ClickPair

Phase 5b: Bob's two detectors, split in efficiency and dark rate

9/9 passed in 945ms

Test	What it does	Result
test_basis_mismatch	With no dark floor or misalignment the basis-keyed detectors partition the clicks with no bit read wrong, the mismatch showing only as imbalance on basis_model, and with both restored the pooled QBER sits on the matched model.	✅ pass
test_basis_symmetric	run_basis with detector 1 given detector 0's numbers reproduces every count and float row, and at sifting 1 the two detectors sum to the clicks plus the doubles.	✅ pass
test_dark_mismatch	Dark counts on one detector lift its click rate past its partner's, give it over twice the errors, and move the QBER from 0.0301 to 0.0526, eleven binomial widths off the symmetric form.	✅ pass
test_efficiency_mismatch	With contrast and a dark floor each detector's click rate and lone errors land on slot_model in a four-sigma binomial band, and the pooled QBER on pair_model, not the symmetric form.	✅ pass
test_error_split	errors_d0 and errors_d1 are each detector's lone errors, so what they omit is the squashing coin's share, within [0, doubles] and half of them.	✅ pass
test_model_reduces	At equal detectors the bit-averaged pair_model returns qber_model to 1e-15 over flux, contrast, dark probability and phase spread.	✅ pass
test_rejects_nonsense	A zero or above-unit efficiency and an always-firing gate are refused as eta_d1 and dark_d1, not under their partner's names, by the DPS and basis-keyed runs alike.	✅ pass
test_split_is_invisible	At perfect contrast with no dark counts an efficiency mismatch makes no errors, so the QBER reads 0 and only detector 1 answering a third of the gates shows it.	✅ pass
test_symmetric_unchanged	Naming detector 1 its partner's efficiency and dark probability reproduces records, counts and floats over the flat path, phase walk, dead-time scan and timing convolution.	✅ pass