and tracks the finite-flux form from 0 to 50 MHz, a walk ${\sigma }^2=2\pi \mathrm{\Delta }\nu d/R$ from 0 to 0.31 rad^2.	✅ pass
test_memory_exact	An independent serial reference on the engine's Threefry stages agrees on all seven counters to the integer across five hold-off, afterpulse and curve configurations.	✅ pass
test_phase_walk_variance	The differential phase variance is the quadrature pipeline's Wiener walk $2\pi \mathrm{\Delta }\nu d/R$, leaving visibility $V{e}^{-{\sigma }^2/2}$.	✅ pass
test_recovery_refused	A partial-efficiency recovery ramp is refused naming bias current, kinetic inductance, the eta/dark/response split and latching, while zero and None run as the gated detector.	✅ pass
test_seed_decorrelates	A second seed reproduces the same physics while differing bitwise.	✅ pass
test_split_memory	Hold-off and afterpulse rate split per detector as eta and dark do: equal values reproduce the pooled run bit for bit, and either on the pi-phase detector alone moves only its count.	✅ pass
test_thread_invariance	The stateful dead-time and afterpulse scan runs in index order, so 1 and 8 workers agree bitwise.	✅ pass
test_visibility_law	With no dark counts and a noiseless laser the QBER over a visibility sweep lands on $e=(1-V)/2$ and on the finite-flux form.	✅ pass
test_zero_noise	With no dark counts, linewidth or dead time the dim port is empty: QBER is exactly 0.0 at sift rate $1-e^{-\eta \mu }$.	✅ pass